SeeThrough: a rationally designed skull clearing technique for in vivo brain imaging

Abstract

Light scattering in the skull limits optical access to the brain. Here we present SeeThrough, a skull-clearing technique that enables simple, high-resolution, and minimally-invasive brain imaging without skull removal. Through systematic screening of over 1600 chemicals, we rationally developed a refractive index-matching solution that combines water- and organic solvent-based components, achieving both high clearing efficiency and biocompatibility. The reagents exhibit minimal brain penetration, maintain tissue integrity, and avoid inflammatory responses. Notably, SeeThrough provides imaging sensitivity and contrast comparable to open-skull window imaging, while permitting minimally-invasive monitoring of brain border macrophages as well as blood and cerebrospinal fluid dynamics. Combined with two-photon imaging, SeeThrough enables spatially and temporally scalable imaging applications in the mouse brain, including ~400 µm deep imaging, one-month longitudinal imaging, and mesoscale, cellular-resolution monitoring of brain activity for network-level analysis. Thus, SeeThrough offers a broadly accessible platform for high-throughput, physiology-preserving imaging of the brain parenchyma and brain-skull interface.

Fig. 1. Development of the RI matching solution suitable for in vivo skull clearing.

a Schematic workflow for the development of woRIMS. To develop a skull clearing protocol that achieves both high RI (RI = 1.56) and biocompatibility, an aqueous intermediate solvent (AqIS) was employed, along with a mixture of a wRIMS and an oRIMS that is miscible with the AqIS. dD dispersion parameter, dP polar parameter, dH hydrogen-bonding parameter, eRI estimated RI. b A plot of dP and dH distances between a selected set of 20 organic solvents and a 50–75% ethanol aqueous solution. Filled and open circles indicate “miscible” and “immiscible,” respectively. The RIs of each organic solvent were displayed using the indicated color code. In a 50–75% ethanol aqueous solution, the miscible radius of the HSP sphere is ~20, indicated by the red curve. c Hydration scores for 640 independent, non-detergent, and salt-free chemicals re-analyzed from ref. (Supplementary Fig. 2 and Supplementary Data 2) are displayed as box-and-whisker plots (boxes, 25–75 %; whiskers, 10–90 %; vertical line, median; cross, mean); for each chemical the value plotted is the average of two technical replicates, whereas reference compounds (C#0001–0012) used for normalization are excluded from n. Normality (Kolmogorov–Smirnov, α = 0.05) failed only for the alcohol group; therefore, functional groups were compared with the alcohol group by a two-sided Steel test (family-wise α = 0.05). Exact results for significant increases: 1° amine (T = 6.76, ρ = 0.194, P = 1.88 × 10⁻¹⁰), 3° amine (T = 4.13, P = 5.05 × 10⁻⁴), urea (T = 4.41, P = 1.47 × 10⁻⁴), nitrile (T = 4.74, P = 3.00 × 10⁻⁵), aromatic (T = 5.79, P = 9.61 × 10⁻⁸), phenyl (T = 3.99, P = 9.02 × 10⁻⁴), and pyridine (T = 5.08, P = 5.41 × 10⁻⁶); all other groups showed P > 0.05. The median estimated refractive index (eRI) for each chemical group is indicated by color coding. d A plot of RI values for saturated mixed solutions of the selected wRIMS candidates with BA (black square) or VA (red circle) against the HSP distance between those wRIMS and BA or VA (n = 3). The RIs of BA and VA are indicated by the black and red lines, respectively. Data were presented as mean ± SD. e The visual appearance of the BA/#0640 and BA/#1050 mixture. f Bright-field images of the isolated skull before and after clearing, and a plot of quantified transmittance after clearing. Data were shown as mean ± SD (n = 3 independent mice per treatment, one skull per mouse, ***P < 0.001, ns: P > 0.9999, one-way ANOVA with Tukey’s multiple comparison test. P < 0.0001 for PBS + PBS, AqIS + SDBA, AqIS + glycerol, EDTA + glycerol, and AqIS + RIMS-1.51, and P = 0.0008 for AqIS + VA/ANP, when compared with AqIS + BA/ANP). Each solution was referenced from past literature; *1: ref. , *2: ref. , *3: ref. , *4: ref. . g The chemical structures of benzyl alcohol and antipyrine, and the solution composition of woRIMS. Scale bar, 2 mm (f). Credits. Panel a includes illustration elements © Shutterstock.com (licensed), not covered by the article’s CC‑BY 4.0 licence.

Fig. 2. In vivo brain imaging using SeeThrough.

a Schematic of the SeeThrough procedures. b Camera and two-photon images of the vasculature (top) and the tdTomato-expressing neuronal dendrite (bottom) captured through the untreated (left) and SeeThrough-treated (middle) skull and through the open-skull glass window (right) in the same mouse (7 weeks old). Arrows indicate the blood vessels that disappeared after the open-skull surgery. c Raw (left) and normalized (right) tdTomato intensity profiles along the dashed lines in (b). d Comparison of the tdTomato intensity measured from the same dendritic segments before/after SeeThrough and after the open-skull surgery with durotomy (no treatment, n = 7 dendritic spines from three mice; SeeThrough, n = 15 dendritic spines from five mice; open-skull surgery with durotomy, n = 15 dendritic spines from five mice). Plots obtained from the same dendritic segments are connected with gray lines. **P = 0.0027, ns not significant (P = 0.5914), two-tailed Kruskal–Wallis test. Data were presented as mean ± SEM. e Comparison of the full width at half maximum (FWHM) of the tdTomato intensity plots along the same dendritic spines imaged through the SeeThrough-treated skull and the open-skull (with durotomy) glass window (n = 39 dendritic spines, P = 0.7985, two-tailed Wilcoxon test). Data were presented as mean ± SEM. f Schematic of the experimental design. Ca²⁺ imaging was performed using the same mice, first through the SeeThrough-treated skull and then through the open-skull glass window, targeting similar locations and depths in the primary motor cortex. g, h, Images (g) and traces (h) for jGCaMP8f signals in dendrites through the SeeThrough-treated skull (top) and the open-skull glass window (bottom). i Comparison of ΔF/F between SeeThrough and open-skull experiments (SeeThrough, n = 14 dendrites from two mice; open skull, n = 12 dendrites from two mice). ns not significant, P = 0.5604, two-tailed Mann–Whitney U-test. Data were presented as mean ± SEM. See also Supplementary Movie 1–4. Scale bars, 100 μm (b: top), 20 μm (g), 5 μm (b: bottom). Credits. Panels a, f include illustration elements © Shutterstock.com (licensed), not covered by the article’s CC‑BY 4.0 licence.

Fig. 3. SeeThrough enables minimally invasive brain imaging.

a Schematic of the experimental design. b–f Calibration curves (black lines) for each chemical to determine its concentration in the brain parenchyma (b EDTA; c SDBA; d EtOH; e urea; f ANP). Black squares indicate signals measured by standard concentrations for each chemical (standard concentrations: b 3.125, 6.25, 12.5 µM; c 0.001, 0.002, 0.005, 0.008%; d 0.005, 0.010, 0.020, 0.025%; e 0, 1.25, 5, 10, 12.5 mg/dl; f 1, 2, 5, 10 µM). Blue triangles and red circles indicate the determined concentrations in control (PBS-treated) and SeeThrough-treated samples (n = 3), respectively (PBS: d 0.015%; e 7.54 mg/dl; SeeThrough: d 0.007%; e 6.63 mg/dl; f <1 µM). Detection kits (b–e) and LC-MS (f) were used for the quantification assay. Data were presented as mean ± SEM. g Monitoring tdTomato-expressing microglia in the visual cortex in Iba1-tdTomato transgenic mice 0 h (left) and 24 h (right) after the SeeThrough treatment (top), and the open-skull surgery with (w/, middle) or without (w/o, bottom) durotomy. h Comparison of the tdTomato fluorescence between 0 and 24 h after the SeeThrough treatment (left, n = 9 areas from three mice, P = 0.5703, two-tailed Wilcoxon test) and open-skull surgery with (middle, n = 9, **P = 0.0039, two-tailed Wilcoxon test) or without (left, n = 7 from three mice, P = 0.2968, two-tailed Wilcoxon test) durotomy. ns not significant. Data were presented as mean ± SEM. i Immunofluorescent images of the visual cortex in the coronal brain sections using anti-GFAP antibody. Brains were fixed without any treatment (left), 24 h after the SeeThrough treatment (second left), and 24 h after the open-skull surgery with (middle) or without durotomy (second right). To assess the long-term effect, SeeThrough was performed three times over seven days (on the first, fourth and seventh days) and then the brains were fixed (right). Cortical regions just below the treated areas were examined. j Comparison of the GFAP immunoreactivity among no treatment (n = 4 areas from two mice), 24 h after the SeeThrough treatment (n = 6 from three mice), 24 h after the open-skull surgery without or with durotomy (n = 6 from three mice for each), and 7 d after the initial SeeThrough treatment (n = 12 from three mice). *P = 0.0275 (no treatment vs open-skull with durotomy), 0.0305 (open-skull with durotomy vs SeeThrough, 7 d), and 0.0493 (open-skull with durotomy vs open-skull without durotomy), **P = 0.0029, two-tailed Kruskal–Wallis test. Data were presented as mean ± SEM. k Schematic of the experimental design. l Measurement of the bone mechanical strength. m Mechanical strength of parietal bone samples surface-treated for ten sessions with PBS or SeeThrough solutions, and whole-immersed for ten sessions with SeeThrough solutions (n = 7 for surface immersion with PBS and n = 9 for surface and whole immersion with SeeThrough solutions, **P = 0.0038, ***P = 0.0008, one-way ANOVA with Tukey’s multiple comparison test). Data were presented as mean ± SEM. Scale bars, 100 μm. Credits. Panel a include illustration elements © Shutterstock.com (licensed) and panel k include illustration elements © ACworks/illustAC, not covered by the article’s CC‑BY 4.0 licence.

Fig. 4. Spatially and temporally scalable brain imaging using SeeThrough.

a A 3D reconstructed image of the EYFP-expressing neurons in the visual cortex captured through the SeeThrough-treated skull (9 weeks old). b Focal plane images of the layer 1, 2/3, and 5 in the cortex. c Time-lapse images of the same dendrite captured through the SeeThrough-treated skull. Yellow and blue arrowheads indicate newly formed and eliminated dendritic spines. d Wide field-of-view (left) and focused (middle) camera images of the cerebral cortex containing a sparse subset of tdTomato-expressing neurons. Two-photon images of the dendrites located in the colored rectangles in the middle images (right). Three series of dendritic branches from different cortical areas were tested in three mice (3 weeks old). e Two-photon images of the dendrite shown in the yellow rectangle in (d). The XY, XZ, and YZ images projected with maximum fluorescence intensity exhibit minimal aberration and satisfactory Z resolution. A total of ten dendritic branches (3, 3, and 4 branches per mouse) were tested in three mice (3 weeks old). Scale bars, 5 μm (c, d, right, e), 100 μm (a, b, d, middle), 1 mm (d, left).

Fig. 5. Mesoscale, cellular-resolution Ca²⁺ imaging using SeeThrough.

a Schematic of the integration of SeeThrough and FASHIO-2PM. b A camera image. The yellow square indicates the field-of-view (FOV). Overlaid white lines show borders according to the Allen mouse common coordinate framework. c A standard deviation projection image for Ca²⁺ imaging of a contiguous full FOV including layer 2 cortical neurons labeled with G-CaMP7.09 at 7.65 Hz sampling rate (150 μm depth from the cortical surface). Representative neurons are shown around the image. d An overlaid diagram showing 2965 regions of interest (ROIs). The colors correspond to the Ca²⁺ signals shown in (e). e Representative Ca²⁺ signals and deconvolved signals from 14 neurons. f Z-scored deconvolved signals of individual neurons recorded using FASHIO-2PM across 13 cortical regions in the same representative mouse. Neurons are sorted by region, and the global mean activity of all neurons is plotted above the plot. The color bar indicates the corresponding cortical regions. g Partial correlation coefficient matrix quantifies the pairwise relationships between the activities of individual neurons in the representative mouse, after removing the influence of the global mean activity. h Distribution of partial correlation coefficients for all neuron pairs in the representative mouse. i Distribution of partial correlation coefficients for individual mice. The inner lines indicate the quartile ranges (25th–75th percentiles) and the median. j Mean partial correlation coefficients calculated in 100 µm bins as a function of distance, averaged across individuals (n = 3 mice). Error bars represent the SEM. k Binary adjacency matrix for the same mouse as in (g), obtained by thresholding the partial correlation coefficients at the 99th percentile (partial correlation >0.184). l Visualization of 2000 randomly selected links from (k), color-coded by distance. Links shorter than 500 µm are shown in cyan, those between 500–1500 µm in magenta, and links longer than 1500 µm in yellow. m Degree distribution of neurons for the same mouse in (k). n Visualization of links for high-degree neurons. Links are color-coded based on degree rank: cyan for neurons in the top 5–2.5%, magenta for the top 2.5–1%, and yellow for the top 1%.