High-speed volumetric two-photon fluorescence imaging of neurovascular dynamics

Abstract

Understanding the structure and function of vasculature in the brain requires us to monitor distributed hemodynamics at high spatial and temporal resolution in three-dimensional (3D) volumes in vivo. Currently, a volumetric vasculature imaging method with sub-capillary spatial resolution and blood flow-resolving speed is lacking. Here, using two-photon laser scanning microscopy (TPLSM) with an axially extended Bessel focus, we capture volumetric hemodynamics in the awake mouse brain at a spatiotemporal resolution sufficient for measuring capillary size and blood flow. With Bessel TPLSM, the fluorescence signal of a vessel becomes proportional to its size, which enables convenient intensity-based analysis of vessel dilation and constriction dynamics in large volumes. We observe entrainment of vasodilation and vasoconstriction with pupil diameter and measure 3D blood flow at 99 volumes/second. Demonstrating high-throughput monitoring of hemodynamics in the awake brain, we expect Bessel TPLSM to make broad impacts on neurovasculature research.

Fig. 1. Design and characterization of a commercial two-photon laser scanning microscope with a Bessel focus module.

a Schematic of the microscope. A half-wave plate (HWP), a polarizing beamsplitter (PBS), and a removable mirror allow switching between Bessel (red) and Gaussian (yellow) beam paths. In the Bessel path, a spatial light modulator (SLM) and a lens generate an annular illumination pattern, which after spatial filtering by an annular mask is imaged via a 4f system onto the galvos and subsequently imaged via a scan and tube lens pair onto the objective lens back focal plane. b Lateral and axial point spread functions for Gaussian and Bessel foci. X and Y scale bars: 1 µm. Z scale bar: 5 µm. Results from one 0.2-µm-diameter bead. c Schematic comparison of Gaussian and Bessel volumetric TPLSM methods. Gaussian volumetric imaging requires multiple 2D frames taken at different Z-positions, while Bessel volumetric imaging is achieved with a single frame.

Fig. 2. In vivo volumetric structural imaging of vasculature and glia with Bessel TPLSM.

a–c Gaussian TPLSM images of vasculature labeled with dextran-conjugated Texas Red at 55 µm, 225 µm, and 420 µm depths, respectively, over a 1.4 mm × 1.4 mm area in the mouse cortex in vivo. d–f Gaussian TPLSM image stacks of vasculature at 0–110 µm, 170–280 µm, and 370–470 µm depths, respectively, color-coded by depth. g–i Scanning the Bessel focus in 2D captured all vasculature in the volumes within (d–f). j–l Gaussian (j, k) and Bessel (l) images of GFP-expressing glia imaged concurrently with a, d, and g, respectively. Insets: zoomed-in views of the white-boxed regions. Red arrowheads: vessels oriented parallel to the Bessel focus; a and g use a grayscale on the normalized square root of fluorescence signal to highlight dim structures without saturating bright structures. All other panels use grayscale on the normalized linear fluorescence signal. Representative data from four mice. Scale bars: 200 µm for full FOV; 20 µm for insets. Post-objective excitation power: Gaussian: 45–177 mW; Bessel: 217 mW.

Fig. 3. Bessel TPLSM signal is correlated with vessel size and captures distributed dynamics of vasodilation and vasoconstriction in 3D.

a A 1.4 mm × 1.4 mm × 0.1 mm volume of vasculature imaged at 15 Hz using Bessel TPLSM, visualized in grayscale on the normalized square root of fluorescence signal. Insets: zoomed-in views of the white-boxed region at two time points, showing changes in vessel size. b Gaussian (single plane at Z = 50 µm) and Bessel images of the red-boxed region in a captured at different times, visualized in grayscale on the normalized fluorescence signal. Red arrows point to three vessels (large, medium, small) to highlight the differences in their fluorescence signal strength between Gaussian and Bessel TPLSM. c Fluorescence vs. vessel diameter for 60 vessel segments in a (see Supplementary Fig. 1) imaged with Gaussian or Bessel TPLSM. d Time traces of fluorescence signal changes of the magenta regions of interest (ROIs) and blood vessel diameter measured along the green lines in a for four vessel segments. e Scatter plot of the data in d. f, g Maps of cross-correlation coefficients between ROIs tiling the FOV in a, and a reference ROI (indicated by arrows and black squares). Representative data from four mice. Scale bars: 200 µm for a, f, g; 20 µm for insets in a; 100 µm for b. Post-objective power: Gaussian: 45 mW; Bessel: 217 mW.

Fig. 4. Entrainment of vasodilation and vasoconstriction of a 3D vasculature network with pupil diameter measured by Bessel TPLSM.

a A 1.4 mm × 1.4 mm × 0.1 mm volume of vasculature imaged at 15 Hz with Gaussian TPLSM, color-coded by depth. b Bessel TPLSM image of the same volume in a, visualized in grayscale on the normalized square root of fluorescence signal. c Example pupil images at time points 1 and 2 of the ipsilateral eye acquired concurrently with vasculature imaging. Dashed yellow ovals: pupil profiles automatically segmented from video data. d Pupil diameter and signal time traces of two regions of interest (ROIs, arrows and black squares in e), showing positive and negative correlation with pupil diameter, respectively. CC: correlation coefficients. Orange circles on pupil time trace indicate time points 1 and 2. e Map of cross-correlation coefficients between ROIs tiling the FOV in b and pupil diameter. Representative data from four mice. Scale bars: 200 µm for a, b, and e; 1 mm for c. Post-objective power: Gaussian: 47 mW; Bessel: 167 mW.

Fig. 5. High-speed volumetric measurement of cerebral blood flow speed with Bessel TPLSM.

a Bessel image of a 416 µm × 416 µm × 80 µm volume of vasculature that was imaged at 99 Hz with Bessel TPLSM for blood flow speed measurements, with structures visualized in grayscale on the normalized square root of fluorescence signal. Blue and red lines trace two example vessel segments. b Gaussian image stack of the same volume in a, color-coded by depth and overlaid with median blood flow speeds (in mm/s) of 11 blood vessel segments. c Skeletonized 3D Gaussian stack from b for the measurements of blood vessel segment 3D lengths. d Kymographs of two example vessel segments in a, obtained by plotting pixels along the traced blood vessels (horizontal) across time (vertical). Dark diagonal streaks are caused by RBCs traveling along the vessel segment. Note that the horizontal axis represents projected 2D distance. e Kymographs of the same vessels after nonlinear transformation so that the horizontal axis represents 3D distance. For flow speed measurement, each kymograph was divided into 0.5-s-long blocks (cyan dashed lines, Methods). Yellow lines represent the distance–time relationship of the blood flow measured for each block. f Changes in blood flow speed over 1 min for the blue vessel segment in a. a–f: Representative data from eight volumes in four mice. g Median blood flow speed plotted against blood vessel diameter for 63 vessel segments between 0 and 180 µm below dura, collected from four mice. Scale bars: 100 µm for a–c. Post-objective power: Gaussian: 45 mW; Bessel: 217 mW.