Tunable Bessel beam two-photon fluorescence microscopy for high-speed volumetric imaging of brain dynamics
- Department of Electrical and Computer Engineering, Princeton University, United States
- Department of Diagnostic Radiology and Nuclear Medicine, University of Maryland School of Medicine, United States
- Omenn Darling Bioengineering Institute, Princeton University, United States
Figures
Figure 1 with 6 supplements
Concept, design, and characterization of tunable Bessel two-photon microscopy.
(A) Comparison of Gaussian (top) and Bessel (bottom) two-photon fluorescence microscopy (TPFM). Gaussian imaging requires sequential acquisition of 2D images at different axial positions, whereas Bessel imaging captures the entire volume in a single frame. (B) Fundamental trade-offs in Bessel beam imaging between spatial resolution, temporal resolution, and axial distinguishability, governed by the numerical aperture (NA), beam length, and side-ring confinement. (C) For targeted opto-stimulation, Bessel beam tuning must preserve a fixed axial beam center to maintain co-registration with stimulation or ablation beams. (D) Schematic of the tunable Bessel module (A1–A3, axicons; L1–L3, lenses; iris) and its integration with Gaussian-TPFM. A1 with L1 generates the Bessel beam; the iris adjusts ΔNA; A2–A3 spacing sets NA; L2–L3 form a 4f relay to the scanning galvo. (E) Zemax simulations showing pupil (left), sample (middle), and propagation (right) images for 0.8 NA with open iris (top), 0.4 NA with open iris (middle), and 0.4 NA with 10% iris opening (bottom). Scale bar: 25λ/n, where n is the refractive index of the immersion medium. (F) Simulated versus experimental Bessel beam cross-section profiles at NA = 0.4 (top) and 0.8 (bottom). Experimental images were acquired using a camera conjugate to the objective’s sample plane. Scale bar: λ/n. (G) Theoretical and experimental results showing adjustment of the Bessel beam NA (blue), resolution (orange), and length (green) as a function of A2-A3 spacing. (H) Simulated versus experimental Bessel beam pupil profiles for ΔNA = 0.005 (top) and 0.041 (bottom). Experimental images were acquired using a camera conjugate to the objective’s back focal plane. (I) Theoretical and experimental results showing adjustment of the Bessel beam ΔNA (blue), half total power radius (orange), and length (green).
Figure 1—figure supplement 1
Comparison of Bessel beams tuning method.
(A) Axicon-lens design. This method adjusts the numerical aperture (NA) and ΔNA through defocusing the last lens. It lacks independent control over NA and ΔNA, and the waist of the generated Bessel beam shifts axially when being tuned. (B) Lens-axicon design. Similar to (A), this method relies on the last lens for tuning NA and ΔNA. It also leads to the generated Bessel beam waist shift axially when being tuned. (C) Dual-axicon design. This method provides independent control of NA and ΔNA, but the generated pupil ring has a non-flat wavefront, causing the Bessel beam waist to shift axially. (D) Focused Bessel beam generation method utilizing an axicon pair. This approach achieves independent NA and ΔNA control without shifting the Bessel beam waist. However, this method is limited to generate short Bessel beams. (E) This design enables independent NA and ΔNA control, but the waist of the generated Bessel beam shift axially when being tuned.
Figure 1—figure supplement 2
Tunable Bessel module.
(A) Input and output of an axicon. (B) Optical diagram of tunable Bessel module with system parameters defined. (C) Optical diagrams showing the module’s capability to independently tune the numerical aperture (NA) (comparing top and middle row) and ΔNA (comparing bottom and middle row). (D) Photo of the tunable Bessel module integrated into a Gaussian-two-photon fluorescence microscopy (TPFM). (E) Experimental comparison of simultaneously two-color Bessel beams with 520 nm (left) and 635 nm (top right) laser. The top right panel presents the 635 nm Bessel beam in physical units (µm), showing its intrinsically larger focus due to the longer wavelength. In the bottom right panel, the 635 nm beam is rescaled by a factor of 520/635 to normalize for wavelength-dependent diffraction. Since the lateral size of the Bessel focus scales linearly with wavelength, this normalized comparison demonstrates that the two beams have identical profiles after accounting for their wavelength difference. (F) Intensity profiles from the vertical dashed white line of (E) comparing the two colors. The horizontal axis is rescheduled by their corresponding wavelengths. (G) Experimental comparison of the two colors pupil profile. (H) Experimental NA comparison between the two colors at various NAs. (I) Experimental axial point spread function (PSF) of NA at 0.4, 0.6, and 0.8 NA. The slight non-uniformities in axial intensity are consistent across different NA settings. This is not induced by the tunable Bessel (tBessel) module itself but rather the centrally peaked intensity distribution of the input Gaussian beam, and axicon tip imperfections (Dudutis et al., 2016).
Figure 1—figure supplement 3
Zemax simulation of the tunable Bessel module: numerical aperture (NA) tuning.
(A) Zemax diagram of the tunable Bessel module. (B) Pupil intensity profiles (top) of three Bessel beams with NA = 0.17, 0.37, and 0.57 (same ΔNA) by varying the separation between the two 5o axicons. Line cuts of the intensity profiles (bottom) show that ΔNA remains constant, demonstrating that adjustments to the axicon pair distance do not impact the delta NA of the system. (C) Pupil phase profiles (top) and line cuts (bottom) of the three Bessel beams in (B).
Figure 1—figure supplement 4
Zemax simulation of the tunable Bessel module: ΔNA tuning, pupil plane characterization.
(A) Pupil intensity profiles (top) and line cuts (bottom) of three Bessel beams with ΔNA = 0.019, 0.025, and 0.038 (same numerical aperture, NA) by varying the iris opening diameter. (B) Pupil phase profiles (top) and line cuts (bottom) of the three Bessel beams in (A).
Figure 1—figure supplement 5
Zemax simulation of the tunable Bessel module: ΔNA tuning, sample plane.
(A) Sample intensity profiles (top) of three Bessel beams with ΔNA = 0.019, 0.025, and 0.038 (same numerical aperture, NA) by varying the iris opening diameter. The larger the ΔNA, the less side-ring excitation the Bessel beam has. (B) Cumulative total energy distribution. The red lines indicate where the half total energy resides.
Figure 1—video 1
Independent continuous adjustment of numerical aperture (NA) and ΔNA with the tunable Bessel module.
Figure 2 with 3 supplements
Large-scale volumetric vascular imaging and neurovascular coupling.
(A) Left: 3D rendering of a 350×350×450 μm3 Gaussian stack showing dense cortical vasculature. Right: representative Bessel images with beam lengths of 17 μm (orange), 35 μm (magenta), and 140 μm (blue), illustrating the trade-off between information content and structural overlap. (B) Tunable Bessel (tBessel) scan combining lateral tiling (4×4) and focal jumps (3, color-coded), yielding 48 tiles covering a 2,500×2,500×450 μm3 cortical volume. Scale bar: 200 μm. (C) Combined millimeter-scale tBessel image from (B). White box: sub-capillary resolution. Green box: penetrating arterioles/venules (arrows). Right: vasoconstriction and vasodilation of the vessel marked in red. Scale bar: 200 μm. (D) Experimental setup for functional imaging of primary visual cortex (V1). Mice with cranial windows over V1 were presented moving gratings in eight directions. Five planes spanning a 260×260×400 μm3 volume were acquired at 2.5 Hz (five trials per direction). Two neurons were analyzed for fluorescence dynamics, and vessels were outlined for resliced space–time plots of diameter changes. (E) Top: averaged calcium traces (green) from two neurons. Middle: raw and segmented resliced images from lines 1 and 2 in (D). Bottom: overlay of neuronal activity (green) and vessel diameter (magenta) showing strong correlation, with vascular changes lagging neuronal responses (arrows).
Figure 2—figure supplement 1
Analysis of inter-vessel distance at different depths in mouse cortex.
(A) Histogram (top) and representative image (bottom) of inter-vessel distance 0–165 μm below pia. The data is fitted to a gamma function, with the mode indicating an optimal Bessel beam length of 44 μm for imaging (red dashed line). (B) Histogram (top) and representative image (bottom) 165–330 μm below pia, with a mode of 36 μm (red dashed line), showing slightly closer vessel spacing compared to the surface. (C) Histogram (top) and representative image (bottom) 330–500 μm below pia, where the mode of vessel separation is 38 μm (red dashed line). (D) Plots illustrating vessel occupancy and overlap along the axial direction from surface to depth. The orange line represents the fraction of pixels in the mean intensity projection that are occupied by blood vessels, while the blue line shows the fraction of these vessel-containing pixels that exhibit overlap. (E) Average blood vessel distance varies at different depths.
Figure 2—figure supplement 2
Neurovascular coupling in response to visual stimulation.
(A) Schematic of experimental setup for functional brain imaging of primary visual cortex (V1). Mice with cranial windows installed over V1 area were imaged while being exposed to moving gratings towards eight directions. (B) Mean intensity projection of an 80-μm-thick image stack at different depths, which was collected with Gaussian focus scanning at 2 μm z steps, with structures color-coded by depth. (C) Images of the same volume of brain collected by scanning a Bessel focus with the length of 80 μm for each depth. (D) Schematic diagram of how to use tunable Bessel (tBessel) foci for cross-layer deep brain imaging. A Bessel focus with the axial length of 80 μm can cover 80 μm deep brain volume. When it is moved by piezoelectric actuators along z direction four times with each time jumping 80 μm, we can cover the complete 0 through 400 μm deep brain volume with five Bessel planes at fast speed (2.5 Hz in this case). (E) Field-of-view (FOV) from Plane #1 in Figure 2D showing where lines were drawn. (F) The intensity over Line 3 from a venous across five trials aligned by eight drifting grating angles from 0° to 315°. Raw and segmented images from resliced stacks for Line 3 were shown. Summed intensity plot of lines (Line 3, in magenta) was overlaid with calcium trace of Neuron 1 with correlation coefficient being 0.035. (G) Cross-correlation matrix among the two selected neurons and the 15 blood vessels. (H) Raster plot of the normalized activities of the two selected neurons and diameter changes of the 15 blood vessels (Line 1–15) in (D). (I) Correlation coefficients between the temporal curves from the 15 blood vessels and Neuron 1 (blue) or Neuron 2 (red).
Figure 2—video 1
Time-lapse hemodynamics imaging over a 1400×1400×150 μm3 volume showing spontaneous neurovascular activities at 15 Hz and responses to moving grating stimulations at 3 Hz.
Figure 3 with 5 supplements
High-speed blood flow mapping in normal and ischemic stroke mice.
(A) Widefield and schematic view of mice brain vasculature. (B) Schematics (left) and images (right) comparing Gaussian (top; 2 μm depth of focus) and tunable Bessel (tBessel) (bottom; 80 μm long) scanning. Gaussian imaging restricts measurements to vessels nearly parallel to the imaging plane, whereas tBessel enables blood flow measurement across a broad range of vessel orientations. Scale bars: 20 μm. (C) Workflow for blood flow measurement. Left: 3D Gaussian stack highlighting a vessel segment (blue). Bottom: real-time tBessel imaging showing moving blood cells. Right: kymographs, with blood cells appearing as dark streaks (arrows). Velocities are extracted from 2D tBessel streak slopes and corrected for vessel 3D length. (D) Mean intensity projection of a Gaussian stack (400×400×120 μm3, 40–160 μm below pia, depth color-coded) versus tBessel scans with NA = 0.4, 0.6, and 0.8. Insets: experimentally measured point spread functions (PSFs, scale bar: 1.5 μm) and optical transfer functions (OTFs). (E) Trade-off between spatial resolution and imaging speed: high resolution for capillaries versus low resolution for larger vessels. (F) Blood flow speed versus vessel diameter for 42 segments imaged with 0.4 NA tBessel at 58 Hz. (G) Frame scanning versus line scanning: restricting scans to vessel regions improves speed and accuracy for large vessels. (H) Blood flow speed versus vessel diameter measured with 1 kHz line scans. Red: arteries; blue: veins. (I) Schematics of stroke induction. (J) tBessel-TPFM hemodynamics of a 1400×1400×140 μm³ volume before stroke induction. Blood flow speeds (mm/s) measured by 1 kHz line scans; numbers indicate speed, arrows indicate direction (red: arterioles, blue: veins). Representative kymographs of three linecuts shown at right. Scale bar: 100 μm. (K) Same measurement after stroke induction. Scale bar: 100 μm. (L) Paired plot of blood flow speed before and after stroke induction (N=17). Wilcoxon test, p=1×10–5.
Figure 3—figure supplement 1
Tracking hemodynamics in live mouse brains.
(A) Schematics of Gaussian imaging: axial motion can move the structures of interest in and out of the focal plane. (B) A representative time-lapse vasculature image (400×400 μm2) collected with Gaussian two-photon fluorescence microscopy (TPFM). Scale Bar: 100 μm. On the right, kymograph of the green line cut. Red boxes highlighted axial motion-induced artifacts. (C) A color-coded projection of the time-lapse image series. White color indicates structures that stay unchanged, while other colors highlight structural changes due to mouse axial motion. (D)-(F), Schematics (D), representative time-lapse and kymograph (E), and color-coded projection (F) of the same blood vessel images collected by an 80 μm long tunable Bessel (tBessel) beam. (G) Color-coded blood vessel orientation map showing a relatively uniform distribution of blood vessels orientations. Scale bar: 50 μm. (H) Diagram of the line-scan measurement under the conventional Gaussian foci with axial point-spread function of 2 μm. When the length of the line at the xy plane is 15 μm, since arctan(2/15) is 7.6o, blood vessels tiling higher than this angle (−7.6 to –90 degrees and –7.6 to –90 degrees) will not be available for line-scan blood flow measurement. (I) Cumulative distribution of orientation angles of blood vessels shown in (G) averaged over five projections. Gray area shows the standard deviation. Blue area indicates the measurable area and red area indicates the non-measurable area in the setting shown in (H). Numbers show the probability correspond to –7.6 and 7.6 degrees at the averaged line. (J) Schematic of red blood cells moving through blood vessel with an angle to the horizontal plane. The same red blood cell must be captured between two frames to be registered for blood flow speed measurements. (K) Simulated maximum measurable flow speed plotted against tBessel resolution with same field-of-view (FOV) and Bessel length, of blood vessels with different orientation angles (color-coded). (L) Simulated maximum measurable flow speed plotted against tBessel length of blood vessels with different orientation angles (color-coded). A Gaussian focus is equivalent to a 2 μm long tBessel beam as shown in the gray shaded area.
Figure 3—figure supplement 2
Measuring blood flow speed with line-scan tunable Bessel (tBessel)-two-photon fluorescence microscopy (TPFM).
(A) Schematic of line-scan for blood flow measurement. The line-scan path is represented as a red line and a stack of sequential line-scans showing the red blood cell at three different locations in the vessel. (B) The 2D imaging plane before line-scanning the same blood vessel under Gaussian foci (upper panel) or 80-μm-length Bessel foci (lower panel). (C) Space-time images and the plot of blood flow speed as a function of time. The repetition rate of the line-scan is 1 kHz. (D) Violin plot showing the distribution of flow speed (in millimeter per second, mm/s) calculated through the streak angle of RBCs in each imaging paradigm. N=234 angles for Gaussian and 97 for Bessel. P-values were calculated by the two-sided Student’s t-test. p=0.21. (E) Gaussian image stack of image of a 200×200×80 µm3 volume of vasculature and tBessel-TPFM volumetric imaging of the same volume. (F) A segment of blood vessel was chosen for line scan at 1 kHz for 5 s (G) Kymographs were generated and analyzed using the line-scanning particle image velocimetry method. Median velocity was calculated. (H) and (I), the line width was corrected by reslicing the stack and found the spanning range of the line along z direction. The true line width was calculated for correction of the flow speed. (J) Blood flow speed before and after correction from 16 line measurements.
Figure 3—figure supplement 3
Characterization of photothrombosis model of ischemic stroke in the mouse visual cortex.
(A) The bright field images of the cranial window (3 mm diameter) and LSI images before and after stroke induction. (B) TTC staining of the mouse brain 24 hr (day 1) after stroke induction.
Figure 3—video 1
Time-lapse imaging showing direct comparison of axial motion artifact between Gaussian-two-photon fluorescence microscopy (TPFM) (left) and Bessel-TPFM (right) over a 400×400 μm2 field of view at 30 Hz.
Figure 3—video 2
Time-lapse imaging of blood vessel for blood flow velocity measurement over a 400×400×120 μm3 volume at 58 Hz.
Figure 4 with 1 supplement
Video-rate volumetric functional imaging with targeted optogenetic stimulation.
(A) Experimental setup. A Gaussian beam (blue) delivers 3D localized optogenetic stimulation to a neuron, while a tunable Bessel (tBessel) beam enables volumetric imaging symmetrically above and below the stimulation plane. (B) Schematics and axial point spread functions (PSFs) of the stimulation Gaussian beam (blue), and short (yellow) and long (red) tBessel imaging beams, showing that the extended projection is always symmetric with respect to the Gaussian stimulation plane. The non-uniformity in intensity comes from the input Gaussian beam. Scale bar: 10 μm. (C) Depth color-coded projection of volumetric functional imaging showing the stimulated neuron (neuron 1) and surrounding neurons; 16 neurons were selected for analysis. Scale bar: 10 μm. (D) Representative calcium transients from neurons numbered in (C). Light blue bars mark the timing of optogenetic stimulation pulses delivered to neuron 1. (E) Pearson correlation map of calcium dynamics among the neurons numbered in (C).
Figure 4—figure supplement 1
Integrated optogenetic stimulation and volumetric imaging.
(A) Schematic (left) of the experimental configuration. Right: In fixed tissue, arbitrary stimulation patterns can be planned and executed with high precision using the stimulation beam (top). In vivo stimulation begins with selecting a target neuron, followed by planning a spiral (vortex) pattern to deliver light precisely to the soma for single-cell activation. (B) Schematic of the in vivo experiment. Mice expressed a bicistronic construct (syn-GCaMP6s-P2A-ChRmine-Kv2.1) that enables co-expression of the calcium indicator GCaMP6s and the optogenetic actuator ChRmine in the same neurons. Right: Example calcium trace showing reliable responses to repeated optogenetic stimulation (red arrows). (C) Sequential stimulation of four individual neurons (3 ms per site with 1 ms inter-site pause). Bottom: Calcium traces from each ROI. Yellow arrows mark stimulation onset. Scale bars: 100 μm.
Figure 5 with 2 supplements
Rapid volumetric imaging of microglial responses to targeted ablation.
(A) Experimental setup: a Gaussian beam (green) delivers targeted ablation of a single microglial cell, while a tunable Bessel (tBessel) beam (red) provides symmetric volumetric imaging above and below the ablation plane. (B) Mean intensity projection of a 200×200×120 μm3 Gaussian stack, depth color-coded. White box marks the targeted microglia. Scale bar: 20 μm. (C) Gaussian plane images of the targeted cell before (left, arrow) and after (right, arrow) ablation. Scale bar: 10 μm. (D) Representative 15 Hz volumetric tBessel data with a 120-μm-long beam (NA = 0.7). Raw (left), denoised (center), and deconvolved (right) images. Scale bar: 10 μm. (E) Time-lapse snapshots showing process extension toward the ablation site (arrowheads) at 1 min 15 s, 4 min 0 s, and 9 min 0 s, revealing a two-wave response. Scale bar: 10 μm. (F) Zoomed views of two regions (green and blue boxes in (D)). Arrowheads indicate process extension toward the ablation site (top, green box) and retraction opposite the lesion (bottom, blue box) across the 10 min imaging period. Scale bar: 5 μm.
Figure 5—figure supplement 1
Denoising and deconvolution pipeline.
(A) Schematic of the Content-Aware Image Restoration (CARE)-based U-Net architecture for denoising. Training data were generated by averaging 10 consecutive raw frames to produce high-signal-to-noise-ratio (SNR) ‘ground truth’ images, paired with corresponding single-frame low-SNR inputs. (B) Representative example comparing raw, denoised, and ground truth images. Line profile analysis (right) demonstrates high concordance with the ground truth. Scale bar: 20 μm. (C) Comparison images of raw, denoised, and deconvolved images. Scale bars: 20 μm.
Figure 5—video 1
Time-lapse imaging of microglia resoponses immediate after microablation of one microglial cell over a 200×200×120-μm3 volume at 15-Hz for 10-minutes.
Author response image 1
Validation of CARE-based denoising for microglial imaging.
(a) Comparison of 10-frame averaged normalized raw (left), CARE-denoised (middle), and their pixel-wise difference (right) images. The second row shows a zoomed-in view of the boxed region. (b) Color-coded time-lapse projections over a 10-min imaging session for the raw (left) and CARE-denoised (middle) data, along with their pixel-wise difference (right).
Additional files
Download links
A two-part list of links to download the article, or parts of the article, in various formats.
Downloads (link to download the article as PDF)
Open citations (links to open the citations from this article in various online reference manager services)
Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)
Tunable Bessel beam two-photon fluorescence microscopy for high-speed volumetric imaging of brain dynamics
eLife 15:RP110228.
https://doi.org/10.7554/eLife.110228.2
