Autofluorescence enhancement for label-free imaging of myelinated fibers in mammalian brains

Analyzing the structure of neuronal fibers with single axon resolution in large volumes is a challenge in connectomics. Different technologies try to address this goal; however, they are limited either by the ineffective labeling of the fibers or in the achievable resolution. The possibility of discriminating between different adjacent myelinated axons gives the opportunity of providing more information about the fiber composition and architecture within a specific area. Here, we propose MAGIC (Myelin Autofluorescence imaging by Glycerol Induced Contrast enhancement), a tissue preparation method to perform label-free fluorescence imaging of myelinated fibers that is user friendly and easy to handle. We exploit the high axial and radial resolution of two-photon fluorescence microscopy (TPFM) optical sectioning to decipher the mixture of various fiber orientations within the sample of interest. We demonstrate its broad applicability by performing mesoscopic reconstruction at a sub-micron resolution of mouse, rat, monkey, and human brain samples and by quantifying the different fiber organization in control and Reeler mouse's hippocampal sections. Our study provides a novel method for 3D label-free imaging of nerve fibers in fixed samples at high resolution, below micrometer level, that overcomes the limitation related to the myelinated axons exogenous labeling, improving the possibility of analyzing brain connectivity.


Supplementary information 1. Raman measurements and analysis
As reported in the main text, we used Raman spectroscopy for probing the glycerol content within myelinated fibers and the surrounding tissue before and after deglycerolization. In order to do so, we first recorded the Raman spectrum of the medium solution used for preparing and mounting glycerolized samples (20% glycerol, 2% DMSO, 4% formaldehyde). Supplementary figure S1a and S1b show the comparison between our measurement and the known spectra of glycerol and DMSO, respectively (spectra obtained from https://sdbs.db.aist.go.jp/sdbs/cgi-bin/cre_index.cgi). The main glycerol peaks can be easily identified in the recorded spectrum, while no major contribution can be attributed to the presence of DMSO. Then, we selected three major glycerol Raman bands (550, 850, and 1465 cm -1 ) to be used for spectral projection: in fact, the "glycerol score" was calculated as the scalar product between each recorded spectrum and these three bands of the spectrum. Hence, a score = 1 means a perfect match with the spectral signatures of glycerol, while a score < 1 implies lower abundance (or no presence at all) of that compound. Figure S1: Glycerol and DMSO Raman spectra. (a) Overlaying Raman spectra of mounting medium (magenta) and literature glycerol spectra (black). (b) Overlaying Raman spectra of mounting medium (magenta) and literature DMSO spectra (black). Graphs were prepared using OriginPro 9.0 (www.originlab.com).

Compatibility with one photon imaging
The MAGIC protocol was developed to help the study of myelinated fibers with fluorescence microscopy techniques. In order to demonstrate the versatility of the protocol, we performed acquisitions with a

MAGIC protocol can provide details of myelinated axon substructures
To establish the possibility of studying the different organization of the myelin sheaths surrounding axons, we used a high magnification objective (Apo Plan Nikon 100× immersion oil objective) to detect the autofluorescence signal coming from them, as shown in Supplementary figure S4. Figure S4. Myelinated axon substructures. Image obtained with a 100x objective (λexc=488nm), Scale bar = 10 µm. The graph shows the intensity profile of the lines corresponding to the two axons (blue and red) of the image. Image and graph were prepared using Fiji (www.fiji.sc/Fiji).

MAGIC and Immunostaining
Representative images of human brain sections treated with MAGIC and stained with an anti-GFAP antibody are shown in Supplementary figure S5, and in Supplementary videos SV2 and SV3.
To perform the staining, the following protocol was applied. Brain sections previously treated with MAGIC protocol were permeabilized, incubating the samples for 45 min with PBS with 0.5% Triton X-100. After, samples were incubated with the primary antibody, an anti-GFAP antibody (abcam; ab7260) with a dilution of 1:200 in PBST 0.1% overnight at 4°C in the dark. The solution containing the primary antibody was removed and the sample was washed three times for 8 hours in PBST 0.1% solution. The day after, the sample was incubated with the secondary antibody, a donkey anti-rabbit IgG antibody conjugated with an alexa fluor 568 with a dilution of 1:200 (abcam, ab175470), in PBST 0.1% for 2 h at room temperature in the dark. The secondary antibody was then removed and the samples were washed three times with PBST 0.1% for 2h each in the dark at RT. Finally, the samples were washed twice with PBS for 10 min each at @RT, a coverglass was mounted on each section with transparent nailpolish, and TPFM imaging was performed. Figure S5. Anti-GFAP Immunostaining. Images showing fibers (in green; MAGIC) and astrocytes (in red; anti-GFAP antibody) in both white (top row) and grey matter (bottom row) of a human brain section. Scale bar = 100 µm. Images were prepared using Fiji (www.fiji.sc/Fiji).

Control and Reeler fiber analysis pre-processing
In order to perform the analysis on different regions of interest of the control mouse (Control) and the Reeler mouse (Reeler), a manual segmentation of the areas was performed using Fiji (http://fiji.sc/Fiji). The

Structure tensor analysis and orientation distribution functions evaluation
In order to compensate for the relative inclination of the brain sections with respect to the coronal image plane, the whole mesoscale stack mosaic of the hippocampus and its segmented ROIs were initially rotated along the frontal and longitudinal axes using Fiji. Next, image blurring was applied in the coronal plane so as to match the in-plane FWHMXY of the optical system PSF (FWHMXY = 0.692 μm) to the sagittal FWHMZ (2.612 μm). After these pre-processing operations, stack volumes were virtually decomposed into macro-voxels of (12, 12, 5) pixel size for obtaining a 5-μm tissue analysis resolution, and then downscaled within the image plane in order to set an isotropic spatial sampling period of 1 μm. As mentioned in the main text, a threshold of 85% non-zero voxels was applied to the separated 5-μm image sub-blocks with the aim to exclude the background and local regions associated with a scarce proportion of the brain tissue data. In practice, this macro-voxel selection step was instrumental in improving the accuracy of the resulting orientation estimates in boundary macro-voxels and in rejecting the contribution of spurious dark regions.
SV4 Video4: 3D rendering navigation of a representative stack of the human hippocampus after MAGIC.