Detailed knowledge of the brain’s nerve fiber network is crucial for understanding its function in health and disease. However, mapping fibers with high resolution remains prohibitive in most histological sections because state-of-the-art techniques are incompatible with their preparation. Here, we present a micron-resolution light-scattering-based technique that reveals intricate fiber networks independent of sample preparation for extended fields of view. We uncover fiber structures in both label-free and stained, paraffin-embedded and deparaffinized, newly-prepared and archived, animal and human brain tissues – including whole-brain sections from the BigBrain atlas. We identify altered microstructures in demyelination and hippocampal neurodegeneration and show key advantages over diffusion magnetic resonance imaging, polarization microscopy, and structure tensor analysis. We also reveal structures in non-brain tissues – including muscle, bone, and blood vessels. Our cost-effective, versatile technique enables studies of intricate fiber networks in any type of histological tissue section, offering a new dimension to neuroscientific and biomedical research.

Sample preparation

Whole human brain (BigBrain) sections

Two human brains, obtained within 24 hours after death and without neurological disorders, were fixed in 4% formaldehyde, dehydrated in increasing alcohol series (80%, 90%, 96%, 100% ethanol for at least one week each), embedded in a solvent (chloroform) for one week, and embedded in 57-60°C paraffin solution for two months. After hardening, the brains were coronally cut into 20 µm-thin sections from anterior to posterior with a large-scale microtome (Leica SM2500 Microtome) and mounted on a 37°C heat plate in gelatin solution. The sections were placed in a decreasing alcohol series to remove the paraffin and then placed in a staining solution to highlight neuronal cell bodies: the sections of the first brain (30 years old, male) were stained with silver following the protocol of Merker (53); the sections of the second brain (71 years old, male) were stained with Cresyl-violet. After staining, the sections were again dehydrated in increasing alcohol series, mounted/sealed on glass slides, and archived. The silver-stained sections were from the second BigBrain data set (54), 3D-reconstructed with the same spatial resolution of 20μm isotropic such as the original Big BigBrain (24), archived for 32 years, and section no. 3452 was selected for evaluation (Fig. 1 and Fig. S6). The Cresyl-violet sections were archived for one and a half years, and sections no. 3301 (Fig. 4, left) and 2520 (fig. S7) were selected for evaluation. Body donors gave written informed consent for the general use of postmortem tissue used in this study for the aims of research and education. The usage is covered by a vote of the ethics committee of the medical faculty of the Heinrich Heine University Düsseldorf (#4863).

120-year-old myelin-stained brain section

The myelin-stained human brain section (Fig. 2B) comes from the brain collection of the Cécile and Oskar Vogt Institute for Brain Research, Heinrich Heine University Düsseldorf, Germany. The brain of a 25-year-old male was embedded in celloidin and stained according to Weigert’s iron hematoxylin myelin staining in 1904 (55).

Human hippocampus, cortex, and pathology brain sections

Four-millimeter thick specimens of formalin-fixed human brains were dehydrated in increasing ethanol steps (70% x2, 95% x2, 100% x3, 3.5hrs each step), cleared in xylene (3.5hrs x2), paraffin-embedded (3.5hrs x2), and sectioned into 5µm-thin sections. The sections were de-waxed and stained with different agents as indicated. The hippocampal sections in Fig. 2A and S1 were from an 89-year-old male with Alzheimer’s pathology, stained against microglia (CD163), Perl’s iron with Diaminobenzidine (DAB) enhancement, tau, and amyloid, with hematoxylin counterstain where indicated. Sections from brains with multiple sclerosis (80 years old, male, from temporal periventricular white matter and cortex) and leukoencephalopathy (43 years old, male, from periventricular white matter and cingulum) were stained with hematoxylin and eosin, luxol fast blue plus hematoxylin and eosin, and neurofilament (2F11) (Fig. 3 and S3). Hippocampal and visual cortex sections in Fig. S2 were from a 60-year-old male stained with hematoxylin & eosin and a 67-year-old female stained with luxol fast blue respectively. The sclerotic hippocampal section in Fig. S4 was from a 69-year-old female with epilepsy, the control was from a 66-year-old female with no neuropathologic abnormality. Specimens were acquired under Stanford ADRC IRB (Assurance nr. FWA00000935).

Mouse brain section

A female ~10-week-old C57BL/6 mouse (Jackson Laboratories) was housed in a temperature-controlled environment, with a 12-hour light/dark schedule and ad libitum food/water access. It was euthanized for a different study (APLAC #32577) under anesthesia with 2-3% isoflurane followed by cardiac puncture and perfusion with 20 mL phosphate-buffered saline (PBS). The brain was harvested, kept in 4% paraformaldehyde (PFA) in PBS for 24 hours at 4°C, transferred to 10%, 20%, and 30% sucrose in PBS, embedded in Tissue-Tek O.C.T. in dry ice for 1 hour, and cut sagittally into 10μm sections using a cryotome (Leica CM1860). The sections were subsequently washed, mounted on a glass slide, intubated with Iba1 antibody (dilution 1:200), secondary antibody (goat anti-rabbit Cy3 1:200), and cover-slipped. A mid-sagittal section was selected for evaluation (Fig. 2F-I).

Pig brain section

A 4-week female Yorkshire pig (#2) was euthanized for a different study (Stanford APLAC protocol nr 33684), the brain was harvested, cut into 5-mm coronal slabs using a brain slicer, and a mid-frontal slab (#5) was paraffin-embedded, similar to the human pathologic specimen preparation above. The slab was cut into 10μm sections using a Leica HistoCore AUTOCUT microtome. After deparaffinization, a section (#127) was stained with hematoxylin and eosin and cover-slipped (Fig. 2J-M).

Human tongue, colorectal, bone, and artery wall sections

The non-brain tissue sections (Fig. 5 and S8) were obtained from a tissue archive at Erasmus Medical Center, Rotterdam, the Netherlands, approved by the Medisch Ethische Toetsing Commissie (METC) under number MEC-2023-0587. The tissue samples were obtained from patients during surgery. The bone sample was decalcified first using DecalMATE by Milestone Medical. Afterward, all samples were fixed in 4% formaldehyde for 24 hours, dehydrated in increasing alcohol series (70%, 80%, 90%, 96%, 100% ethanol), treated with xylene, embedded in paraffin, and cut with a microtome (Leica RM2165) into 4μm-thin sections. The sections were placed in a decreasing alcohol series to remove the paraffin, mounted on glass slides, stained with hematoxylin and eosin (artery wall with Verhoeff-Van Gieson elastin staining), and then cover-slipped.

Brightfield microscopy

The whole human brain sections were scanned with the TissueScope LE120 Slide Scanner by Huron Digital Pathology, Huron Technologies International Inc. The device measures in brightfield mode with 20X magnification and 0.74 NA, providing a pixel size of 0.4µm. The final images were stored with a pixel size of 1µm.

The hippocampus, cortex, pathology, and animal brain sections were scanned using an Aperio AT2 whole slide scanner with the ImageScope software and a 20X magnification, resulting in brightfield images with a pixel size of 0.5μm.

The stained non-brain microscopy slides were scanned using the Nanozoomer 2.0 HT digital slide scanner by Hamamatsu Photonics K.K., offering a 20X magnification and a pixel size of 0.46µm. The unstained non-brain microscopy slides were scanned using the Keyence VHX-6000 Digital Microscope (with VH-ZST objective, 20X), with a pixel size of 10μm.

ComSLI

Whole human brain (silver-stained), hippocampus, cortex, pathology, and animal brain sections

Measurements were performed with a rotating light source and camera, using a Flexacam C3 12 MP microscope camera (Leica) and a Navitar 12X Zoom Lens with a 0.67X Standard Adapter and a 0.5X Lens Attachment, with 4.25-9μm pixel size. As a light source, an ADJ Pinspot LED II was used, with a 5.1cm diameter and 3.5° full-angle of divergence, oriented at ~45° with respect to the sample plane. A motorized specimen stage enabled the whole human brain section to be scanned in 8x5 tiles, all other brain sections were scanned in a single tile. Images were acquired at 10° rotation steps (36 images/sample) with a 125ms exposure time, except the sections in paraffin that gave very strong scattering and were imaged with a 7ms exposure time. Before the measurement, a 100mm diameter diffuser plate (Thorlabs) was measured for calibration (see below for calibration details).

Whole human brain (Cresyl-violet) and non-brain tissue sections

Measurements were similarly performed with a rotating light source and camera, using a fiber-coupled LED light source consisting of an ultra-high power LED (UHP-FB-W50 by Prizmatix) with 400-750nm wavelength (peak at 443nm), 2-meter long step-index multimode silica (low OH) fiber patch cord (Thorlabs), a 25.4 mm diameter collimating optics (FCM1-0.5-CN by Prizmatix), and a 25.4 mm diameter engineered diffuser (beam shaper) for homogenizing the illumination (ED1-S20-MD by Thorlabs), yielding top-hat beam with 20° full-angle of divergence. The exposure time was adjusted manually per sample to maximize the dynamic range of the captured signal while avoiding saturation (range: 50-100 ms). The light source was oriented at ~45° with respect to the sample and rotated with a motorized specimen stage (ZABER X-RSB060AD-E01-KX14A) in steps of 15° (24 images/sample). Images were taken with a 20 MP monochromatic CMOS camera (BASLER acA5472-17um) and a Rodenstock Apo-Rodagon-D120 Lens, yielding a pixel size of 3μm (4μm optical resolution) and a field-of-view of 16x11mm². A motorized specimen stage was used to perform whole-slide scanning. Before the measurement, a diffuser plate (DG100x100 N-BK7 ground glass diffuser, 1500 grit, Thorlabs) was measured for calibration.

120-year-old myelin-stained brain section

The measurement was performed with a similar camera and lens as for the non-brain tissue sections (BASLER acA5472-17uc and Rodenstock Apo-Rodagon-D120), using an LED display instead of a focused light source (50x50cm², 128x128 RGB-LEDs, Absen Polaris 3.9pro In/Outdoor LED Cabinet). The sample was illuminated by a green circle segment (9° azimuthal and polar widths, respectively) with an effective illumination angle of 47°, which was rotated in 15° steps. Images were taken with 10 seconds exposure time and a gain of 10, and 4 images were averaged per illumination angle to increase signal to noise.

Calibration

Before each measurement session, a diffuser plate was measured under similar conditions. The resulting images were blurred using a Gaussian blur with a 100-pixel radius to homogenize diffuser defects. Subsequently, the blurred images of all angles were divided by the average of their maxima for normalization. These normalized diffuser images were used to calibrate the measured tissue images, aiming to account for the different light intensities across the field of view for each image: Each tissue image was divided by its corresponding normalized diffuser image of the same illumination angle.

Generation of fiber orientation and vector maps

Each calibrated image series from a ComSLI measurement was evaluated with the open-source software SLIX (30), which analyzes the position of scattering peaks to compute the fiber orientations and visualize them in color-encoded maps, using multi-colored pixels and colored vector lines. Measurements with 15° azimuthal steps were processed without filtering. Measurements with 10° azimuthal steps were processed with a Fourier low pass filter (40% cutoff frequency, 0.225 window width) before generating the parameter maps, as described in (15).

Nissl-ST

The fiber orientation maps were computed in Matlab following the procedure described and code shared by Schurr & Mezer (25), using default settings and 100μm as kernel to compute the structure tensor (effective resolution).

3D-PLI

The 3D-PLI measurements were performed using the LMP3D microscope (Taorad GmbH, Germany), containing an evo4070MFLGEC (2048x2048) camera and a Nikon 4x (NA 0.2) lens, which achieves a pixel size of 1.85μm and an in-plane optical resolution of 2.2μm (determined by US-Airforce target). The sample was illuminated by linearly polarized light in 20° rotation angles and analyzed by a circular analyzer as described by Axer et al. (12). 3D-PLI FOM was computed on the supercomputer JURECA at Forschungszentrum Jülich (grant no. 28954).

Diffusion MRI

The diffusion MRI dataset (26) is from a 30-year-old male who underwent 18 hours of diffusion MRI scanning in the MGH-USC 3T Connectom scanner using gSlider-SMS (56). After manually identifying the MR plane that most closely matched the BigBrain histology sections, the entire dataset was rotated using FreeSurfer’s freeview, and the b-vectors were rotated at the same angles (rotation angles for the Silver-stained section were -34° sagittal and 1.5° axial, for the Cresyl-violet -30.4° sagittal). Fiber responses and orientation distributions were computed using the dwi2response and dwi2fod functions in mrtrix3 (31), using the multi-tissue, multi-shell algorithm (57), and visualized in mrview. To generate whole-brain colormaps in Fig. 4 and S6, mrtrix3’s sh2amp function was used to probe fiber orientations at the coronal plane at 5° intervals, and colormaps were generated using SLIX (30).

REFERENCES

12. M. Axer, D. Grässel, M. Kleiner, J. Dammers, T. Dickscheid, J. Reckfort, T. Hütz, B. Eiben, U. Pietrzyk, K. Zilles, K. Amunts, High-resolution fiber tract reconstruction in the human brain by means of three-dimensional polarized light imaging. Front. Neuroinform. 5, 34 (2011).

15. M. Menzel, M. Ritzkowski, J. A. Reuter, D. Gräßel, K. Amunts, M. Axer, Scatterometry Measurements With Scattered Light Imaging Enable New Insights Into the Nerve Fiber Architecture of the Brain. Front. Neuroanat. 15 (2021).

24. K. Amunts, C. Lepage, L. Borgeat, H. Mohlberg, T. Dickscheid, M.-É. Rousseau, S. Bludau, P.-L. Bazin, L. B. Lewis, A.-M. Oros-Peusquens, N. J. Shah, T. Lippert, K. Zilles, A. C. Evans, BigBrain: An Ultrahigh-Resolution 3D Human Brain Model. Science 340, 1472–1475 (2013).

25. R. Schurr, A. A. Mezer, The glial framework reveals white matter fiber architecture in human and primate brains. Science 374, 762–767 (2021).

26. F. Wang, Z. Dong, Q. Tian, C. Liao, Q. Fan, W. S. Hoge, B. Keil, J. R. Polimeni, L. L. Wald, S. Y. Huang, K. Setsompop, In vivo human whole-brain Connectom diffusion MRI dataset at 760 µm isotropic resolution. Sci. data. 8, 122 (2021).

30. J. A. Reuter, M. Menzel, SLIX: A Python package for fully automated evaluation of Scattered Light Imaging measurements on brain tissue. J. Open Source Softw. 5, 2675 (2020).

31. J.-D. Tournier, R. Smith, D. Raffelt, R. Tabbara, T. Dhollander, M. Pietsch, D. Christiaens, B. Jeurissen, C.-H. Yeh, A. Connelly, MRtrix3: A fast, flexible and open software framework for medical image processing and visualisation. Neuroimage. 202, 116137 (2019).

53. B. Merker, Silver staining of cell bodies by means of physical development. J. Neurosci. Methods. 9, 235–241 (1983).

54. H. Mohlberg, B. Tweddell, T. Lippert, K. Amunts, "Workflows for Ultra-High Resolution 3D Models of the Human Brain on Massively Parallel Supercomputers" in International Workshop on Brain-Inspired Computing (2015).

55. O. Vogt, Die myelogenetische Gliederung des Cortex cerebelli. J. für Psychol. und Neurol. 5, 235–258 (1905).

56. K. Setsompop, Q. Fan, J. Stockmann, B. Bilgic, S. Huang, S. F. Cauley, A. Nummenmaa, F. Wang, Y. Rathi, T. Witzel, L. L. Wald, High-resolution in vivo diffusion imaging of the human brain with generalized slice dithered enhanced resolution: Simultaneous multislice (gSlider-SMS). Magn. Reson. Med. 79, 141–151 (2018).

57. B. Jeurissen, J.-D. Tournier, T. Dhollander, A. Connelly, J. Sijbers, Multi-tissue constrained spherical deconvolution for improved analysis of multi-shell diffusion MRI data. Neuroimage. 103, 411–426 (2014).