2.1.

Animals

SV129 mouse embryos imaged in this study were generously provided by Dr. Quenten Schwarz, University of Adelaide. Pregnant mice were euthanized at E11.5 and the embryos were collected, decapitated, and fixed in ice cold 4% paraformaldehyde (PFA) (Electron Microscopy Sciences, Hatfield, Pennsylvania) diluted in phosphate buffer saline (PBS) (141 mM NaCl, 2.7 mM KCl, 10 mM ${\mathrm{Na}}_2{\mathrm{HPO}}_4$, and 1.5 mM ${\mathrm{KH}}_2{\mathrm{PO}}_4$) overnight. For nerve preparations, adult C57BL/6 were raised according to the guidelines of the animal facility at the Montreal Neurological Institute. Adult mice were euthanized and nerves were collected and postfixed for 2 h in 4% PFA in PBS. Samples were washed in PBS and stored at 4°C prior to use.

2.2.

Sample Preparation for Imaging

Whole mount fluorescent immunohistochemistry was performed according to a modified protocol.²³ Samples were dehydrated using a gradient of methanol (50%, 75%, 100%) diluted in PBS, incubating for 1 h per methanol solution. Autofluorescence was quenched by incubating samples overnight in an ice cold mixture of 70% methanol (Fischer Scientific, Geel, Belgium), 20% dimethylsulfoxide (DMSO) (Sigma, St. Louis, Missouri), and 10% hydrogen peroxide (Fisher Scientific, Fair Lawn, New Jersey).

Samples were rehydrated with methanol solutions in the reverse order (75%, 50%, and PBS) and fixed embryos were blocked at room temperature for 12 h in 6% donkey serum (Sigma, St. Louis, Missouri), and 0.1% Triton-100 (Sigma, St. Louis, Missouri) in PBS. For visualization of the spinal cord commissural projections, embryos were stained for Tag-1 (AF4439, R&D Systems, Minneapolis, Minnesota) 1:200 in wash solution (0.6% donkey serum, 0.1% Triton X-100 PBS) for 48 h at 37°C. Samples were vigorously washed for 12 h and subsequently incubated in Alexa 568 conjugated antigoat donkey (IgG A11056, Invitrogen, Eugene, Oregon) (1:250 in wash solution) for 36 h at 37°C. Finally, samples were washed for 12 h with PBS and kept at 4°C until imaging.

Nerve samples were prepared following a similar protocol. After methanol dehydration and peroxide bleaching, rehydrated samples were blocked at room temperature for 12 h in 10% goat serum (Sigma, St. Louis, Missouri) and 0.2% Triton-100 in PBS. Sciatic nerves were costained with mouse anti-βIII-Tubulin (801201, BioLegend, San Diego, California, 1:33 dilution) and rabbit SCG10 (Stathmin-2) (NBP149461, Novus Biologicals, Oakville, Canada, 1:200 dilution), in nerve wash solution (3% goat serum, 0.2% Triton-X, $10\mu \mathrm{g}/\mathrm{mL}$ heparin in PBS) for 5 to 7 days at 37°C. After extensive washing, samples were incubated with Alexa 488 conjugated antimouse IgG (Invitrogen, Eugene, Oregon) and Alexa 568 conjugated antirabbit IgG, at a dilution of 1:33 and 1:200, respectively, for 5 to 7 days at 37°C. Finally, nerves were washed for 12 h in nerve wash solution and 30 min in PBS before storage at 4°C until clearing.

Clearing was performed in accordance with a previously described protocol.²⁶ Fluorescently stained embryos or nerves were first dehydrated using a tetrahydrofuran gradient (50%, 75%, and 100%) (Sigma, St. Louis, Missouri) at room temperature, for 1 h per solution. Samples were then immersed in DBE (Sigma, St. Louis, Missouri) for at least 30 min prior to imaging.

2.3.

Cell Culture

Human embryonic kidney 293T cells were cultured in Dulbecco’s modified Eagle’s medium, supplemented with 10% fetal bovine serum, 4 mM L-glutamine, $100\text{units}/\mathrm{mL}$ penicillin, $0.1\mathrm{mg}/\mathrm{mL}$ streptomycin, and 0.1 mM nonessential amino acids (Gibco, Carlsbad, California). Cells were maintained in a humidified 5% ${\mathrm{CO}}_2$ atmosphere at 37°C. Cells were grown for 2 days on coverslips before imaging. At 24 h postplating, plasmids encoding transmembrane proteins attached to EGFP and mCherry were transfected into the cells using LTX Lipofectamine with the Plus Reagent, as specified by the manufacturer (Life Technologies, Carlsbad, California). Imaging was performed in media consisting of Hanks’ balanced salt solution and 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (Gibco, Carlsbad, California).

2.4.

Interchangeable Sample Holder Design

Samples holders were designed (SOLIDWORKS, Waltham, Massachusetts) and cut out of acrylic sheets (1.5, 3, and 6 mm thick, Johnston Industrial Plastics, Montreal, QC, Canada) using a Roland MDX-40A CNC milling machine (Roland DGA Corporation, Irvine, California). Sample holders are displayed below in Fig. 1.

For encapsulation of samples containing DBE, two parts are milled, aligned, and assembled using acetone (ACP, Montreal, QC, Canada). After the sample holder has dried ($\sim 10\min$), a coverslip (#1) is attached to the front side using clear nail polish, and left to dry overnight. The attached coverslip is then shaped using a diamond burr to fit the front side of the holder.

2.5.

Multimodal Light Sheet Setup with Dynamic Decoupling Between Excitation and Detection

We designed and built a highly modular system around a commercial microscope base creating a user-friendly light sheet microscopy system. We used a Nikon FN-1 upright microscope base (Nikon Canada Inc., Mississauga, ON, Canada), where the back left port of the NI-TT Quadrocular tilting tube on top of the microscope was used for detection. The microscope was equipped with halogen and mercury lamps for rapid sample visualization and positioning. An overview of the experimental setup for this microscope is shown below Fig. 2.

Fig. 2

Overview of the experimental setup. Layout showing the illumination path for brightfield (gray), fluorescence excitation and emission (dark blue, green, and red), and mercury lamp (cyan). Detection path is made of an objective, a $Z$ piezoscanner, dichroic (D1), flipping mirror (M1), and an EM-CCD camera to record the filtered fluorescence signal coming from the sample. The gray dotted line represents the software synchronization between the sample and the detection objective. Light sheet generation is achieved using an optical fiber, lens (L1), cylindrical lens, and a $10\times$ air objective or $40\times$ immersion objective.

Imaging was performed with three different detection objectives for a variety of resolutions and fields of view; CFI Plan Apo Lambda $10\times$ (0.45 NA, Nikon), CFI75 Apo LWD $25\times$ (1.1 NA, Nikon), and CFO Apo $40\times$ (0.8 NA, Nikon). A long travel objective piezoscanner (P-725.4CD, Physik Instrumente L.P., Auburn, Massachusetts) allowed rapid and reproducible movement of the objective lens. There are two main light pathways; 100% binocular and 100% camera port, with four different tube lenses ($0.6\times$, $1\times$, $1.5\times$, $2\times$). The whole setup was assembled on an optical breadboard table (Newport Corporation, Irvine, California).

For excitation, we designed a laser box with five OBIS diode lasers (405, 445, 488, 561, 604 nm, Coherent Inc., Santa Clara, California) to cover excitation wavelengths for the majority of fluorophores, with two fibered and one free space outputs. All five laser lines were combined using reflective and dichroic mirrors, and directed to one of the three outputs using switchable mirrors. For light sheet use, the laser lines were focused onto a fiber (OZ Optics Ltd., Ottawa, ON, Canada) and the fiber output was directed to an excitation arm, where light was collimated using a lens and then shaped into a sheet of light via a cylindrical lens placed before a CFO Apo $40\times$ (0.8 NA, Nikon) or CFI Plan Fluor $10\times$ (0.3 NA, Nikon) excitation objective.¹⁷ The excitation arms are assembled on a magnetic base to be easily interchangeable and are mounted on two linear stages to adjust the plane of light.

The sample holder module was placed on the right side of the microscope, opposite to the excitation arm, specially designed to fit the range of sample holders, Fig. 1, within the working distance of the appropriate objective. It was assembled on two sets of $X$, $Y$, and $Z$ linear motors: one manual set for rapid sample positioning at the point of convergence between the excitation and detection objectives, and a second automated set for reproducible movement within the sample and sample stitching for larger fields of view. Additionally, two linear ($X$ and $Z$) piezoactuators were added for fast 3-D imaging at nm step resolution. The system can also be modified for cuvette samples as shown in Fig. 3.

Fig. 3

Schematics of the different excitation arms developed for greater sample versatility. (a,b) Using the $10\times$ excitation objective described before. The $10\times$ objective can be used: (a) without or (b) with a cuvette with a cover slip on the side. (a) can be used with holders displayed in Figs. 1(d)–1(f), while (b) permits the use of Figs. 1(b) and 1(c) holders. No detection objective is shown on these two arms, because both can accommodate any detection objective due to a very long excitation working distance of 16 mm. (c) The $40\times$ excitation arm in combination with the $40\times$ detection objective. The cuvette was designed to hold a homemade PDMS O-ring in order to prevent leaks. A homemade support is used under the cuvette to prevent extra weight on the objective. In (a–c), the optical fiber is connected on the left side. (d) The excitation arm for the lattice light sheet using a custom objective for excitation¹² and the $25\times$ detection objective. This arm serves to bring the laser excitation in at the correct height after passing through a complex optical path also described previously.¹²

The entire setup, including excitation, detection, and all motorized parts, was controlled using LabVIEW and an NI-DAQ acquisition card (NI-PCIe-6353, National Instruments, Toronto, ON, Canada). The lasers, camera, and piezocontrollers are triggered using a digital or analog signal from the NI-DAQ for synchronization with the internal clock. Emitted photons are collected by the objective, filtered through a quad-band fluorescent cube (Chroma Technology Corp., Bellows Falls, Vermont), and detected on a $1024\times 1024$ Andor iXon Ultra 888 EMCCD camera (SnowHouse Solutions Inc., Lac-Beauport, Quebec).

A highly detailed supplementary file is available to any interested researcher via simple e-mail request. All parts are listed in this document with all CAD drawings and procedures for alignment and programming.

3.1.

Combining Multiple Light Sheet Modes into One System

An important goal in microscope development is to design a way to combine many different modalities into one system. In the specific case of light sheet microscopy, the objective working distance is a limitation, as it needs to accommodate the sample size while maintaining a large numerical aperture for detection purposes. In the generation of a light sheet, only a section of the sheet can be considered homogenous, restricting the field of view. Furthermore, better resolution in the $Z$ direction can only be achieved by decreasing the field of view, e.g., a light sheet thickness of $4\mu \mathrm{m}$ typically has a field of view of $300\mu \mathrm{m}$, whereas to attain a thinner $1\mu \mathrm{m}$ sheet, the field of view will be reduced to $25\mu \mathrm{m}$. To our knowledge, outside of high-end solutions such as isotropic multiview^10,11 or lattice light sheet¹² that can allow, respectively, a greater $Z$ or both better $Z$ and field of view, there is no solution to this trade-off. However, we introduce a very low cost simple adaptive solution which can be used on any sample size and desired resolution.

In our system, three different excitation arms are made easily interchangeable to fit the user’s needs, see Fig. 3. The first excitation arm employs the $10\times$ objective, Figs. 3(a) and 3(b), to provide a relatively large $300\mu \mathrm{m}$ field of view and a sizeable working distance of 16 mm, which can be used with any detection objective. The system can be easily adapted for imaging in a cuvette with a coverslip for side illumination, or for the homemade sample holders, shown in Fig. 1, to envelope specimens and protect objectives when used with organic solvents. However, this size excitation objective generates a sheet of $\sim 4\mu \mathrm{m}$.

The second excitation arm utilizes a $40\times$ objective [Fig. 3(c)] with a working distance of 3.5 mm, which can only be used with an identical $40\times$ detection objective. A cuvette is needed for the immersion of both objectives and a sample holder can be used to encase cleared samples in organic solvent, while protecting both objectives. The light sheet produced from this objective is $\sim 1\mu \mathrm{m}$ thick, with a relatively small field of view of $25\mu \mathrm{m}$.

3.2.

Decoupling Excitation and Detection Planes

Here, we present a straightforward solution to compensate for changes in refractive index that can be easily implemented on the majority of light sheet systems. Clearing techniques are required for the imaging of relatively large samples with fluorescence microscopy. However, these methods often employ solvents with optical indices significantly different from the immersion liquid, i.e., water. Our implementation aims to overcome this limitation and can be applied to liquids of any refractive index.

When imaging in media with a different refractive index than that required for the detection objective, the detection cone [dotted line in Fig. 4(a)] is modified [solid line in Fig. 4(a)]. The numerical aperture and working distance for detection will be altered depending on the distance needed to travel through each medium to reach the objective. This can become problematic when scanning the sample along the detection axis. The defocusing effect of an index mismatch is illustrated in Figs. 4(b) and 4(c). As light sheet microscopy decouples excitation and detection into two objectives, when moving the sample the excitation and detection planes will quickly dissociate from each other. To compensate for the discrepancy in working distance, we implement a facile solution by relocating the objective Fig. 4(d), which recovers the overlap needed between the excitation sheet and detection plane.

Fig. 4

Schematic of the optomechanical solution developed in order to image optically cleared samples in any solvent. (a) Schematic description of the sample holder used to protect objectives from the organic solvents. The specimen (red) is encapsulated in a custom designed holder (gray). Emitted light from the specimen passes through three media with different refractive indices before reaching the objective. The black line shows the light path if $n_1$ is air, $n_2$ is a borosilicate coverslip, and $n_3$ is DBE. The black dotted line illustrates the theoretical path if $n_1=n_2=n_3=1$. (b–d) The specimen and cover slip are moved together along the $Z$ detection axis. $\mathrm{d}Z$ represents the displacement along this axis and $\delta Z$ represents the correction for a given $\mathrm{d}Z$ movement using the piezoactuator. (b) The light sheet excitation plane (blue) is equivalent to the detection imaging plane (gray dotted lines) on the top of the specimen. (c) The specimen is moved up to acquire a different detection plane. Due to the change in refractive index, the excitation plane is no longer in the same position as the imaging plane, resulting in an unfocused image in Fig. 5(a). (d) To overcome this effect, we move the detection objective a calculated distance to recombine both planes.

3.3.

Rapid Correction of the Imaging Plane

For correction of the imaging plane in fast acquisition mode (20 to $50\mathrm{ms}/\text{frame}$), the objective needs to be rapidly repositioned for each frame. In our system, we implemented this via a piezoscanner for the objective to synchronize with the piezoactuator scanning the sample in the light sheet. Synchronization was made possible by simultaneously sending an analog signal via NI-DAQ to both piezosystems and a digital signal to the lasers and the camera.

For imaging in any refractive index medium, the compensation displacement of the objective is configured by the user determining the correct position for the piezoactuator for the top and the bottom of the $Z$ stack in order to obtain a focused image.

When these parameters are known for a given imaging medium, the user can utilize the same displacement or redefine the parameters. During our experiments, a typical correction for the $10\times$ objective in DBE is $\sim 85\mu \mathrm{m}$ for a $250\mu \mathrm{m}$ $Z$ stack and $\sim 25\mu \mathrm{m}$ for the $40\times$ objective. After user setup, the software simply divides the entire compensation displacement by $N$ number of planes in the $Z$ stack in order to apply the compensation during the imaging of each plane

By using these live corrections, users can image in any refractive index environment and we have demonstrated this in DBE, Fig. 5. We show the detrimental impact on imaging without the piezoelectric correction, exhibited after a small displacement of $50\mu \mathrm{m}$ in Fig. 5(a), and we present the acquisition with the correction from our technique in Fig. 5(b), revealing structures that were previously obscured. Users can acquire large stacks without a loss of information or the need to periodically refocus. Video 1 highlights the same corrective implementation for 3-D imaging of a mouse embryo.

Fig. 5

Active refractive index correction on murine embryonic spinal cord and dorsal root ganglion neurons stained for Tag-1. (a,b) $250\mu \mathrm{m}$ $Z$ stack acquisition on the same part of the same embryo (a) without and (b) with the plane-correction, which uses the displacement of the objective to correct for the high-refractive indices (DBE, $n=1.56$). (c) A max projection was applied to highlight the axons. There is a noticeable difference in the fine commissural fibers as they can be observed from the inner part of the spinal cord without (left) and with (right) correction, obtained by reslicing the $Z$ stack in (a/b) in the $X$ and $Y$ directions to view the back of the embryo. The correction allowed us to observe fine structures of these commissural tracts in detail (red arrows) and their parallel disposition in the ventral midline of the neural tube. $X$ and $Y$ resolution with this objective is limited by pixel size which is 650 nm. The $Z$ resolution is limited by the light sheet axial dimension measured to be between 3.5 and $4\mu \mathrm{m}$ depending on the alignment and the $Z$ stack is oversampled every $1\mu \mathrm{m}$ (Video 1, mp4, 5,689 KB) [URL: http://dx.doi.org/10.1117/1.JBO.21.12.126008.1].

3.4.

Acquisition of an Entire Specimen

When using the excitation arm equipped with the $10\times$ objective, the field of view is considered to be $\sim 300\mu \mathrm{m}$ (i.e., homogeneous part of the light sheet) by $650\mu \mathrm{m}$ (camera field of view) and the maximum movement of the piezoelectric actuator with respect to the $Z$ stack is $250\mu \mathrm{m}$. In order to record the entire specimen (i.e., 12.5-day mouse embryo, 4 $4\times 3\times 3\times {\mathrm{mm}}^3$), the acquisition needs to take place over multiple fields of view. For this, we use three motorized stages to take $Z$ stacks at different positions, crop each stack to keep the thinnest homogeneous part of the light sheet, and stitch the $Z$ stacks together as shown in Fig. 6.

Fig. 6

Schematic for data treatment; process and stitching for a full 3-D rendering of a mouse embryo. Multiple full frame $Z$ stacks of $250\mu \mathrm{m}$ are acquired following the order shown in (a) (where each square represents a $250\mu \mathrm{m}$ stack), in multiple $X$, $Y$, and $Z$ positions. (b) For each $X$ and $Y$ position, the $Z$ positions are concatenated into a single $Z$ stack. (c) All $Z$ stacks are cropped in $X$ and $Y$, to only retain data collected from the thinnest part of the light sheet (d). (e) The cropped stacks are then stitched according to the motor displacements using Fiji stitching plugin.²⁷ (f) The entire specimen is rendered in Imaris. This 11.5-day mouse embryo sample was $1000\times 2500\times 1250\times {\mu \mathrm{m}}^3$. For dynamic view of the mouse embryo, see (Video 2, mp4, 14,198 KB) [URL: http://dx.doi.org/10.1117/1.JBO.21.12.126008.2]. $X$ and $Y$ resolution with this objective is limited by pixel size which is 650 nm. The $Z$ resolution is limited by the light sheet axial dimension measured to be between 3.5 and $4\mu \mathrm{m}$ depending on the alignment, and the $Z$ stack is oversampled every $1\mu \mathrm{m}$.

Automated data treatment for the acquisition of the entire specimen was performed using the Fiji stitching plugin²⁷ and 3-D rendered using Imaris (Bitplane, South Windsor, Connecticut). For the sample presented in Figs. 5 and 6, acquisition time was around $10\min /{\mathrm{mm}}^3$. The acquisition time varies depending on the integration time and camera mode.

3.5.

Neurobiology Applications

Spinal cord commissural neurons are an important model system for axon guidance in mammals. This population of neurons projects their axons radially toward the ventral midline, where they cross to the opposite hemicord and change their growth direction to the higher brain regions.²⁸ These directional changes make it difficult to quantify the different possible phenotypes for these axons. One limitation in the field is the need to physically slice the samples and perform immunostaining for the neuronal populations of interest, but this has recently been overcome by the development of clearing techniques.^23,26,29,30

As hundreds of thousands of commissural neurons project in parallel, only an imaging system capable of resolving very fine structures would allow us to study the neuronal projections. This can be achieved by reducing the light scattering of the sample through clearing using a range of different chemicals that act by equilibrating refractive indices through the tissue.³¹ Each clearing method uses a mounting medium with a specific refractive index³¹ and this often poses a problem for microscope objectives, which are usually designed for a single refractive index.

As can be observed in Fig. 5(a), the structure of thick spinal cord commissural axons ($\sim 5$ to $10\mu \mathrm{m}$) is lost as we advance deeper in the DBE-immersed tissue. By mechanically correcting the focal plane deviation caused by DBE, we are able to visualize the crossing fibers in the midline, Fig. 5(c). To our knowledge, this is one of the sharpest images of the developing commissural spinal cord tracts. Using our method for focus correction, we were able to image all of the spinal cord commissural neurons using the general marker Tag-1, at a similar resolution to what has been previously published with neuronal subpopulation markers.²⁹ Thus, the imaging improvement brought by mechanical focal correction allows for a detailed inspection of spinal cord phenotypes without incurring a cell population bias in order to resolve the axons.

In addition, clearing techniques enable the observation of whole mouse embryos without tissue sectioning, Fig. 6. As the neurons in the spinal cord differentiate in waves, in which the cells at the brachial level mature earlier than the caudally located neurons,³² it is advantageous to be able to resolve the complete structure. In comparison to commercially available light sheet microscopes, mechanical focal correction can be used to reach the same resolution without specific sample preparation constraints and index compatibility.

3.6.

Nerve Imaging

The benefits of our correction method are readily apparent when applied to nerve imaging. Understanding how nerves develop and regenerate is a major challenge for the neuroscience field. The current gold standard method for this analysis is the manual sectioning of nerves in transverse or longitudinal sections and subsequent reconstruction. For a rodent nerve, that usually implies combining images of hundreds of slices from single nerves in order to find growing fibers. In such cases, one nerve can have a diameter of 0.5 mm and length of several centimeters. As the sectioning process transposes a 3-D system to two-dimensional fragments, it is susceptible to many interpretation mistakes, including the confusion of spared and regenerating fibers.³³

While multiphoton and point scanning confocal microscopy have been used together with tissue clearing, these techniques are time consuming and can give rise to a significant bleaching problem.³⁴ These drawbacks can be overcome with spinning disk microscopy if pinhole crosstalk did not pose a challenge for scanning deep tissues.³⁵ To our knowledge, some of the best results for the imaging of regenerating nerves in whole mount preparations have been obtained by light sheet imaging of cleared tissue using organic solvents.^36,37 Here, we show that our system is capable of imaging both DBE- immersed sciatic nerves, which are important models for the study of the peripheral and central nervous systems, respectively. We can observe the expression of neuronal markers, such as β-III-tubulin, to follow axon morphology across an entire nerve, as shown in Fig. 7. Regenerating fibers can be identified by Stathmin-2 expression, a marker of growing and sensory axons,³⁸ also shown in Fig. 7. Our system could enable us to count and trace growing and regenerating fibers throughout the extension of the nerves.

Fig. 7

(a,b) 3-D rendering of light sheet images representing the reconstruction on an intact mouse sciatic nerve, fluorescently tagged with Alexa 488 (β-III-Tubulin) and Alexa 584 (Stathmin 2). The entire sciatic nerve sample is $1050\times 7900\times 470{\mu \mathrm{m}}^3$. Total acquisition time in two color imaging was 1 h, including 36 positions for stitching. (c,d) The resolution of the light sheet enabled visualization of separate nerve fibers. (c) Max projection of several images of a stack showing optically separated nerve fibers. (d) Single image projection of a perpendicular nerve plane (Video 3, mp4, 14,197 KB) [URL: http://dx.doi.org/10.1117/1.JBO.21.12.126008.3] shows a 3-D reconstruction of the nerve. $X$ and $Y$ resolution with this objective is limited by pixel size which is 650 nm. The $Z$ resolution is limited by the light sheet axial dimension measured to be between 3.5 and $4\mu \mathrm{m}$ depending on the alignment and the $Z$ stack is oversampled every $1\mu \mathrm{m}$.

Altogether, light sheet microscopy consists of a practical and accessible method to image whole mount preparations. Combined with paradigms of nerve lesion, this system can be used to accurately assess nerve regrowth in future studies.

3.7.

Imaging of Live Single Cells

Equipping the light sheet microscope with $40\times$ objectives for excitation and detection, alongside the 45 deg sample holder, Fig. 1(b), single cell imaging was achieved at a 3-D acquisition timescale of ms. However, using this objective, the field of view was limited to 20 to $30\mu \mathrm{m}$. The piezoelectric actuators were completely synchronized, as one was used to scan the sample in the $Z$ direction, while the other simultaneously moves the sample along the excitation axis in order to center the sample at the thinnest part of the light sheet for the best resolution. Images of single cells are shown in Fig. 8.

Fig. 8

Image of live HEK293T cells using the $40\times$ objectives for both excitation and detection. The area between the white lines represents the area that is subjected to a homogenous light sheet ($\sim 25\mu \mathrm{m}$). Crop of the camera field of view to show the effect of the light sheet edges. To acquire a $Z$ stack, the $X$ and $Z$ piezosystems are used simultaneously to image on the 45 deg coverslip and for the thinnest part of the light sheet to be kept in the middle of the frame. To reconstruct the $Z$ stack, the central part should be cropped and realigned, as the images are shifted by the $X$ piezosteps.

3.8.

Implementation of Further Corrections

For compensation of the defocusing effect when stitching large specimens, the excitation arms can be further motorized along the excitation axis, analogous to our solution employed to correct for excitation in high-refractive index media. In our system, the light sheet displacement along the excitation axis was almost negligible due to the numerical aperture and working distance of the objective used, therefore, we did not need to motorize this part. However, other objectives would need this correction. The cropping of data to include only sampling from the thinnest part of the light sheet enables the best resolution possible.

Aberrations due to material scattering and absorption (i.e., stripes) can be avoided by the use of two light sheets, generated by two objectives on different sides of the sample,⁸ by pivoting the cylindrical lens,⁸ by implementing a rotational stage,^7,19,20 or by software treatment.³⁹ Currently, we are adding a third excitation arm, Fig. 3(d), to our modular system to enable the generation of a lattice light sheet, for improved resolution in $Z$ while keeping a reasonable field of view.¹²