Breaking the volumetric imaging barrier.

(a) Current fluorescence microscopy approaches are bounded by a volumetric imaging barrier (thick pink line; inspired by (Daetwyler & Fiolka 2023)). Resolution is limited by the diffraction limit. The accessible imaging volume is limited by specifications of microscope objectives for life sciences. The former can be surpassed by using tissue expansion, and the latter can be overcome by leveraging highly engineered lenses from the electronics metrology industry. (b) The etendue (G) of 90% of life sciences objectives (magenta) are bounded by 0.0243 mm2 < G < 0.9503 mm2 (Yueqian & Herbert 2019), apart from custom lenses (black), such as the Mesolens (G = 6.25 mm2) (McConnell et al. 2016) and 2p-RAM lens (7.07 mm2) (Sofroniew et al. 2016). In contrast, lenses developed for electronics metrology can have G > 10 mm2. The lens used in the ExA-SPIM system provides a field of view of 16.8 mm2 with NA = 0.305 (G = 19.65 mm2).

Microscope overview.

(a) Schematic of the ExA-SPIM system. Light enters the system from the laser combiner and is reflected by mirror M1. A cylindrical lens focuses the light in one dimension onto the surface of a tunable lens, which is magnified onto the back focal plane of the excitation through a 1.5× relay consisting of lenses L1 and L2 and mirrors M2 and M3. The excitation objective is oriented vertically and dipped into a liquid immersion chamber. The tunable lens is conjugated to the back focal plane of the excitation objective to enable axial sweeping. A pair of galvo mirrors are used in tandem to translate the position of the light sheet in z (along the optical axis of the detection objective). The detection objective is oriented horizontally. A beam splitter is removed from the lens and replaced with approximately 35 mm of water. A large-format CMOS camera captures images from the detection lens, at a back focusing distance of 50 cm. (b) The field of view of the system is 10.6×8.0 mm, which is digitized by the camera into a 151-megapixel (MP) image with 0.75 µm/px sampling. This large field of view dramatically reduces the need for tiling. For example, a 3× expanded mouse brain can be captured in only 15 tiles. (c) The PSF is shown in the xy, xz, and yz planes. The mean and standard deviation of the lateral and axial full-width half-maximum are shown as a function of x and y position across the full field of view. (d) The field curvature and distortion of the system as a function of field position is shown for different wavebands. The field curvature is <2.5× the depth of focus (DoF) for all wavebands. This performance is better than “Plan” specified life sciences objectives (Yueqian & Herbert 2019). (e) The relative signal-to-noise-ratio (rSNR) of the VP-151MX CMOS camera and an Orca Flash V3 sCMOS camera as a function of imaging speed. The VP-151MX camera provides equivalent SNR at nearly twice the imaging speed.

Nanoscale imaging of entire intact mouse brains.

(a-b) Intact mouse brains were expanded (3×) and sparsely labeled neurons expressing tdTomato were imaged using the ExA-SPIM microscope ((Ouellette et al. 2023), (c) Single neurons were traced and reconstructed from the resulting imaging data. (d) Due to a lack of sectioning, stable specimen mounting, fast imaging times, and minimal imaging distortions, adjacent imaging tiles are easily aligned in overlapping regions (Supplementary Videos 1-2). Nanoscale imaging resolves individual dendritic spines (e-h) and axonal varicosities (i-k) with near-isotropic resolution (Supplementary Videos 3-6). Images are displayed as maximum intensity projections with the following thicknesses: (b) 26 mm, (d) 2 mm, (e-h) 1 mm, (i-k), 100 µm.

Expansion and imaging of a large volume of macaque brain.

A 1×1×1.5 cm block of macaque primary motor cortex was expanded (3×) and imaged on the ExA-SPIM (Supplementary Videos 7-9). Corticospinal neurons were transduced by injecting tdTomato-expressing retro-AAV into the spinal cord. (a-b) Maximum intensity projections of the imaged volume pseudo colored by depth. The axes descriptors in (a) indicate the 5×3 tiling used to image this volume. (c-f) Fine axonal and dendritic structures including descending axons, collaterals and dendritic spines are clearly discernable in the images throughout the entire volume. See Supplementary Video 10. Images are displayed as maximum intensity projections with the following thicknesses: (a) 23 mm, (b) 45 mm, (c-f) 1 mm.

ExA-SPIM imaging of human tissue.

(a-b) A region from the medial temporal-occipital cortex was manually dissected into a ∼1×1 cm block, which was subsequently sectioned into ∼100 µm sections for tissue expansion (4×), labeling, and ExA-SPIM imaging. (c-d) Maximum intensity projection of a region of interest from white and gray matter, pseudo colored by depth. (e-f) Individual axons and their trajectories are clearly resolved with high contrast. Images are displayed as maximum intensity projections across 350 µm (b-f).

Comparison of large-scale volumetric imaging modalities.

ExA-SPIM imaging across scales.

Comparison between standard cleared-tissue and expansion-assisted SPIM.

(a-c) A traditional cleared-tissue SPIM system uses standard 100 mW or less lasers, a scientific CMOS camera, and life sciences objectives with higher NA and <10 mm working distance. In theory, these systems can image an entire cleared mouse brain at 500 nm or less resolution without any physical cutting. However, this would require 400+ individual image tiles, and high camera framerates which is problematic for techniques such as axial sweeping. (b-d) By comparison, the ExA-SPIM system uses 1000+ mW lasers, a large-format CMOS camera, and electronics metrology lenses with a moderate NA and ∼35 mm working distance. After expanded a mouse brain 3×, the system is still capable of imaging the entire brain in only 15 tiles.

Comparison of methods for large-scale volumetric imaging.

Summary plot comparing the resolution, isotropy, and imaging speed of various existing large-scale volumetric imaging methods. Methods which use sectioning (cyan) or do not require sectioning (yellow) are highlighted. The aspect ratio of each marker indicates the relative isotropy of the method.

Comparison of electronics metrology and life sciences technologies for large-scale volumetric imaging.

(a) Nikon 20× GLYC next to the VEO_JM DIAMOND 5.0× / F1.3 lens. (b) Traditional sCMOS camera with 2048×2048 pixels next to the VP-151MX camera with the Sony IMX411 sensor with 14192×10640 pixels.

Inverted and open-top ExA-SPIM designs.

Schematics for configuring the ExA-SPIM system in an (a) inverted or (b) open-top architecture. In both geometries, the working distance and mechanical housing of the lenses provides >1 cm of clearance, enabling imaging large tissue sections up to 1 cm thick. This geometry also reduces pathlengths through the tissue, reducing the demands on optical clearing and tissue clarity.

Photobleaching at high excitation laser powers.

The same region in an expanded mouse brain sample was irradiated repeatedly to measure photobleaching at the excitation powers used by the ExA-SPIM. Decay curves for a soma (blue) was measured over 2000 repeated exposures. Because the higher excitation power of the ExA-SPIM is distributed over an 11 mm wide light sheet, the light intensity is similar to typical standard SPIM (105 - 106 mW/cm2). Only modest photobleaching is observed, even after hundreds of exposures.

Factors to consider for expansion-assisted imaging.

A continuum of NA and expansion factor combinations can achieve a desired effective resolution. Light collection efficiency (which decreases quadratically with NA) and required working distance (which increases linearly with expansion factor) should be considered when deciding on exact parameters to use for expansion-assisted imaging.

Cleared-tissue imaging with the ExA-SPIM system.

(a) Although we have focused on imaging expanded tissues, the ExA-SPIM microscope does not require expansion. The chamber can be filled with any refractive-index matching media, and the liquid working distance of the electronics metrology lens can be adjusted slightly to recover diffraction-limited imaging performance. (b) As an example, an entire cleared mouse brain could be imaged coronally in a single tile.

ExA-SPIM mechanical layout.

CAD renderings of the microscope detailing (a) the complete system, (b) the detection assembly, (c) the illumination assembly, (d) the chamber assembly, and (e) the stage assembly.

ExA-SPIM laser combiner.

The system uses three 1000 mW lasers at 488, 561, and 639 nm (Genesis MX-STM series, Coherent) with an optional 405 nm laser. The beam from each laser (∼1 mm in diameter) first passes through a half waveplate for polarization rotation. The beams are all then combined using a series of dichroic mirrors mounted in kinematic mounts. An acousto-optic tunable filter (AOTF) is used to both select wavelength and modulate the power of each laser. The 0-order beam from the AOTF terminates in a beam dump, whereas the modulated 1st order beam is reflected off a final kinematic mirror before being injected into the microscope. An AR coated glass plate is used to reflect ∼0.5% of the output beam to a power meter for monitoring during dataset acquisition.

ExA-SPIM image acquisition scheme.

A block diagram summarizing the acquisition procedure for an ExA-SPIM dataset is shown. Each dataset requires looping over the total frames within a given tile, and then all tiles within a given dataset. Each tile results in a single file on disk, which is copied over the network to a local VAST storage system. The transfer of the previous tile occurs synchronously with the acquisition of the next tile, and the transfer speed to the VAST system outpaces the data generation speed of the microscope. For the ImarisWriter workflow, once tiles are transferred to the VAST storage system, they are synchronously converted to OME-Zarr files with ZSTD Shuffle compression. After compression, the files are written directly to cloud storage. The compression and conversion to OME-Zarr also runs at a speed which outpaces the microscope.

Simulations of camera sensitivity.

(a) Simulations of SNR versus collected photons for sCMOS (cyan) and the Sony IMX411 sensor with 16-bit (green), 14-bit (yellow), and 12-bit (red) readout. The SNR for an ideal perfect sensor is highlighted. (b) Relative SNR values for the curves shown in (a), normalized to the ideal sensor curve.