Benchtop mesoSPIM: a next-generation open-source light-sheet microscope for cleared samples

In 2015, we launched the mesoSPIM initiative, an open-source project for making light-sheet microscopy of large cleared tissues more accessible. Meanwhile, the demand for imaging larger samples at higher speed and resolution has increased, requiring major improvements in the capabilities of such microscopes. Here, we introduce the next-generation mesoSPIM (“Benchtop”) with a significantly increased field of view, improved resolution, higher throughput, more affordable cost, and simpler assembly compared to the original version. We develop an optical method for testing detection objectives that enables us to select objectives optimal for light-sheet imaging with large-sensor cameras. The improved mesoSPIM achieves high spatial resolution (1.5 µm laterally, 3.3 µm axially) across the entire field of view, magnification up to 20×, and supports sample sizes ranging from sub-mm up to several centimeters while being compatible with multiple clearing techniques. The microscope serves a broad range of applications in neuroscience, developmental biology, pathology, and even physics.

Supplementary Figure 11.The SPIM-tower sample holder and SPIM-mold: a standardized highthroughput imaging on the mesoSPIM system.a, CAD model of the SPIM-mold for reproduceable embedding of biological samples in agarose blocks.The mold has been optimized to allow for the positioning of 6 X. tropicalis tadpoles simultaneously in a standardized way.A pin on the SPIM-mold generates a negative hole for future mounting into the SPIM-tower.b, The resulting agar block with embedded sample (tadpole, position indicated by cartoon drawing).The larger part of the block is for sample mounting into the tower, and the smaller part contains the sample (arrow) and protrudes from the tower, like a small balcony.This design facilitates access of the light-sheet to the sample.c, One section (level) of the SPIM-tower holder, with 4 holes in the center for push-magnets, and 4 sample slots positioned for mounting agar blocks.Each slot has a mounting pin to fit securely into the hole in the agar block.This allows consistent and reproducible mounting of samples into the SPIM-tower.d, Four agar blocks containing samples, mounted onto one level of the SPIM-tower.The push-magnets allow stacking multiple levels to form a tower.e, Top view of the SPIM-tower with light-sheet illuminating the sample, and the detection objective shown.Consecutive levels are rotated by 45° from each other to avoid samples from different levels interfering with each other.f, The SPIM-tower in the Benchtop mesoSPIM during imaging.Each sample can be accessed by 90° rotation within the same level and Y translation (7.5 mm) + 45° rotation from one level to the next.g,h Visualization of the deviation in position of 4 stage-37 tadpoles (pseudo-colored) on one SPIM-tower level in the X-Y (g) and XZ (h) axes.Due to the SPIM-mold design and standardized agarose blocks, the X, Y deviations are within 1 mm and Z within 0.3 mm.* F-mount directly to the camera.Filter wheel (2"-filters) in front of the lens.Galvos require active (fan) cooling.** Requires Mitutoyo MT-1 (f=200m) or similar large-FOV tube lens.Filter wheel (1"-filters) behind the objective (infinity space).The BD version has larger diameter and mounting thread and requires 3D printed screen ring.Supplementary Table 3. Samples, labelling, clearing, and acquisition parameters.

Supplementary Table 4. sCMOS cameras supported by the mesoSPIM control software.
The camera needs programmable rolling shutter readout ("light-sheet") mode.

Model
where FOVx and FOVy are field of view dimensions on the sample side (camera chip dimensions divided by the system magnification), Poverlap = 0.9 is the tile overlap ratio in X and Y used for stitching (10%, default setting), ∆ is the z-step between planes.The total acquisition time is where FPS is the effective frame rate (frames/s).Fast mode.In this mode one can choose under-sampling in Z while still maintaining sufficient resolution for axons and dendrites detection.
Nyquist mode.This mode samples the axial PSF of the system at the optimal step, about 2 planes per PSF axial size, FWHMz.3) in fast mode takes 333 instead of 200 min and produces 419 instead of 897 GB of data.** Actual time is larger than (volume) average time because the the number of tiles in actual acquisition is always integer, which may be non-optimal for small volumes.At high magnifications and/or large volumes the difference between the two estimates becomes negligible.

Supplementary Table 8. Estimated acquisition time and file size for imaging 1 cm 3 sample on the Benchtop mesoSPIM
The code for calculating the acquisition speed and file size is available as Python notebook: https://github.com/mesoSPIM/image-processing/blob/main/notebooks/volumetric-speedcalculation.ipynb

Nuclear reactor neutrinos
To detect another elusive particle, the neutrino, ton-sized detectors or powerful accelerators are usually required to obtain a few events per day [15][16][17] .While elastic collisions of neutrinos with nuclei have just recently been observed using accelerator neutrino sources 16 , these interactions from reactor neutrinos have not yet been detected.The challenge is to bring a very sensitive detector close enough to the reactor.Small passive detectors, such as transparent crystals, could be more easily placed close to the reactor and work as nuclearnonproliferation safeguards 10 .In this application, crystals would be exposed to the high flux of neutrinos from the reactor and scanned with SPIM (ex-situ) to image neutrino-induced color centers.By counting the number of new centers in the crystal, it would be possible to understand the reactor power 10 .This knowledge could be then used to estimate (or exclude) the production of plutonium in the monitored reactor 10 .

Imaging of color centers
While the interaction of elusive particles (W, ) with the crystal lattice cause atom dislocations of only a few nanometers (Supplementary Figure 17), color centers can be probed with fluorescence microscopy at optical wavelengths.The fluorescence response of color centers enables faster and less expensive imaging than other techniques used to probe nm-sized features, such as transmission electron microscopy (TEM) and atomic force microscopy (AFM).The imaging of single color centers has already been performed with widefield and confocal fluorescence microscopes 18,19 .Imaging these color centers with SPIM offers further advantages: good optical sectioning compared to widefield imaging, less bleaching, higher throughput, and larger maximum sample size compared to confocal imaging.The non-destructive, fast and isotropic scan of large 3D-volumes offered by mesoSPIM is thus well suited for imaging large amounts of color-center based particle detectors.Another possible application of color center imaging with the mesoSPIM is fission track dating of transparent rocks: the mesoSPIM could offer a fast imaging method for large amounts of minerals without the need of etching, as described in Ref 12 .

Benchtop mesoSPIM imaging of particle-induced color centers in CaF2
In this work, we show the first tests of imaging particle-induced color centers with the mesoSPIM.This is part of the international collaborative work of PALEOCCENE (Passive lowenergy optical color center nuclear recoil) -a group of scientists working on the R&D of passive detectors of dark matter and neutrinos at low energy thresholds 11 .For the purpose of initial tests, we acquired transparent CaF2 crystals, irradiated them, and imaged them with the mesoSPIM, as described in the Methods section.The Supplementary Figure 12a shows the crystal ready for imaging before and while the light sheet is on.A custom-made crystal holder, shown in Figure 12b was made for keeping the samples in a stable position and allowing for changing crystals in consecutive scans without the need of complete refocusing.The fluorescence of an irradiated crystal in response to the 405 nm light sheet is clearly distinguished from the background, as shown by the contrast of the illuminated area in Figure 12c.While the fluorescence is mostly homogeneous across the crystal (presumably from many nm-sized color centers), a few clusters of high-intensity pixels appear at small scales (~10 µm).These structures are especially clear when imaged with the 20x/0.28objective.A few examples are shown in Figure 12d.The structures shown in Supplementary Figure 12d are identified by an algorithm, which scans the area of the illuminated ROI and outputs the contours of clusters of high-intensity pixels.The parameters used to define the selection thresholds of intensity and clustering are preliminary and will be benchmarked and improved in the future with data from dedicated ionirradiation (track-inducing) campaigns.
To verify whether the selected structures are intrinsic to the crystal -and not some random noise from the camera -repeated scans are compared.The matching correlation between them is calculated with the skimage.feature.template_matchingmethod, where the smallest rectangle enclosing the feature with N pixels is compared to the repeated scan using a normalized cross-correlation.The result ranges from -1 to 1, where 1 represents a perfect match.Most of the values are around zero, which is expected from random matching of any given group of N pixels which do not contain structures.The first two examples of the identified structures shown in Supplementary Figure 12d-e are observed at the same respective spots in the repeated scans and thus display large matching correlation coefficients (shown in red in e).Their matching coefficients are 7.9 and 7.3 sigma away from the mean of their respective distributions.The other values similar to the ones in red are due to the matching with neighboring z-planes, as the imaged structures span across a few z-scans.These distributions vary as the N number of pixels in the rectangles also vary.While this analysis shows that these "track-like" structures are intrinsic fluorescent features in the crystal, the origin of these features is not yet clear.One possibility is that the passage of particles such as cosmic rays may have created a track of color centers.Cosmic rays hit the Earth constantly, especially at high altitudes.The crystals were likely exposed to this natural irradiation since their production and more intensively while they were mailed by air after the purchase and for the irradiation campaigns.Despite the unknown origin of these features, this study demonstrates the capability of the benchtop mesoSPIM in identifying "track-like" structures of color centers.While the tracks from dark matter or neutrino interactions would have sizes below the resolution power of the mesoSPIM, understanding track formation is part of the R&D process of the concept.Furthermore, "the energy reconstruction of dark matter / neutrino interactions may be possible through the larger intensity of fluorescence (pixel brightness) measured from a region containing a full track in comparison to a single-site color center" 12 .The improved resolution provided at 20x magnification will be especially relevant when imaging single color centers or single tracks from nuclear recoils produced by neutrons -a proof-of-concept test planned as a next step.With the mesoSPIM images of the crystals irradiated with gamma rays, we performed further analyses with the aim of understanding: i) the distribution/homogeneity of color centers at larger scales (across milliliters of material), ii) the sources of background, and iii ) the color of the imaged color centers, and how the results obtained with the mesoSPIM compare to farfield spectroscopy.To quantify the level and homogeneity of fluorescence from blank and irradiated crystals, we imaged them at 1x magnification and calculated the average pixel intensity from every z-image of 350 scans taken at 10 µm steps inside the crystals.The obtained mean values and the respective gaussian standard deviations are shown in Supplementary Figure 12f.For this estimation, only pixels inside the illumination ROI #1 and at a distance larger than ~0.2 mm from the surface were selected to avoid surface background (as we found that the surface of blank crystals present a slightly higher signal intensity than their bulk).Possible sources of the higher surface fluorescents include dust or any other residual material which is autofluorescent and difficult to remove by usual cleaning methods, such as machine oils or surface coating, depending on the manufacturing of the crystal.The bulk of blank (not irradiated or before irradiation) crystals displayed low net fluorescence level, as observed in Supplementary Figure 12f: the fluorescence intensities in response to all the selected laser excitations were very close to the background level, which was estimated with the laser off.By imaging these crystals in the mesoSPIM with different laser excitations (405, 488, and 561 nm) and filters (quadrupole or long pass), we can understand the colors / wavelengths absorbed and reemitted as well as verify whether the signal corresponds to fluorescence or Raman scattering.Supplementary Figure 12f shows that irradiated crystals absorb 405 nm light and fluoresce in the blue: the most intense emission is in response to 405 nm light and faints when the blue spectrum is cut by a 515 nm long-pass filter.This blue fluorescence has been confirmed by the absorption and emission spectra measured in response to light from 250 to 800 nm with an Edinburgh Instruments FS5 spectrofluorometer.An example of the measured fluorescence from an irradiated crystal in response to 400 ± 10 nm light is shown Figure / movieOrganism, line Figure 5