Whole-Brain Single-Cell Imaging and Analysis of Intact Neonatal Mouse Brains Using MRI, Tissue Clearing, and Light-Sheet Microscopy

Felix A. Kyere; Ian Curtin; Oleh Krupa; Carolyn M. McCormick; Mustafa Dere; Sarah Khan; Minjeong Kim; Tzu-Wen Winnie Wang; Qiuhong He; Guorong Wu; Yen-Yu Ian Shih; Jason L. Stein

doi:10.3791/64096

JoVE Journal > Neuroscience

神经科学

Whole-Brain Single-Cell Imaging and Analysis of Intact Neonatal Mouse Brains Using MRI, Tissue Clearing, and Light-Sheet Microscopy

Published: August 01, 2022

doi:

10.3791/64096

Felix A. Kyere*^1,2, Ian Curtin*^1,2, Oleh Krupa^1,2, Carolyn M. McCormick^1,2, Mustafa Dere³, Sarah Khan^3,7, Minjeong Kim⁷, Tzu-Wen Winnie Wang^4,5,6, Qiuhong He^4,5,6, Guorong Wu³, Yen-Yu Ian Shih^4,5,6, Jason L. Stein^1,2

¹UNC Neuroscience Center,University of North Carolina, Chapel Hill, ²Department of Genetics,University of North Carolina, Chapel Hill, ³Department of Psychiatry,University of North Carolina, Chapel Hill, ⁴Center for Animal Magnetic Resonance Imaging,The University of North Carolina at Chapel Hill, ⁵Biomedical Research Imaging Center,The University of North Carolina at Chapel Hill, ⁶Department of Neurology,The University of North Carolina at Chapel Hill, ⁷Department of Computer Science,The University of North Carolina at Greensboro

Summary

This protocol describes methods for conducting magnetic resonance imaging, clearing, and immunolabeling of intact mouse brains using iDISCO+, followed by a detailed description of imaging using light-sheet microscopy, and downstream analyses using NuMorph.

Abstract

Tissue clearing followed by light-sheet microscopy (LSFM) enables cellular-resolution imaging of intact brain structure, allowing quantitative analysis of structural changes caused by genetic or environmental perturbations. Whole-brain imaging results in more accurate quantification of cells and the study of region-specific differences that may be missed with commonly used microscopy of physically sectioned tissue. Using light-sheet microscopy to image cleared brains greatly increases acquisition speed as compared to confocal microscopy. Although these images produce very large amounts of brain structural data, most computational tools that perform feature quantification in images of cleared tissue are limited to counting sparse cell populations, rather than all nuclei.

Here, we demonstrate NuMorph (Nuclear-Based Morphometry), a group of analysis tools, to quantify all nuclei and nuclear markers within annotated regions of a postnatal day 4 (P4) mouse brain after clearing and imaging on a light-sheet microscope. We describe magnetic resonance imaging (MRI) to measure brain volume prior to shrinkage caused by tissue clearing dehydration steps, tissue clearing using the iDISCO+ method, including immunolabeling, followed by light-sheet microscopy using a commercially available platform to image mouse brains at cellular resolution. We then demonstrate this image analysis pipeline using NuMorph, which is used to correct intensity differences, stitch image tiles, align multiple channels, count nuclei, and annotate brain regions through registration to publicly available atlases.

We designed this approach using publicly available protocols and software, allowing any researcher with the necessary microscope and computational resources to perform these techniques. These tissue clearing, imaging, and computational tools allow measurement and quantification of the three-dimensional (3D) organization of cell-types in the cortex and should be widely applicable to any wild-type/knockout mouse study design.

Introduction

Whole-brain imaging at single-cell resolution is an important challenge in neuroscience. Cellular-resolution brain images allow for detailed analysis and system-level mapping of brain circuitry and how that circuitry is disrupted by genetic or environmental risk factors for neuropsychiatric disorders, cellular behavior in developing embryos, as well as neural circuits in the adult brain¹^,²^,³. There are multiple histological methods that allow for high-resolution images of the reconstructed 3D brain; however, these techniques require expensive, specialized equipment, may not be compatible with immunolabeling, and the two-dimensional (2D) nature of some methods may lead to tissue damage and shearing during sectioning⁴^,⁵.

Recent advancements have provided an alternative approach for imaging entire brains that does not require tissue sectioning; they involve using tissue clearing to make brains transparent. Transparency is achieved in most tissue clearing methods by both removing lipids, as they are a major source of light scattering, and matching the refractive index (RI) of the object with the RI of the sample immersion solution during imaging. Light can then pass through the boundary between materials without being scattered⁶^,⁷^,⁸^,⁹.

Tissue clearing methods, such as iDISCO+, are often combined with rapid 3D imaging using single-photon excitation microscopy, such as LSFM⁶^,⁷^,¹⁰. Within transparent tissues labeled with a fluorophore, light-sheet fluorescence microscopy images sections by excitation with a thin plane of light¹¹. The main advantage of LSFM is that a single optical section is illuminated at a time, with all the fluorescence from the molecules within that section being excited, which minimizes photobleaching. Moreover, imaging an entire optical slice enables camera-based detection of that excited slice, increasing speed relative to point scanning¹². LSFM nondestructively produces well-registered optical sections that are suitable for 3D reconstruction.

While the iDISCO+ method allows for inexpensive tissue clearing within ~3 weeks, dehydration steps within the protocol may lead to tissue shrinkage and potential alteration of the sample morphology, thus affecting volumetric measurements⁶^,¹⁰. Adding a secondary imaging method, such as MRI, to be used prior to the tissue clearing procedure can measure the degree of tissue clearing-induced shrinkage across the sample. During the dehydration steps, differences in mechanical properties between gray and white matter may lead to nonuniform brain matter deformations, resulting in dissimilar tissue clearing-induced volume deformations between wild-type and mutant samples and may confound interpretations of volumetric differences in these samples¹⁰^,¹³. MRI is performed by first perfusing the animal with a contrast agent (e.g., gadolinium), followed by incubating the extracted tissue of interest in an immersion solution (e.g., fomblin) before imaging¹⁴. MRI is compatible with tissue clearing and performing LSFM on the same sample.

LSFM is often used to create large-scale microscopy images for qualitative visualization of the brain tissue of interest rather than quantitative evaluation of brain structure (Figure 1). Without quantitative evaluation, it is difficult to demonstrate structural differences resulting from genetic or environmental insults. As tissue-clearing and imaging technologies improve, along with decreased costs of storage and computing power, quantifying cell type localizations within the tissue of interest is becoming more accessible, allowing more researchers to include these data in their studies.

With over 100 million cells in the mouse brain¹⁵ and whole-brain imaging sessions that can generate terabytes of data, there is increased demand for advanced image analysis tools that allow accurate quantification of features within the images, such as cells. A host of segmentation methods exist for tissue-cleared images that apply thresholding for nuclear staining intensity and filter objects with predefined shapes, sizes, or densities¹⁰^,¹⁶^,¹⁷^,¹⁸. However, inaccurate interpretations of results can arise from variations in parameters such as cell size, image contrast, and labeling intensity. This paper describes our established protocol to quantify cell nuclei in the mouse brain. First, we detail steps for tissue collection of the P4 mouse brain, followed by a tissue clearing and immunolabeling protocol optimized from the publicly available iDISCO+ method¹⁰. Second, we describe image acquisition using MRI and light-sheet microscopy, including the parameters used for capturing images. Finally, we describe and demonstrate NuMorph¹⁹, a set of image analysis tools our group has developed that allows cell-type specific quantification after tissue clearing, immunolabeling with nuclear markers, and light-sheet imaging of annotated regions.

Protocol

All mice were used in accordance with and approved by the Institutional Animal Care and Use Committee (IACUC) at the University of North Carolina at Chapel Hill.

1. Mouse dissection and perfusion

NOTE: The following dissections were performed on P4 and P14 mice using a syringe. The volume of perfusion fluid will vary depending on the age of the animal.

Perfusion
CAUTION: Paraformaldehyde (PFA) is a hazardous chemical. Perform all perfusion steps in a chemical fume hood.
1. Prior to surgery, administer pentobarbital via intraperitoneal injection (100 mg/kg) and allow the anesthetic to take effect.
2. Once the animal has reached a surgical plane of anesthesia, use the toe-pinching response method to confirm unresponsiveness.
3. Make a lateral incision beneath the rib cage to expose the thoracic cavity of the animal.
4. Using curved, surgical scissors, carefully cut through the rib cage up to the collarbone on one side of the animal and make an identical cut on the opposite side, allowing the sternum to be lifted away, exposing the heart.
5. Without damaging the descending aorta, carefully trim any tissue connected to the heart before making a small incision on the right atrium to allow blood to flow out of the vasculature.
6. Using a syringe-based method, perfuse the mouse through the left ventricle with 10 mL and 7 mL of phosphate-buffered saline (PBS) for P14 and P4, respectively, with a perfusion rate of 1.5 mL/min through the system.
7. Once the blood is cleared, perfuse again with 10 mL of PBS + 4% PFA and 7 mL of PBS + 4% PFA for P14 and P4, respectively, at 4 °C with a perfusion rate of 1.5 mL/min to fix the animal.
  NOTE: Fixation tremors will be observed, and the animal will be stiff upon completion. If using MRI on samples, perfuse with similar volumes of PBS and PFA for each timepoint + 20% gadolinium-based MRI contrast agent in the PFA solution.
8. Remove the head using surgical scissors and drop-fix with PBS + 4% PFA for 24 h at 4 °C for complete fixation.
  NOTE: At this stage, the brain remains intact with the skull on (see Section 2). Pause point: brains can be stored for several months at this stage in PBS + 0.1% sodium azide at 4 °C.

2. MR-based gross brain structure imaging with intact skull and analysis

NOTE: The brain must be perfused and incubated in gadolinium as described above without being removed from the skull. All MRI occurs before removal of the brain from the skull to avoid unintended tissue loss during dissection. Imaging with an intact skull also provides support to the brain in the sample holder (i.e., syringe) during sample preparation and imaging.

Sample preparation
NOTE: The following steps are optimized for P4 and P14 mouse brain samples. The syringe size needed will depend on the physical size of the sample.
1. If performing MRI on the sample, remove the skin from the skull and incubate in PBS + 3% gadolinium for 23 days at 4 °C before imaging¹⁴. After 23 days, rinse the samples quickly in PBS.
2. Use 5 mL syringes to create a sample holder for MRI, using syringe pistons to close each end of the holder made with the syringe²⁰. Use plastic pieces to hold the skull tightly in place in the holder (Figure 2A). Remove the markings on the syringe with ethanol to prevent artifacts upon imaging.
3. Securely place the skull in the sample holder and fill with an immersion solution compatible with MRI (see the Table of Materials). Close the holder and remove all the air bubbles using a syringe.
  NOTE: Pause point: The skull may be stored in the immersion solution for several months before imaging.
Gross brain structure imaging (MRI)
1. Image the samples with a 9.4T/30 cm horizontal-bore, animal MRI system using a 15 mm volume coil and a spin-echo-based sequence with the following parameters: Spatial resolution: 60 µm x 60 µm x 60 µm; total scan time: 7 h 12 min; time to echo (TE): 6.83 ms; repetition time (TR): 40 ms; excitation/refocusing flip angles: 90/180 degrees; image size:166 x 168 x 209 voxels; Field of view (FOV): 9.9 mm x 10.1 mm x 12.4 mm; bandwidth: 100,000 kHz.
Computational gross brain structure analysis
1. Remove the surrounding skull from raw MRI images by manually tracing the mouse brain using segmentation software. Next, apply the voxel-wise multiplication operation between the mask image and raw MRI image to generate the skull-stripped brain MRI image.
  NOTE: The output is a binary mask image where the intensity for brain voxels is set to 1 and 0 otherwise.
2. Apply rigid image registration (estimating only translation and rotation) using 'flirt' in the FSL package²¹^,²² to align the skull-stripped MRI image (moving image) to the corresponding light-sheet microscopy image (reference image).
3. Apply non-rigid registration (using 'SyN' in ANTS software²³) to find the point-to-point correspondences between the rigid-aligned MRI image in step 2.3.2 to the light-sheet image (same reference image in step 2.3.2).
  NOTE: The output includes the warped MRI image and the deformation field associated with the volume changes from MRI to the light-sheet images.
4. Calculate the Jacobian determinant on the deformation field generated in step 2.3.3, which quantifies the volume change in a 3 x 3 x 3 voxel local neighborhood.
5. Align skull-stripped images to the Allen Developmental Mouse Brain Atlas using deformable image registration.
  NOTE: The established spatial point-to-point correspondence allows automatic annotation of brain regions of interest in the new mouse image (Figure 2C-H).

3. Brain dissection from the skull

Make a midline incision along the top of the skull from the neck to the nose to expose the skull.
Expose the base of the skull by trimming away the remaining neck muscle and all other residual muscle.
Using sharp surgical scissors, carefully cut along the inner surface of the skull, taking care not to damage the brain by maintaining a gentle upward pressure while cutting with the sharp surgical equipment.
Use tweezers to peel the two cut halves of the skull away from the brain and carefully trim away excess fat attached to the brain.
Use a surgical scissors to trim any dura that connects the brain to the skull, and then use a spatula to gently remove the brain from the head.
Remove the brain, wash with PBS, and then swap to PBS with 0.1% sodium azide and keep at 4 °C for long-term storage.

4. Tissue clearing

NOTE: This protocol is adapted from the iDISCO+ protocol for P4 mice⁶, with minor changes. Some details may change for different time points/species/experiments). CAUTION: Methanol, dichloromethane (DCM), and dibenzyl ether (DBE) are hazardous chemicals. These tissue clearing steps are performed in a chemical fume hood.

Antibody validation
NOTE: Methanol compatibility of untested antibodies needs to be checked as they may be negatively affected by the harsh methanol washes required in the iDISCO+ protocol. For a list of antibodies that have been shown to work in iDISCO+, see the website²⁴.
1. Harvest 10 µm frozen sections of the PFA-fixed tissue of interest onto stereological slides.
2. Incubate the sections in 100% methanol for 3 h at room temperature.
3. Rehydrate in PBS before proceeding with standard immunohistochemistry protocols to determine whether the antibody shows the expected pattern of fluorescence after methanol washes. For positive control, use a slide not treated with methanol.
Buffer preparation
1. Prepare buffers according to the official iDISCO protocol. See the Table of Materials for composition of the buffers and other solutions used in this protocol.
Pretreatment
1. Dehydrate the sample with methanol/PBS series: 20%, 40%, 60%, 80%, 100%; 1 h each at room temperature.
  NOTE: Using PBS during dehydration helps prevent cracking of the samples in methanol washes.
2. Wash the sample in 100% methanol for 1 h, and then chill at 4 °C for 1 h before incubating overnight with shaking in 66% DCM/33% methanol at room temperature.
3. Wash the sample 2x in 100% methanol at room temperature, and then chill it at 4 °C.
4. Use fresh 5% H₂O₂ in methanol to bleach the sample overnight at 4 °C.
5. Rehydrate the sample with methanol/PBS series: 80%, 60%, 40%, 20%, PBS; 1 h each at room temperature and wash 2x for 1 h in PTx.2 at room temperature.
6. Incubate the sample in 1x PBS/0.2% TritonX-100/20% DMSO at 37 °C overnight.
7. Incubate the sample in 1x PBS/0.1% Tween-20/0.1% TritonX-100/0.1% Deoxycholate/0.1% NP40/20% DMSO at 37 °C overnight.
8. Wash in PTx.2 at room temperature for 1 h twice.
Immunolabeling
1. Incubate the samples in Permeabilization Solution at 37 °C for 2 days (~48 h).
2. Block the samples in Blocking Solution at 37 °C for 2 days (~48 h).
3. Incubate the samples with primary antibody in PTwH / 5%DMSO / 3% Serum at 37 °C for 4 days (~96 h) (e.g., rabbit(Rb) Brn2/POU3F2 mAb (1:100) and anti-Ctip2 rat(Rt) antibody (1:400) (Table of Materials).
4. Wash 3 x 1 h in PTwH. Wash for another 2 h in PTwH. Leave in the wash solution overnight at room temperature.
5. Incubate the samples with secondary antibody and a nuclear dye, such as TO-PRO-3, in PTwH / 3% Serum at 37 °C for 4 days (~96 h; e.g., goat anti-Rb(1:50) and (goat anti-Rt(1:200)) (Table of Materials).
6. Wash 3 x 1 h in PTwH Wash for another 2 h in PTwH. Leave in the wash solution overnight at room temperature.
Clearing
1. Dehydrate in methanol/PBS series-20%, 40%, 60%, 80%, 100%-1 h each at room temperature. Incubate for 3 h, with shaking, in 66% DCM / 33% methanol at room temperature.
  NOTE: The sample can be left overnight at room temperature immediately after dehydration in 100% methanol.
2. Incubate in 100% DCM for 15 min twice at room temperature (with shaking) to wash the MeOH.
3. Incubate in dibenzyl ether (DBE) without shaking at room temperature. Ensure that the tube is filled almost completely with DBE to prevent oxidation of the sample. Finish mixing the solution by inverting a couple of times before imaging.

5. Light-sheet imaging

NOTE: iDISCO tissue-cleared brains were imaged with a light-sheet microscope, equipped with a 2X/0.5 NA objective, a complementary metal oxide semiconductor camera, and microscope operating and image acquisition software at 0.75 x 0.75 x 4 µm/voxel for the P4 timepoint as this allowed single-cell resolution within the cortex (Figure 3A,B).

Sample mounting
1. Carefully mount the sample in the correct sample size holder such that the sample is oriented with the z-dimension no more than 5.2 mm in depth due to the rated working distance of the light-sheet microscope (5.7 mm minus 0.5 mm safety margin)²⁵.
2. Place holder in the sample cradle with the screw of the holder at a 45° angle to the cradle supports (Figure 1B). Position the cradle so that the light path is perpendicular to the sample (Figure 1C).
Imaging parameters
1. Set the zoom body on the microscope to 4x magnification or higher yielding 0.75 µm/pixel.
  NOTE: Single-cell computational analyses on P4 light-sheet images can be done with any commercially available light-sheet microscope that allows resolution of 0.75 x 0.75 x 4 µm/voxel or higher. A lower resolution is sufficient for brains at later time points in which nuclei are more sparsely distributed.
2. In the image acquisition software, select a single light-sheet with an NA = ~0.08 (9 µm thickness/4 µm z step).
  NOTE: This setting combined with horizontal dynamic focusing allows whole-brain imaging at a single-cell resolution of a mouse brain within a reasonable time. For a postnatal day 4 (P4) brain, image acquisition time is estimated to be 11-15 h for three channels depending on the size of the brain.
3. To ensure axial resolution is maintained along the width of the image, select Horizontal Dynamic Focusing and apply the recommended number of steps depending on the laser wavelength. For a whole P4 mouse brain, set the horizontal dynamic focusing to 11. Adjust Fine Focus for each channel with respect to the registration channel.
  NOTE: Here, TO-PRO-3 channel (647 nm) is registered to the Allen Developmental Mouse Brain Atlas as this labels all nuclei.
4. Adjust the laser power per channel with respect to the channel properties.
  NOTE: Longer wavelengths require higher laser power compared to shorter wavelengths. For instance, the 780 nm needs to be imaged at a high laser power (70% – 75%) and low exposure (50 ms), while the 647 nm channel requires an average laser power (40% – 45%) and low exposure (50 ms).
5. Adjust the Light-Sheet Width 到 ~50% to ensure that the sheet power is distributed optimally in the y dimension for this sample size.
  NOTE: In combination with the horizontal dynamic focusing, a 50% sheet width provides an average distribution of power across the image with a reduced risk of photobleaching²⁵.
6. Set Number of Tiles in respect to the size of the sample with a recommended Overlap of 15% between tiles, and capture images for each channel sequentially for each stack at a given tile position.

6. Image processing using NuMorph

NOTE: The NuMorph pipeline has three main parts for 3D image analysis: preprocessing, analysis, and evaluation. These parts have been organized into NMp_template.m, NMa_template.m, and NMe_template.m, respectively, which are discussed below. Additionally, NM_setup.m is added to download and install software packages needed for NuMorph to run smoothly. NM_samples.m also provides a template to input image acquisition information.

NuMorph setup
1. Download and install the conda environment manager for Linux²⁶. Download and install NuMorph¹⁹ image processing tools.
2. On the command line, run Matlab. Run NM_setup.m from NuMorph to download and install image analyses software packages needed for analyses.
  NOTE: This step ensures the conda environment is set up properly as well as ensures all tools and add-ons needed for Matlab to run each of the three pipelines are downloaded and installed correctly. Most notable here are Elastix for running registration and 3D-Unet for cell detection and counting.
Specify sample names, input and output directories, channel information, and light-sheet imaging parameters by editing the file NM_samples.m.
NOTE: Here, it is recommended to double-check to make sure the right information, especially image input directory, is specified properly. Errors are not usually called here until running subsequent steps.
Image preprocessing
1. Intensity adjustment
  1. In the NMp_template, set intensity adjustment = true.
    NOTE: Set to true if intensity adjustment is required. If not, set intensity adjustment = false. There is also an option to use 'update' to overwrite previous adjustment parameters.
  2. Set use processed images = false when working with a new set of images. Otherwise, indicate any previously saved image datasets in the output directory (e.g., "aligned", "stitched") to use for subsequent processing steps.
    NOTE: This option is provided in the case where input images have already been preprocessed. In this case, preprocessed images will be used as input images and the option will be set to the name of the subdirectory in the output directory.
  3. Set save images = true.
    NOTE: Using this option ensures processed images are saved in the output directory; otherwise, only parameters will be calculated and saved.
  4. Set save samples = true.
    NOTE: This option ensures sample results are saved for each major step.
  5. Set adjust tile shading = basic to apply shading correction using the BaSiC algorithm²⁷ or manual to apply tile shading correction using measurements from the light-sheet microscope at specific light-sheet widths.
    NOTE: This option corrects for the uneven illumination along the y dimension caused by the shape of the sheet waist.
2. Image channel alignment
  1. In NMp_template, set channel alignment = true. Set this option to true if channel alignment is required. If not, set to false. Set channel alignment method to either translation (rigid) or elastix (nonrigid).
    NOTE: The translation method utilizes rigid 2D registration approaches in aligning multiple channels while the elastix method utilizes non-rigid B-splines²⁸ to correct for rotational drift, which may occur during long image acquisition¹⁹.
3. Iterative image stitching
  1. In NMp_template, set stitch images = true.
    NOTE: Set this option to true if stitching is required.
  2. Set sift refinement = true.
    NOTE: This option is used to further refine translation in xy using the Scale Invariant Feature Transform²⁹.
  3. Set load alignment params = true.
    NOTE: This option utilizes the channel alignment translations during stitching. This option is recommended with multichannel imaging. Otherwise, set to false.
  4. Set overlap = 0.15 to match tile overlaps during imaging.
4. To run any of these preprocessing steps, run the following in Matlab outside NMp_template environment:
  1. Specify sample name. Set config = NM_config(process,sample).
  2. Run any of the preprocessing steps by specifying NM_process(config,stage) and specify the stage using intensity, align, or stitch for any of the processes. Check the output directory for output files for each of the stages (Figure 3 and Figure 4).
Image analysis
1. Before NuMorph
  1. Start with a 3D atlas image and an associated annotation image that assigns each voxel to a particular structure.
    NOTE: The P4 Allen Developmental Mouse Brain Atlas generated from the MagellanMapper³⁰ is used here.
  2. Align both the atlas image and annotation file to ensure they match correctly in the right orientation.
2. Within NuMorph
  NOTE: Now that the atlas and its annotations are aligned correctly, the files have to be "munged" or processed within NuMorph so that they can be saved for later use. To do this, use the munge_atlas function to specify inputs as shown below.
  1. Specify Atlas_file: (string). Provide the full path to the atlas file.
  2. Specify Annotation_file: (string). Provide the full path to the associated annotations.
  3. Specify Resolution: (1×3 numeric). Specify the atlas y,x,z resolution as micron per pixel.
  4. Specify Orientation: (1 x 3 char). Provide the atlas orientation and ensure it matches the setup of the sample in the cradle (anterior(a)/posterior(p),superior(s)/inferior(i),left(l)/right(r)).
  5. Specify Hemisphere: Specify which brain hemisphere was imaged ("left", "right", "both", "none").
  6. Specify out_resolution:(int). Specify the isotropic resolution of atlas output in microns. (default: 25).
  7. Run the command "munge_atlas(atlas_file, annotation_file, resolution, orientation, hemisphere)" to generate munged annotations in /data/annotation_data and a copy of the atlas image in /data/atlas.
  8. Read the Matlab structure and atlas file to verify both files are munged correctly in the right orientation.
    NOTE: An additional cell classification step can be performed to quantify cell-types based on co-localization of immunolabeled protein markers.
3. Resampling
  1. In NMa_template, set resample images = true, if performing image registration to reference the atlas or to generate downsampled volumes of high-resolution datasets.
    NOTE: The NMa_template.m will be used to set the parameters for resampling, registration, nuclei detection, and cell counting.
  2. Set resample resolution to match the atlas.
    NOTE: Here, 25 µm³/voxel isotropic resolution is used because the reference atlas is at this resolution.
  3. Specify the channel number to be resampled using resample channels = [ ].
    NOTE: Here, channel number is set to match the nuclear channel. If this option is empty, only the registration channel will be resampled.
4. Registration
  1. In NMa_template, set register images = true. Set to true if registration is required. If not, set registration = false.
  2. Specify the atlas file to match the file in the atlas directory.
  3. Set registration parameters = default.
    NOTE: This option utilizes an affine followed by B spline transformation to estimate spatial correspondence. Otherwise, define a new set of registration parameters via Elastix in /data/elastix_parameter_files/atlas_registration.
  4. Set save registered images = true.
    NOTE: Output files from registration and resampling can be downloaded and visually inspected in Matlab or other visualization tools such as FIJI³¹.
5. Nuclei Detection, Cell Counting and Classification
  NOTE: Errors occurring here may be due to not specifying the annotations file correctly or not matching the age of the sample with the right annotation.
  1. In NMa_template, set both count nuclei and classify cells = true.
  2. Set count method = 3dunet.
    NOTE: This option allows the use of the trained 3D-Unet model¹⁹. Otherwise, select Hessian that utilizes the Hessian blob detection method.
  3. Set min_intensity to define a minimum intensity threshold for detected objects.
    NOTE: An appropriate threshold is determined empirically based on the signal-to-noise ratio of nuclear labeling.
  4. Set classify_method to either threshold, which is based on an unsupervised fluorescence intensities at centroid positions or svm, which models a supervised linear Support Vector Machine (SVM) classifier.
    NOTE: This step will classify all the detected cells into four major classes with 3-channel imaging. With this protocol, Ctip2⁺/Brn2^–, Ctip2^–/Brn2⁺, Ctip2^–/Brn2^–, and Outliers are generated.
6. Analysis steps
  1. Specify sample name. Set config =NM_config(analyze,sample).
  2. Run any of the analysis steps by specifying NM_analyze(config,stage) and specify the stage using resample, register, count, or classify. Check the output directory for output files for each of the stages (Figure 5).
7. Cell-type classification and group analysis
  1. In NMe_template, set update = true and overwrite all previous statistics calculated.
    NOTE: The NMe_template.m provides the option to perform cell-type group analysis across brain regions of the same brain being analyzed.
  2. Set compare_structures_by to either index to compare by all unique annotations or table to compare structures according to the table.
  3. Set the template_file, which specifies all the possible structure indexes and must exist in /annotations.
  4. Set structure_table and specify structures to evaluate.
  5. Specify cell counting and cell-type classification as described in NMa_template.m.
  6. Set compare_groups to specify groups to compare.
  7. Set paired to either true or false to either perform paired t-test or two-sample t-test.
8. Run analysis.
  NOTE: To perform this step, run the following in Matlab outside NMe_template.m environment.
  1. Specify sample name.Set config =NM_config(evaluate,sample).
  2. Run the analysis step by specifying NM_evaluate(config,stage) and specify the stage. Check the output directory for output files for the group analysis.

Representative Results

As the iDISCO+ protocol introduces significant tissue shrinkage, which is easily noticeable by eye (Figure 2B), we added an MRI step to this pipeline prior to tissue clearing to quantify the shrinkage induced by tissue clearing. The workflow starts with removal of the non-brain tissue from the MR image (Figure 2C). Next, a rigid transformation (3 translation and 3 rotation angles) is applied to align the MR image to the light-sheet image (Figure 2D). By doing so, we observed a 60% total volume loss induced by the tissue clearing procedure (Figure 2E), which was consistent with volume change estimates by eye (Figure 2B). This result is quite different from a previous report where a shrinkage of 5%-10%¹⁰^,³² was reported. The difference in shrinkage may be caused by the difference in animal age between the two studies (P4 vs. adult brains) or due to time differences in the methanol wash in the clearing step of the iDISCO+ protocol (16 h vs. 2 h). To map degrees of tissue shrinkage across different regions of the brain, we further deployed a deformable image registration where the light-sheet image is used as the reference. The registered MR image is shown in (Figure 2F). The estimated volume change due to the tissue clearing procedure is color-coded, where a larger degree of shrinking is observed in the cortex as compared to other brain regions (Figure 2G).

To measure volumes of brain regions, we performed segmentation of the MRI through registration to an annotated atlas. Segmentation of MR-imaged brains was considered successful if the reference atlas overlay closely matched the anatomical boundaries identified by differences in contrast in the raw MRI images (Figure 2H). The overlay during registration is dependent on good sample preparation (Figure 2A), image acquisition, and skull stripping. The skull stripping step is crucial for good registration results as the registration will attempt to warp the reference image from the atlas to the entire MR image, including any skull that remains in the image. Skull inclusion in registration can distort the results and produce an incorrectly registered 3D volume rendering of the segmentation. The final 3D volume rendering can then be visualized using ITK-SNAP³³ (Figure 2H). This method is used to assess gross brain volume differences across sample groups. The cellular basis underlying volume differences discovered with MRI can be pursued using cell counting with tissue clearing, lightsheet microscopy, and NuMorph.

The goal of tissue clearing and analysis is to assess the contributions of different cell types to differences across experimental conditions (e.g., genotype, environmental exposure) by counting individual nuclei using NuMorph. Tissue clearing using the iDISCO+ protocol and neuronal layer specific nuclei markers resulted in clearly defined cell groups of upper and lower layer neurons in the isocortex (Figure 3).

Cell counting using NuMorph is dependent on successful preprocessing steps involving intensity adjustment, channel alignment, and stitching (Figure 4A). Errors occurring in any of these steps can result in improper stitching (Figure 4B,C). For instance, improper image acquisition can result in images with in-focus and out-of-focus patterns during preprocessing (Figure 4C). To count nuclei from specific brain regions, the stitched images are annotated using a publicly available atlas. We registered our stitched images to a corrected version of the P4 Allen Developmental Mouse Brain Atlas³⁴. Here, the stitched images were downsampled from a high resolution (0.75 x 0.75 x 4 µm³/voxel) to a lower resolution (25 µm³/voxel isotropic resolution) to match the resolution of the atlas. Matching the resolution of the atlas ensures correct registration of the images to the atlas and provides regional annotations in the acquired images (Figure 5A).

To perform specific cell-type specific counting, we labeled upper layer neurons expressing Brn2, lower layer neurons expressing Ctip2, and all nuclei with TO-PRO-3 (Figure 3B). Imaging is performed at sufficient spatial resolution to separate individual nuclei (0.75 x 0.75 x 4 µm³/voxel). The centroids of nuclei are detected with a trained 3D-Unet model in NuMorph (Figure 5B,C). We detected ~12 × 10⁶ total nuclei that were To-Pro-3⁺ in the isocortex, including ~2.6 × 10⁶ Brn2⁺ and 1.6 × 10⁶ Ctip2⁺ nuclei. We also detected ~3.7 × 10⁶ and 2.9 × 10⁶ To-Pro-3⁺ total nuclei in the basal ganglia and hippocampal allocortex, respectively. Approximately 1.5 × 10⁶ and <1 × 10⁶ Ctip2⁺ cells were counted in the basal ganglia and hippocampal allocortex, respectively, although a negligible number of Brn2⁺ cells were detected (Figure 5D). The total nuclei counts in the isocortex and hippocampal allocortex combined (~15 million cells) are similar to previously reported total cell counts in the cerebral cortex of the mouse¹⁵. These results demonstrate the utility of MRI imaging, iDISCO+ tissue clearing methodology, and NuMorph computational analysis to reveal volumetric, total cell count, and cell-type specific counts underlying differences in mouse brain structure across experimental groups.

Figure 1: Overview of whole-brain single-cell analysis of intact neonatal 3D tissue cleared mouse brains. (A) Overview of iDISCO+ tissue clearing, light sheet imaging, and 3D image analysis. (B) Images showing correct sample mounting on the light-sheet microscope. (C) Cartoon showing sample mounting relative to light-sheet path. Abbreviation: ab = antibody. Please click here to view a larger version of this figure.

Figure 2: Gross brain structure measurements with MRI. (A) Image showing sample preparation for MRI. (B) Images of representative P4 brains before and after iDISCO+ tissue clearing. Scale bar = 5.0 cm. (C) Raw MR image with skull attached at 60 x 60 x 60 µm³. Scale bar = 800 µm. (D) Light-sheet image of mouse brain at 25 x 25 x 25 µm³. Total volume = 68.7 mm³. Note that a small section of the dorsal isocortex was unintentionally removed during dissection marked by arrowhead (white). Scale bar = 800 µm. (E) MR image of mouse brain after skull stripping and rigid registration. Total volume = 171.9 mm³. Insert (yellow) shows brain contour from the light-sheet image. Scale bar = 800 µm. (F) Registered MR image in reference to light-sheet image to assess deformation. Scale bar = 800 µm. (G) Voxel-wise brain mapping to assess volume changes between light-sheet image and MR image where blue and red indicate shrinkage and expansion from MR image to the light-sheet image, respectively. Overall there was a 60% volume loss, but some regions such as the isocortex showed greater shrinkage relative to others. (H) Left panel: Axial view of MRI image with registered segmentations from the Allen Developmental Brain Atlas overlaid. Right panel: 3D volume rendering of segmentation. Scale bar = 800 µm. Abbreviations: OB = olfactory bulb; Dpall = dorsal pallium/isocortex; Mb = midbrain; Cb = cerebellum. Orientation: R = Rostral; L = Lateral; D = Dorsal. Please click here to view a larger version of this figure.

Figure 3: Cellular resolution images and channel alignment. (A) Optical axial section of TO-PRO-3 (TP3) nuclear staining and immunolabeling for Ctip2 (lower layer neuron) and Brn2 (upper layer neuron) markers in P4 mouse brain. Scale bar = 1 mm. (B) Enlarged inset of cortical areas in A (box) showing correct channel alignment with expected localization of Brn2-immunolabeled upper layer and Ctip2-immunolabeled lower layer neurons. Scale bar = 50 µm. Please click here to view a larger version of this figure.

Figure 4: Image stitching examples. (A) Sample results from a 2D iterative, correctly stitched image. Scale bar = 1 mm. Insert shows an example of correct stitching zoomed in. Scale bar = 500 µm. (B) Sample results from a 2D iterative, incorrectly stitched image. Scale bar = 1 mm. Insert shows an example of incorrect stitching zoomed in (overhang). Scale bar = 100 µm. (C) Sample results from a 2D iterative incorrectly stitched image. Scale bar = 1 mm. Insert shows an example of incorrect stitching zoomed in (out-of-focus). Scale bar = 200 µm. Please click here to view a larger version of this figure.

Figure 5: Low resolution image registration, high resolution nuclei detection, and cell counting in the P4 mouse brain. (A) Left panel: An example of a z-slice from whole brain light sheet 3D images resampled to 25 µm isotropic resolution. Scale bar = 1 mm. Middle panel: Annotations from the Allen Developmental Brain Atlas (P4) registered to the microscopy image. Scale bar = 1 mm. Right panel: Overlay of the left and middle panels. Scale bar = 1 mm. (B) Example of an intensity adjusted stitched image in axial view. The boxed region is shown in (C). Scale bar = 1 mm. (C) Automated detection of nuclei (red). Scale bar = 50 µm. (D) Quantification of cell types in different brain regions of the P4 brain (Basal ganglia, Hippocampal Allocortex, and Isocortex). Red = Upper layer neurons labeled with Brn2, Green = Lower Layer neurons labeled with Ctip2, and Blue = all nuclei labeled with To-Pro-3 dye. Abbreviations: DPall = Dorsal pallium (isocortex); MPall = Medial pallium (hippocampal allocortex); CSPall = Central subpallium(classic basal ganglia). Please click here to view a larger version of this figure.

NuMorph Analysis Stage	Estimated Time
Intensity Adjustment	1 h
Channel Alignment	2 min
2D Iterative Stitching	12 h
Resampling	3 min
Registration	3 min
Cell Counting	5 h
Cell Classification	10 min

Table 1: NuMorph analysis times.

Discussion

Tissue clearing methods are useful techniques for measuring 3D cellular organization of the brain. There are a host of tissue clearing methods described in the literature, each with its advantages and limitations⁶^,⁷^,⁸^,⁹. The options for computational tools to analyze the cell types in the tissue-cleared images are relatively limited. Other available tools have been implemented to sparse cell populations in which segmentation is less difficult¹⁰^,³⁵ or have taken advantage of the tissue expansion¹⁶. This paper describes the preparation of a mouse brain for imaging with iDISCO+ tissue clearing and demonstrates NuMorph, a computational pipeline for processing and quantifying nuclei within structures of a mouse cortex captured by LSFM. It is designed to strike an appropriate balance between the length of imaging time, detection accuracy of cell nuclei, and computational resources. This pipeline contrasts with other currently available tools by reliably segmenting all nuclei, rather than sparse populations of labeled cells¹⁰^,³⁶^,³⁷. We have previously shown that cell detection accuracy is high, with significantly shorter imaging times and much reduced acquired data sizes for whole-hemisphere acquisition compared to other methods¹⁹. NuMorph addresses a number of important challenges for the analysis of multi-channel light-sheet images, such as providing tools for channel alignment, correcting for stage movements with a stitching tool, and correcting signal brightness in the image tiles. The workflow presented in this paper is useful for the broader scientific community and accessible for all groups in research institutions. An overview of the process can be seen in Figure 1.

There are several critical steps throughout this process from the mouse perfusion to the computational analysis that can affect the downstream quality of images, quantification, and results. It is essential that the perfusion is performed such that all the blood is removed from the brain, as the blood vessels have a natural autofluorescence that will affect the intensity in the images and will require adjustments in the pipeline, adversely affecting the results. The duration of the PFA fixation is also crucial, as leaving the brains submerged in 4% PFA for longer than 24 h may result in 'overfixing' of the sample and lead to cracking of the brain, which becomes brittle during the iDISCO+ process. The iDISCO protocol described above has been slightly modified from the official protocol found on the website²⁴. One important modification is the addition of PBS to the methanol/PBS series, instead of water in the original protocol. This is to prevent cortical cracking in the embryonic and early postnatal brains, which likely occurs because of the incomplete development of glial cells, axonal projections, and extracellular matrix. Depending on the tissue used in this protocol, modifications may have to be made at every step of the iDISCO+ protocol, such as altering antibody concentrations and increasing incubation times of larger tissue samples, to optimize the ideal parameters to capture the marker of choice and analyze it accurately.

Tissue shrinkage is known to occur during the iDISCO+ tissue clearing process, which may then alter volumetric measurements downstream⁶^,¹⁰. Alternate methods, such as MRI conducted prior to tissue clearing, may be used to determine the extent of any change in tissue size in controls versus experimental animals to determine whether any volumetric differences are due to studied phenotype and not from the tissue clearing process. Heterogeneous changes in the volumes of the cortical regions of the mutants may differ depending on the regional cellular content, as the gray and white matter of the brain may be differentially affected by the methanol dehydrations. The MRI data can be used to determine whether tissue shrinkage is non-uniform throughout the subregions of the brain (Figure 2G), and any changes incurred by the clearing protocol can be adjusted for in the downstream computational analysis. Moreover, the iDISCO+ process uses harsh methanol washes to dehydrate the sample, which may result in damage or alterations in certain antigens on the nuclei surfaces. It is recommended that any antibodies being used are first tested on brain tissue sections that have been washed for ~3 h in 100% methanol to ensure that the antibody is compatible with the iDISCO+ protocol.

For big datasets that require long processing times, we recommend using the no-window version of MATLAB in Linux that allows strategies such as implementing the "nohup" command, which will continue running a process even after the connection to a server is lost. Further, NuMorph requires high computing power, so we recommend at least a 10 GB memory for analyses. We analyze the samples using a Linux workstation running CentOS 7, which is equipped with a 2.6 GHz 14-core processor, 8 x 64 GB DDR4 2400 LRDIMM memory, 4 x 11 GB GPUs, and 2 x 4TB external SSDs. The raw light-sheet data used here was 620 GB and the memory requirement for the processes was 10 GB. Here, we have included a table with the estimated times to complete each step of the NuMorph analyses as seen in Table 1.

The image preprocessing steps involve intensity adjustment in the xy dimension for each z-plane, channel alignment, and stitching. The image intensity is adjusted to account for differences in intensity between tiles and uneven illumination along the y-dimension. The difference in intensity occurs because of inherent technical properties of the light-sheet as it images between tiles²⁵. Next, channel alignment is performed to ensure image channels align correctly in a multichannel imaging. This step is recommended since each channel in multichannel imaging is acquired individually and subtle movements of the stage may cause sample drifts and result in spatial misalignment¹⁹. Finally, the tiles per z-plane are then stitched together to form a single image in each z-plane. In the end, there will be a stack of stitched images per channel representing the entire volume of the sample. The iterative image stitching is based on a custom method to organize all the 2D images per channel in a 3D volume of the sample¹⁹. The method first implements a pairwise z-correspondence in the axial direction to match tile regions that are overlapping in both horizontal and vertical direction. The final z-displacement of each tile is determined using the minimum spanning tree³⁸. 2D iterative stitching is performed in a stack starting at a location in the center of the stack with less background signal. The top left tile is placed first for each stitching iteration to prevent subtle shift of images in the z dimension.

We recommend running each of the preprocessing pipeline sequentially, especially when optimizing to correct for errors. The major problems often occur during the stitching steps. One problem that can arise is the incorrect image acquisition on the light-sheet. This problem may occur when the horizontal dynamic focusing is improperly set on the light-sheet and may result in alternating in-phase/out-of-phase patterns in the images (Figure 4B,C). Additional errors may arise if the sample was not tightly secured, and significant sample movement occurred during the acquisition. In these cases, accurate stitching may not be possible due to a lack of pairwise correspondence between adjacent tiles. Other potential errors can stem from the stitch start site. NuMorph determines the stitch start site roughly in the middle of the stack. However, when there is significant background in the starting slice site, errors may occur in the pairwise comparison in image tiles. It is recommended here to choose a start site that has a high signal-to-noise ratio.

The image analysis described here includes image resampling of stitched images from high resolution (0.75 x 0.75 x 4 µm³/voxel) to a lower isotropic resolution of 25 µm³/voxel. We resample the stitched images from the nuclear channel to register the stitched images to the Allen Developmental Mouse Brain Atlas in the next step³⁴. Cells are then counted for each channel in annotated regions of the images in reference to the atlas at a high resolution (0.75 x 0.75 x 4 µm³/voxel) using a trained 3D-Unet model¹⁹. This method can also be used to detect and count blob objects. When developing NuMorph, we tested segmentation accuracy relative to completely new images never seen during training that were from the same tissue clearing procedure, same microscope, and same nuclear label finding precision of 0.99 and recall of 0.94¹⁹. NuMorph accuracy has not yet been tested on different tissue clearing procedures, microscopes, or markers; hence, the accuracy of segmentation in other experimental designs is unknown. The default atlas in NM_setup.m is the adult Nissl-stained Allen Reference Atlas³⁹ with the third iteration of Allen Mouse Brain Common Coordinate Framework (CCFv3) as the custom annotation.

Although NuMorph effectively analyzes tissues, which are comparatively dense, such as the adult mouse cortex, more challenging experimental designs (such as embryonic mouse brains or highly dense structures such as the cerebellum) may require development of additional computational tools. Current community-based efforts are attempting to produce adequate annotation data for training of deep learning models used to segment these more difficult cases⁴⁰. Advances in light-sheet imaging technologies allow high throughput quantitative investigation of subcellular structures, with new approaches such as dual-view inverted selective plane illumination microscopy (diSPIM) leading to increased resolution and accuracy in cellular-dense brain regions⁴⁰^,⁴¹. As advancements in tissue clearing, imaging and computational tools involved in this research evolve, we hope that they will lead to wider usage of tissue clearing methods for quantitative analyses on how neuropsychiatric disorders can stem from alterations in brain structure induced by genetic or environmental risk factors.

Disclosures

The authors have nothing to disclose.

Acknowledgements

This work was supported by the NIH (R01MH121433, R01MH118349, and R01MH120125 to JLS and R01NS110791 to GW) and the Foundation of Hope. We thank Pablo Ariel of the Microscopy Services Laboratory for assisting in sample imaging. The Microscopy Services Laboratory in the Department of Pathology and Laboratory Medicine is supported in part by Cancer Center Core Support Grant P30 CA016086 to the University of North Carolina (UNC) Lineberger Comprehensive Cancer Center. The Neuroscience Microscopy Core is supported by grant P30 NS045892. Research reported in this publication was supported in part by the North Carolina Biotech Center Institutional Support Grant 2016-IDG-1016.

Materials

Bruker 9.4T/30 cm MRI Scanner	Bruker Biospec		Horizontal Bore Animal MRI System
Dibenzyl ether	Sigma-Aldrich	108014-1KG
Dichloromethane (DCM)	Sigma-Aldrich	270997-1L
Dimethyl sulfoxide (DMSO)	Fisher-Scientific	ICN19605590
Donkey serum	Sigma-Aldrich	S30-100ML
EVO 860 4TB external SSD
Fomblin Y	Speciality Fluids Company	YL-VAC-25-6	perfluoropolyether lubricant
gadolinium contrast agent (ProHance)	Bracco Diagnostics	A9576
gadolinium contrast agent(ProHance)	Bracco Diagnostics	0270-1111-03
GeForce GTX 1080 Ti 11GB GPU	EVGA	08G-P4-6286-KR
Glycine	Sigma-Aldrich	G7126-500G
Heparin sodium salt	Sigma-Aldrich	H3393-10KU	Dissolved in H₂O to 10 mg/mL
Hydrogen peroxide solution, 30%	Sigma-Aldrich	H1009-100ML
ImSpector Pro	LaVision BioTec		Microscope image acquisition software
ITK Snap			segmentation software
Methanol	Fisher-Scientific	A412SK-4
MVPLAPO 2x/0.5 NA Objective	Olympus
Paraformaldehyde, powder, 95% (PFA)	Sigma-Aldrich	30525-89-4	Dissolved in 1x PBS to 4%
Phosphate Buffered Saline 10x (PBS)	Corning	46-013-CM	Diluted to 1x in H₂O
Sodium Azide	Sigma-Aldrich	S2002-100G	Dissolved in H₂O to 10%
Sodium deoxycholate	Sigma-Aldrich	D6750-10G
Tergitol type NP-40	Sigma-Aldrich	NP40S-100ML
TritonX-100	Sigma-Aldrich	T8787-50ML
Tween-20	Fisher-Scientific	BP337 500
Ultramicroscope II Light Sheet Microscope	LaVision BioTec
Xeon Processor E5-2690 v4	Intel	E5-2690
Zyla sCMOS Camera	Andor		Complementary metal oxide semiconductor camera
Antibody			Working concentration
Alexa Fluor Goat 790 Anti-Rabbit	Thermofisher Scientific	A11369	(1:50)
Alexa Fluor Goat 568 Anti-Rat	Thermofisher Scientific	A11077	(1:200)
Rat anti-Ctip2	Abcam	ab18465	(1:400)
Rabbit anti-Brn2	Cell Signaling Technology	12137	(1:100)
To-Pro 3 (TP3)	Thermofisher Scientific	T3605	(1:400)

References

Dodt, H. U., et al. Ultramicroscopy: Three-dimensional visualization of neuronal networks in the whole mouse brain. Nature Methods. 4 (4), 331-336 (2007).
Hägerling, R., et al. A novel multistep mechanism for initial lymphangiogenesis in mouse embryos based on ultramicroscopy. The EMBO Journal. 32 (5), 629-644 (2013).
Tomer, R., Khairy, K., Keller, P. J. Shedding light on the system: studying embryonic development with light sheet microscopy. Current Opinion in Genetics & Development. 21 (5), 558-565 (2011).
Li, A., et al. Micro-optical sectioning tomography to obtain a high-resolution atlas of the mouse brain. Science. 330 (6009), 1404-1408 (2010).
Stoner, R., et al. Patches of disorganization in the neocortex of children with autism. The New England Journal of Medicine. 370 (13), 1209-1219 (2014).
Renier, N., et al. IDISCO: A simple, rapid method to immunolabel large tissue samples for volume imaging. Cell. 159 (4), 896-910 (2014).
Susaki, E. A., et al. Whole-brain imaging with single-cell resolution using chemical cocktails and computational analysis. Cell. 157 (3), 726-739 (2014).
Richardson, D. S., et al. Tissue clearing. Nature Reviews. Methods Primers. 1 (1), 84 (2021).
Richardson, D. S., Lichtman, J. W. Clarifying tissue clearing. Cell. 162 (2), 246-257 (2015).
Renier, N., et al. Mapping of brain activity by automated volume analysis of immediate early genes. Cell. 165 (7), 1789-1802 (2016).
Santi, P. A. Light sheet fluorescence microscopy: a review. The Journal of Histochemistry and Cytochemistry: Official Journal of the Histochemistry Society. 59 (2), 129-138 (2011).
Girkin, J. M., Carvalho, M. T. The light-sheet microscopy revolution. Journal of Optics. 20 (5), 053002 (2018).
Budday, S., et al. Mechanical properties of gray and white matter brain tissue by indentation. Journal of the Mechanical Behavior of Biomedical Materials. 46, 318-330 (2015).
Johnson, G. A., et al. High-throughput morphologic phenotyping of the mouse brain with magnetic resonance histology. NeuroImage. 37 (1), 82-89 (2007).
Herculano-Houzel, S., Mota, B., Lent, R. Cellular scaling rules for rodent brains. Proceedings of the National Academy of Sciences of the United States of America. 103 (32), 12138-12143 (2006).
Matsumoto, K., et al. Advanced CUBIC tissue clearing for whole-organ cell profiling. Nature Protocols. 14 (12), 3506-3537 (2019).
Fürth, D., et al. An interactive framework for whole-brain maps at cellular resolution. Nature Neuroscience. 21 (1), 139-149 (2018).
Chandrashekhar, V., et al. CloudReg: automatic terabyte-scale cross-modal brain volume registration. Nature Methods. 18 (8), 845-846 (2021).
Krupa, O., et al. NuMorph: Tools for cortical cellular phenotyping in tissue-cleared whole-brain images. Cell Reports. 37 (2), 109802 (2021).
Petiet, A., Delatour, B., Dhenain, M. Models of neurodegenerative disease – Alzheimer’s anatomical and amyloid plaque imaging. Methods in Molecular Biology. 771, 293-308 (2011).
Jenkinson, M., Beckmann, C. F., Behrens, T. E. J., Woolrich, M. W., Smith, S. M. FSL. NeuroImage. 62, 782-790 (2012).
Jenkinson, M., Bannister, P., Brady, M., Smith, S. Improved optimization for the robust and accurate linear registration and motion correction of brain images. NeuroImage. 17 (2), 825-841 (2002).
Avants, B. B., Epstein, C. L., Grossman, M., Gee, J. C. Symmetric diffeomorphic image registration with cross-correlation: evaluating automated labeling of elderly and neurodegenerative brain. Medical Image Analysis. 12 (1), 26-41 (2008).
. iDISCO method Available from: https://idisco.info/ (2022)
Ariel, P. UltraMicroscope II – A User Guide. University of North Carolina at Chapel Hill. , (2019).
. Anaconda Distribution – Anaconda documentation Available from: https://docs.anaconda.com/anaconda/ (2022)
Peng, T., et al. A BaSiC tool for background and shading correction of optical microscopy images. Nature Communications. 8, 14836 (2017).
Klein, S., Staring, M., Murphy, K., Viergever, M. A., Pluim, J. P. W. elastix: a toolbox for intensity-based medical image registration. IEEE Transactions on Medical Imaging. 29, 196-205 (2010).
Lowe, G. Sift-the scale invariant feature transform. International Journal. 2 (91-110), 2 (2004).
Young, D. M., et al. Whole-brain image analysis and anatomical atlas 3D generation using MagellanMapper. Current Protocols in Neuroscience. 94 (1), 104 (2020).
Schindelin, J., et al. Fiji: an open-source platform for biological-image analysis. Nature Methods. 9 (7), 676-682 (2012).
Velíšek, L. Under the (Light) sheet after the iDISCO. Epilepsy Currents / American Epilepsy Society. 16 (6), 405-407 (2016).
Young, D. M., et al. Constructing and optimizing 3D atlases from 2D data with application to the developing mouse brain. eLife. 10, (2021).
Yun, D. H., et al. Ultrafast immunostaining of organ-scale tissues for scalable proteomic phenotyping. bioRxiv. , (2019).
Silvestri, L., et al. Universal autofocus for quantitative volumetric microscopy of whole mouse brains. Nature Methods. 18 (8), 953-958 (2021).
Frasconi, P., et al. Large-scale automated identification of mouse brain cells in confocal light sheet microscopy images. Bioinformatics. 30 (17), 587 (2014).
Kruskal, J. B. On the shortest spanning subtree of a graph and the traveling salesman problem. Proceedings of the American Mathematical Society. American Mathematical Society. 7 (1), 48-50 (1956).
Wang, Q., et al. The allen mouse brain common coordinate framework: A 3D reference atlas. Cell. 181, 936-953 (2020).
Borland, D., et al. Segmentor: a tool for manual refinement of 3D microscopy annotations. BMC Bioinformatics. 22 (1), 260 (2021).
Kumar, A., et al. Dual-view plane illumination microscopy for rapid and spatially isotropic imaging. Nature Protocols. 9 (11), 2555-2573 (2014).

Play Video

PDF

DOI

DOWNLOAD MATERIALS LIST

Cite This Article

Kyere, F. A., Curtin, I., Krupa, O., McCormick, C. M., Dere, M., Khan, S., Kim, M., Wang, T. W., He, Q., Wu, G., Shih, Y. I., Stein, J. L. Whole-Brain Single-Cell Imaging and Analysis of Intact Neonatal Mouse Brains Using MRI, Tissue Clearing, and Light-Sheet Microscopy. J. Vis. Exp. (186), e64096, doi:10.3791/64096 (2022).