Access the full text.
Sign up today, get DeepDyve free for 14 days.
Background: Advances in tissue clearing and molecular labeling methods are enabling unprecedented optical access to large intact biological systems. These developments fuel the need for high-speed microscopy approaches to image large samples quantitatively and at high resolution. While light sheet microscopy (LSM), with its high planar imaging speed and low photo-bleaching, can be effective, scaling up to larger imaging volumes has been hindered by the use of orthogonal light sheet illumination. Results: To address this fundamental limitation, we have developed light sheet theta microscopy (LSTM), which uniformly illuminates samples from the same side as the detection objective, thereby eliminating limits on lateral dimensions without sacrificing the imaging resolution, depth, and speed. We present a detailed characterization of LSTM, and demonstrate its complementary advantages over LSM for rapid high-resolution quantitative imaging of large intact samples with high uniform quality. Conclusions: The reported LSTM approach is a significant step for the rapid high-resolution quantitative mapping of the structure and function of very large biological systems, such as a clarified thick coronal slab of human brain and uniformly expanded tissues, and also for rapid volumetric calcium imaging of highly motile animals, such as Hydra, undergoing non-isomorphic body shape changes. Keywords: Light sheet microscopy, Whole brain imaging, Quantitative imaging, Hydra, Calcium imaging, Tissue clearing, Expansion microscopy Background demonstrated ). These approaches have the potential Advances in tissue clearing methods  are enabling un- to accelerate discoveries across multiple domains of life hindered optical access to the structure and function of sciences, including an understanding of the mammalian large intact biological systems such as mouse brain [2– brain architecture, reconstructing tumor microenviron- 5] and tumor biopsies . Most of these approaches em- ments, and in situ transcriptomics. However, taking full ploy a cocktail of chemicals for cellular membrane lipid advantage of these techniques requires rapid high- dissolution and/or refractive index smoothening to ren- resolution three-dimensional (3D) imaging of very large der the tissue transparent . Furthermore, the develop- volumes. ment of physical tissue expansion approaches (expansion Conventional point-scanning approaches, such as con- microscopy, ExM ) is enabling higher (super-reso- focal and two-photon microscopy, provide high imaging lution) effective imaging resolutions, although at the cost quality, but their slow imaging speeds and high photo- of ever-increasing sample sizes (up to 20-fold expansion bleaching rates render them less effective for imaging of large volumes. Variants of confocal microscopy, includ- * Correspondence: email@example.com ing line scanning confocal microscopy (LSCM) [9, 10], Bianca Migliori and Malika S. Datta contributed equally to this work. can provide much higher imaging speeds due to parallel Department of Biological Sciences, Columbia University, New York, NY, USA imaging of multiple points. However, these approaches NeuroTechnology Center, Columbia University, New York, NY, USA Full list of author information is available at the end of the article © Tomeret al. 2018 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Cm Migliori et al. BMC Biology (2018) 16:57 Page 2 of 19 still entail highly redundant illumination of out-of-focus single-plane illumination and simultaneous whole-plane parts of the samples and also reduced axial resolution and detection, are proving to be highly effective due to min- imaging depth, thus limiting their utility for high- imal photo-bleaching and high imaging speeds (2–3or- resolution imaging of large cleared samples such as whole ders of magnitude more than confocal) [11–13]. However, mouse brains (Fig. 1a). On the other hand, light sheet mi- these advantages of orthogonal illumination-detection croscopy (LSM)-based approaches, with their orthogonal geometry also require unhindered optical access from the Standard LSM Inverted LSM SCAPE/OPM Line Scan Confocal LSTM Cm Cm Limits on lateral Limits on imaging Low quality off-focal Detection-coupled line scan: Detection-independent sample dimensions depth and quality plane signal low axial resolution and large-angle thin high photo-bleaching light-sheet illumination bc Synchronized Planar illumination & detection Modes of line scanning Exposed Pixel line 1-Axis Scan Simultaneous 2-Axes Scan Synchronized Focal Plane LSTM uniform optical sectioning LSTM-1AS LSTM-2AS DAPI 0 0.5 1.0 1.5 2.0 0 0.2 0.4 0.6 0.8 1.0 Position (mm) in field-of-view Depth (mm) Fig. 1 Light sheet theta microscopy (LSTM) for high-resolution quantitative imaging of large intact samples. a Light sheet microscopy (LSM) employs orthogonally illumination-detection optics, which limits the lateral dimensions of imaging volume. iSPIM, SCAPE/OPM and line scan confocal microscopy are partially effective in alleviating this limitation, however at the cost of reduction in usable working distance (magenta arrowheads)and image quality (e.g., SCAPE collects low-quality signal from non-native focal planes, and line scan confocal results in lower axial resolution and high photo-bleaching.). The proposed LSTM uses non-orthogonal (< 90°) illumination light sheets to effectively image very large samples, while maintaining high imaging speed and depth and uniform high resolution. b One or two light sheets intersect with the detection plane in a line illumination profile, which is synchronously scanned with the rolling shutter detection of an sCMOS camera to achieve optical sectioning. c Two scanning approaches: 1-axis scanning (1-AS) by perpendicular translation and simultaneous 2-axis scanning (2-AS, default LSTM) by translation along and perpendicular to the illumination axis such that the thinnest part is utilized for uniform planar illumination. d Comparison of point spread function (PSF) in 1-AS, default LSTM, and LSM configurations. Left: x-z maximum intensity projections of ~ 1 μm fluorescent microbeads imaged using the same detection (10×/0.6NA/8mmWD) and illumination (4×/0.28NA/28.5 mmWD) objectives. Axial full width at half maximum values (FWHM) across the field of view (blue LSTM in default 2-AS mode, green LSTM in 1-AS mode, red LSM). LSTM achieves uniform axial resolution (~ 4–6 μm FWHM) over the entire field of view, whereas both the 1-AS and LSM provide lower peripheral resolution (1-AS ~ 5–13 μm; LSM ~ 4–11 μm). Right: x-z projections (20 μm) of an image volume from a DAPI-stained human brain tissue. Additional file 5: Video 2  provides 3D reconstructions. The graph compares the signal for a central and a peripheral region of interest. Scale bars: 100 μm FWHM (µm) LSM LSTM-1AS LSTM-2AS Mean Intensity Migliori et al. BMC Biology (2018) 16:57 Page 3 of 19 sides of the samples, thus limiting the lateral dimensions using independent illumination objectives, for rapid (along the illumination light sheet) of the imaging volumes high-resolution imaging of large samples (Fig. 1). The (Fig. 1a). For example, we previously reported an opti- oblique optical arrangement eliminates the restrictions mized implementation of LSM, called CLARITY opti- on the sample lateral dimensions while ensuring high mized light sheet microscopy (COLM) [5, 14], which imaging speed and resolution and utilization of the allowed high-resolution imaging of entire intact mouse entire available WD of high-NA detection objectives brains in a few hours imaging time, although with pro- (Fig. 1a). We present a detailed characterization of gressively reduced image quality towards the middle of the LSTM approach and demonstrate several real- the samples  due to the scattering of illumination light world high-resolution imaging examples of very large sheets. Similar attempts of LSM imaging of clarified rat samples including mouse and rat brain tissues, a large brains resulted in much poorer image quality in large section of human brain, and a highly expanded ExM parts of the brain . sample. The new capabilities of LSTM for high-speed Alternative optical configurations of LSM have been quantitative imaging of larger samples at high reso- explored to address these limitations, including the lution with low photo-bleaching may facilitate map- rotation of the illumination and the detection axes by ping of an entire post-mortem human brain (thick 45° relative to the sample surface normal in objective- slab-by-slab) in a practical time-frame. coupled planar illumination (OCPI) microscopy, inverted selective plane illumination microscopy (iSPIM), dual Results view iSPIM (diSPIM), and triple-view implementations LSTM implementation [16–21], and the generation of illumination light sheets LSTM includes a standard wide-field detection arm and through the detection objectives themselves in swept two symmetrically arranged non-orthogonal (θ < 90°, confocally aligned planar excitation (SCAPE)/oblique relative to the detection axis) illumination arms for the plane microscopy (OPM) [22, 23]. The iSPIM/diSPIM generation of thin static light sheets that intersect at the approach does alleviate the limits on the lateral detection focal plane (Fig. 1). The resulting thin line illu- dimensions of imaging volumes, although at the cost of mination profile is scanned along the detection focal significant reduction in the usable working distance of plane in synchrony with the row-by-row rolling shutter the detection objective (Fig. 1a); therefore, it remains imaging with a scientific complementary metal-oxide- restricted to relatively low-numerical aperture (NA)/ semiconductor (sCMOS) camera (virtual slit effect ) long-working distance (WD) detection objectives for to achieve thin optical sectioning (Fig. 1b). This is real- imaging of large samples. The triple-view approach  ized by simultaneously translating the light sheet along incorporated an additional objective in the diSPIM im- (using an electrically tunable lens (ETL)) and perpen- plementation for simultaneously detecting the obliquely dicular (using a galvanometer (galvo) scanner) to its illuminated plane from the opposite side by rapid scan- propagation direction such that only the thinnest part ning with the piezo motors, resulting in enhanced spatial intersects the detection plane (Fig. 1c). Note that the resolution for small samples (such as single cells). The translation of light sheets along their propagation direc- SCAPE/OPM implementations use rotation optics to tion may also be achieved using alternative implementa- image an oblique plane illuminated using an oblique tions, including fast piezo motors for translating light sheet generated through the detection objective illumination objectives, or using an acoustic tunable lens itself (Fig. 1a). The use of a single objective for detec- such as the tunable acoustic gradient index of refraction tion as well as illumination is effective for fast volu- (TAG) lens (TAG optics) [14, 24–28]. metric imaging of small samples such as developing The LSTM illumination and detection arms were im- embryos. However, the imaging of an oblique (relative plemented as rigid assemblies (using a caging system to the native detection focal plane) plane provides from Thorlabs) which were connected to a vertically less-than-optimal image uniformity (across the imaged mounted breadboard via x-y manual translation stages plane) and resolution. In summary, the scaling up of for finer adjustments (Fig. 2, Additional file 1: Figure S1, LSM imaging volumes, while maintaining uniform Additional file 2: Figure S2, Additional file 3: Figure S3, high imaging quality and speed, faces steep chal- Additional file 4: Video 1). An open-top sample mount- lenges. Here we address some of these challenges by ing strategy was implemented by using a custom developing a conceptually distinct microscopy frame- 3D-printed sample chamber (Fig. 2, Additional file 1: work, termed light sheet theta microscopy (LSTM), Figure S1) attached to a high-accuracy x-y-z motorized which builds upon the principles of LSM to allow stage assembly (LNR50S, Thorlabs). The imaging samples high-speed quantitative imaging of large intact tissues can be mounted in a quartz glass cuvette of appropriate with uniform high resolution. The LSTM uses two size. The illumination arm consists of a laser source symmetrically arranged oblique light sheets, generated (SOLE-6, 405, 488, 561, and 647 nm, Omicron-Laserage), a Migliori et al. BMC Biology (2018) 16:57 Page 4 of 19 Fig. 2 LSTM microscopy implementation. a LSTM optical path. Two symmetric light sheets are generated by using a cylindrical lens (CL), scan lens (SL), tube lens (TL), and illumination objectives. The galvo scanners are used to translate the light sheets perpendicular to their propagation direction, and the electrically tunable lens (ETL) for translating the thinnest part of the light sheets along the propagation direction. An input beam of ~ 10 mm diameter is then trimmed through an iris. A slit is placed after the ETL to control the effective numerical aperture of the illumination. An additional iris is placed between the SL and TL to control the light sheet width. The illumination axes are arranged at ~ 60° to the detection axis. A custom 3D-printed cap with a quartz coverslip is attached to the illumination objective to allow dipping in the immersion oil to ensure that the low NA illumination rays from an air illumination objective enter perpendicularly to the oil. The detection arm consists of a detection objective (Olympus 10×/0.6NA/8mmWD or 25×/1.0NA/8mmWD, both with correction collars), a tube lens, and an sCMOS camera. b 3D model of the LSTM microscope. A vertical breadboard was used to mount the caged optical assemblies via x-y manual translation stages to allow fine adjustments. A sample chamber was attached to a 3-axis (x, y, z) motorized stage assembly. See also Additional files 1, 2, 3, and Additional file 4: Video 1 for further details and Table 1 for complete parts list Migliori et al. BMC Biology (2018) 16:57 Page 5 of 19 collimator (~ 10-mm beam diameter, Omicron-Laserage), an approximate effective WD in immersion oil was an ETL (Optotune), a cylindrical lens (LJ1695RM-A, f = calculated as shown in Additional file 3: Figure S3. 50 mm, Thorlabs), a galvo scanner (GVS001, Thorlabs), a scan lens (CLS-SL, Thorlabs), a tube lens (f = 200 mm, LSTM characterization ITL200, Thorlabs), and an illumination objective (Macro We first characterized the LSTM point spread function 4×/0.28NA/28.5mmWD, Olympus; the arrangement of dif- (PSF) by imaging micrometer-size fluorescent microbe- ferent components is summarized in Additional file 2:Fig- ads. The same microbeads were also imaged with LSM ure S2, and the parts list is provided in Table 1). Note that as well as a non-optimal 1-axis LSTM scanning proced- even though the ETLs are mounted at an oblique angle (as ure (i.e., the light sheet is only translated perpendicular opposed to vertical), this does not result in any significant to its propagation; marked as 1-AS mode in Fig. 1c), observable aberrations because of the low NA of the illu- using the exact same detection (10×/0.6NA) and mination objectives. In addition, an iris is placed after the illumination objectives (4×/0.28NA). The quantification collimator to remove peripheral spreads of Gaussian beams, and comparison of full width at half maximum (FWHM, a one-dimensional (1D) slit is positioned before the cylin- Fig. 1d) revealed that LSTM indeed allows for uniform drical lens to control the effective NA (hence the light sheet high axial resolution (~ 4 to 6 μm for the combination thickness), and a second iris is placed at the conjugate plane of these objectives) across the entire field of view, between the scan lens and tube lens to control the light whereas as expected, both the LSM and the non-optimal sheet width (i.e., the dimension of light sheet perpendicular 1-AS LSTM scan procedure resulted in lower axial reso- to its propagation direction). The detection arm is composed lution on the peripheries of the field of view (> 11 μm) of a detection objective (Olympus 10×/0.6NA/8mmWD or (Fig. 1d, Fig. 4 and Additional file 5: Video 2 ). Next, 25×/1.0NA/8mmWD, both with refractive index correction we compared the maximum illumination path length collars, from water to oil), a multi-band emission filter (MIPL), i.e., the maximum distance the illumination (FF01-432/515/595/730-50-D, Semrock), a tube lens (f = light sheets need to penetrate inside the tissue to achieve 200 mm, ITL200, Thorlabs), and an sCMOS camera (Orca complete sample coverage. The shorter the illumination Flash 4.0, Hamamatsu; 2048 × 2048 pixels, 6.5 μm× 6.5 μm path length, the lesser the scattering, hence potentially pixel size). Note that since we used a 200-mm tube lens with better the imaging performance. In LSM, the MIPL de- Olympus objectives, the effective magnifications are 11.11 pends on the sample width as the light sheet needs to and 27.78 for the 10× and 25× objectives respectively. The penetrate the entire width (half width for two-sided illu- LSTM assembly was optically aligned by placing a prism mination) of the sample to achieve complete coverage, mirror (with scratches in the center, see Additional file 1:Fig- whereas in LSTM, the MIPL depends on the angular ure S1 for mounting arrangements) in the focal plane of the separation of the illumination-detection arms and the detection optics, to visualize the location and cross section of tissue thickness (t): t/cos(θ) (Fig. 3b). The ratio of LSTM the light sheet relative to the detection focal plane. The light and LSM illumination path lengths was calculated for sheet positioning parameters (i.e., galvo and ETL) were varying sample dimensions and the θ values (Fig. 3b optimized such that the thinnest part was in alignment with shows θ = 60°, and Additional file 6: Figure S4 shows θ the center of the field of view of the detection plane. Next, = 0°, 10°, 20°, 30°, 40°, 50°, 60°, 70°, and 80°). This re- fluorescent beads, embedded in a high concentration (> 2%, vealed that the LSTM illumination path length was to restrict the signal source to the gel surface) agarose gel, smaller than that for LSM for wider samples, and larger were used to find the optimal galvo and ETL parameters by for smaller samples, suggesting the complementary ad- examining the extent and focus quality. All imaging experi- vantages of LSTM and LSM for high-resolution imaging ments were performed using 405 (for 4,6-diamidino-2-phe- ’ of large and small samples respectively. Since the LSTM nylindole (DAPI)) or 488 nm (for eYFP detection) laser lines. illumination path length decreases with decreasing angu- LSTM parameters were adjusted to use a 2 to 5 μmeffective lar separation (t/cos(θ)), minimizing the angular separ- light sheet thickness. ation (θ) will result in potentially higher imaging quality; The angular separation of the illumination and detec- however, the angular separation also affects the effective tion arms is constrained by the physical dimensions and light sheet thickness (approximated as b/sin(θ), Fig. 3c) optical properties of the detection and illumination ob- in an inverse relationship. Therefore, we used the max- jectives. For instance, only an angular separation range imum allowed angular separation (~ 60°) to achieve bet- of 43–62 is feasible for the specific combination of ter axial resolution. illumination (Macro 4×/0.28NA/29.5mmWD, Olympus) Further, we sought to assess and compare the total and detection objectives (10×/0.6NA/8mmWD or 25×/1. energy loads in LSTM vs. LSM imaging (Fig. 3d–f, 0NA/8mmWD, Olympus) used in this study (see Fig. 3a, Additional file 7:FigureS5).Since LSTM employs Additional file 3: Figure S3). Note that the WD of the only the thinnest part of the illumination light sheet illumination objective is specified for use in air; therefore, for imaging, this results in significant redundancy in Migliori et al. BMC Biology (2018) 16:57 Page 6 of 19 Table 1 Parts list of LSTM Vendor Number Qty. Description Detection unit Thorlabs CXY2 1 60-mm Cage System Translating Lens Mount for Ø2” Optics Thorlabs LCP90F 1 60-mm Removable Cage Plate Thorlabs SM2A20 1 SM2-M38 Adapter for Nikon Tube Lens Thorlabs SM2L30 2 SM2 Lens Tube, 3” Thread Depth, One Retaining Ring Included Thorlabs LCP09 2 60-mm Cage Plate with Ø2.2″ Double Bore for SM2 Lens Tube Mounting Thorlabs ER10 4 Cage Assembly Rod, 10″ Long, Ø6 mm Thorlabs SM1A1 1 Adapter with External SM05 Threads and Internal SM1 Threads Thorlabs SM2A31 1 Adapter with External C-Mount Threads and Internal SM2 Threads Hamamatsu C13440 1 sCMOS Orca Flash 4.0 V3.0 camera / Objective Adapter 1 Custom-made Adapter from SM2 to M34 threading Thorlabs SM2V10 1 Ø2” Adjustable Lens Tube, 0.81″ Travel Olympus XLPLN10XSVMP 1 10× Long Working Distance Detection Objective Olympus XLSLPLN25XGMP 1 25× Long Working Distance Detection Objective Thorlabs LCP01B 2 60-mm Cage Mounting Bracket Thorlabs RSH2 2 Ø1” Post Holder with Flexure Lock, Pedestal Base, L = 2” Thorlabs RS2 2 Ø1” Pillar Post, 1/4″-20 Taps, L = 2″,8–32 Adapter Included Thorlabs TBB0606 2 Large-Area Translation Stage, 6″ × 7.66” Thorlabs TTL200 1 f = 200 mm Tube Lenses for Wide Field Imaging Semrock FF01-432/515/595/730-50- 1 Multi-Band Emission Filter Illumination unit Thorlabs SM2V10 2 Ø2” Adjustable Lens Tube, 0.81″ Travel / Objective Adapter 2 Custom-made Adapter from SM2 to M34 Threading Olympus XLFLUOR4X/340 2 4× Air Objective Thorlabs SM2A20 2 SM2-M38 Adapter for Nikon Tube Lens Thorlabs CXY2 2 60-mm Cage System Translating Lens Mount for Ø2” Optics Thorlabs SM2A31 2 Adapter with External C-Mount Threads and Internal SM2 Threads Thorlabs SM2V10 2 Ø2" Adjustable Lens Tube, 0.81" Travel Thorlabs LCP09 2 60-mm Cage Plate with Ø2.2” Double Bore for SM2 Lens Tube Mounting Thorlabs LCP01B 4 60-mm Cage Mounting Bracket Thorlabs RS2 4 Ø1” Pillar Post, 1/4″-20 Taps, L = 2″,8–32 Adapter Included Thorlabs RSH1.5 4 Ø1” Post Holder with Flexure Lock, Pedestal Base, L = 1.5” Thorlabs TBB0606 4 Large-Area Translation Stage, 6″ × 7.66” Thorlabs ER05 8 Cage Assembly Rod, 1/2″ Long, Ø6 mm Thorlabs LCP02 6 30-mm to 60-mm Cage Plate Adapter, 8–32 Tap Thorlabs LJ1695RM-A 2 Ø1”, N-BK7 Mounted Plano-Convex Round Cylindrical Lens Thorlabs CRM1L 2 Cage Rotation Mount for Ø1” Optics, Double Bored with Setscrew, 8–32 Tap Thorlabs CP20S 2 30-mm Cage System Iris, Ø20.0-mm Maximum Aperture Thorlabs CP90F 2 30-mm Removable Cage Plate, Front and Back Plate, Internal SM1 Threading Thorlabs CXY1 2 30-mm Cage System, XY Translating Lens Mount for Ø1” Optics Thorlabs CP12 2 30-mm Cage Plate, Ø1.2″ Double Bore for SM1 Lens Tube Mounting Thorlabs LCP01 4 60-mm Cage Plate, SM2 Threads, 0.5” Thick, 8–32 Tap (Two SM2RR Retaining Rings Included) Thorlabs CLS-SL 2 Scan Lens with Large Field of View, 400 to 750 nm, EFL = 70 mm Migliori et al. BMC Biology (2018) 16:57 Page 7 of 19 Table 1 Parts list of LSTM (Continued) Vendor Number Qty. Description Thorlabs ER18 6 Cage Assembly Rod, 18″ Long, Ø6 mm Thorlabs LCP50S 2 60-mm Cage System Iris, Ø50.0 mm Maximum Aperture Optotune EL-16-40-TC 2 Electrically Tunable Lens Thorlabs ER4 18 Cage Assembly Rod, 4″ Long, Ø6 mm Thorlabs VA100C 2 30-mm Cage System Adjustable Slit, 8–32 Tap, Imperial Micrometer Thorlabs GVS001 2 1D Galvo System, Silver-Coated Mirror, PSU Not Included Thorlabs GCM001 2 1D Galvo 30-mm Cage System Mount Omicron / 2 Collimator with ~ 10-mm Bead Diameter Output (Custom-made) Omicron SOLE-6 1 SOLE-6 Engine Containing Four Laser Lines: 405, 488, 561, 647 nm Thorlabs TTL200 2 f = 200 mm Tube Lenses for Wide Field Imaging Base support Thorlabs MB1236 1 Aluminum Breadboard 12″ ×36″ × 1/2″, 1/4″-20 Taps Thorlabs RS12 4 Ø1” Pillar Post, 1/4″-20 Taps, L = 12″,8–32 Adapter Included Thorlabs C1001 4 Post Mounting Clamp for Ø1” Post Stage and sample mounting Thorlabs LNR50S 3 50-mm (1.97″) TravelMax Translation Stage, 1/4″-20 Taps Thorlabs LNR50P3 1 XY Adapter Plate for LNR50 TravelMax Stages, Imperial Hole Spacings Thorlabs LNR50P2 2 Right-Angle Bracket for LNR50 TravelMax Stages, Imperial Threads / Theta chamber 1 Custom-made 3D Printed Sample Chamber Controls and electronics National Instruments CA1000 4 Configurable Connector Accessory Enclosure National Instruments NI PXIe-1082 1 Modular Electronic Instrumentation Platform Thorlabs GPS011 1 Galvo System Linear Power Supply Thorlabs LEDD1B 2 T-Cube LED Driver with Trigger Mode, 1200 mA Thorlabs BSC203 1 BSC203 - Three-Channel APT™ Benchtop Stepper Motor Controller Dual Xeon / 1 Custom Workstation with Supermicro X10DRHCT Motherboard Workstation illumination. Similarly, in LSM, the imaging of large time of the illumination line profile is the same for samples (i.e., larger than a single field of view of de- both. As summarized in Fig. 3e, the LSTM indeed tection) entails redundant illumination of the parts of imparts a much higher energy load for imaging of samples (along the illumination) not being imaged. As smaller samples, but the ratio approaches 1 with in- summarized in the Fig. 3d schematics, the total re- creasing sample size and higher detection magnifica- dundant energy load depends on the sample thickness tion. The LSTM energy load also depends on the in LSTM (Fig. 3d,top row),and on thesample width angular separation (lower for larger θ); θ =60° was in LSM (Fig. 3d, bottom row). Therefore, for a quantitative used for these calculations. To complement these calcula- comparison of LSTM vs. LSM energy loads, we calculated tions, we also performed empirical assessment of the signal the ratio of the total energy loads (the procedure is summa- photo-bleaching for the LSTM imaging datasets reported rized in Additional file 7:FigureS5) as afunctionofthe in this study. As shown in Fig. 3f, no significant trend is ob- sample width, thickness, angular separation (θ)ofthe served, suggesting minimal photo-bleaching consequences. LSTM, and the detection objective magnification. Note that In summary, the LSTM total energy load is much higher we are comparing the LSTM energy load with the scanned for smaller samples, but is comparable to LSM for high- light sheet microscopy, which is the commonly used imple- resolution imaging of larger samples, further support- mentation for imaging of large samples (e.g., COLM ) ing the overall complementary advantages of LSTM due to the reduced coherent illumination scattering and the and LSM for imaging of larger and smaller samples. synchronization possibilities with the rolling shutter Next, we characterized the effect of the width of the detection of sCMOS cameras. Therefore, the dwell rolling shutter (i.e., the “virtual” slit, controlled by the Migliori et al. BMC Biology (2018) 16:57 Page 8 of 19 a b LSTM geometric constraints Illumination depth comparision t/cos( ) w 43° Illu. path length LSTM vs LSM 60 10 t/cos( ) Illu. path length 62° LSTM < LSM 0 20 40 60 80 20 60 100 140 180 (degrees) w (mm) c Effective Light Sheet Thickness in LSTM d Illumination energy load in LSTM vs. LSM Single plane Single image stack 16 b/sin( ) b/sin( ) 10 30 50 70 90 (degrees) Dependence of LSTM illumination energy load on imaging parameters Sample width Theta Sample thickness Imaging mag. w = 5 mm w = 5 mm t = 5 mm obj = 10x obj = 10x 100 100 obj = 10x Mouse t = 3 mm = 60 d = 3 mm = 60 80 brain 100 t = 1 mm w = 10 mm 60 60 w = 10 mm 60 100 w = 20 mm Human brain coronal 40 40 40 w = 20 mm w = 30 mm w = 5 mm 5 mm thick slice 20 w = 30 mm 0 0 0 w = 200 mm w = 200 mm w = 10 mm 140 40 obj = 25x 100 obj = 25x 100 obj = 25x 120 w = 20 mm d = 3 mm 20 = 60 80 = 60 80 100 w = 200 mm 80 0 60 60 10x 25x Mouse 40 40 Detection objective Human brain coronal brain magnification 5 mm thick slice 20 20 0 0 0 50 100 150 200 20 30 40 50 60 70 80 1 3 5 Sample width, w (mm) (degrees) Sample thickness, t (mm) f Signal intensity across tiles of image volumes reported in this study 1.0 Fig. 5a Fig. 5b Fig. 5c Fig. 6a Fig. 6c 0.8 0.6 0.4 0.2 0 200 0 200 400 0 200 400 0 200 400 0 400 800 1200 Tiles in order of acquisition Fig. 3 LSTM characterization. a Geometric constraints in LSTM. Specific example of using Olympus 4×/0.28NA/29.5WD and 10×/0.6NA/8mmWD for illumination-detection. Note that the working distance of the air illumination objective is elongated in high refractive index immersion media (Additional file 3: Figure S3). Angular separation of ~ 60° was used for all experiments. b Comparison of maximum illumination path length (MIPL) required for full sample coverage in LSTM and LSM. The illumination light sheets need to penetrate the entire width (w) of the sample (or half width for two-sided illumination) in LSM, whereas MIPL depends on the angular arrangement and the tissue thickness (t) in LSTM. Bottom left: dependence of LSTM MIPL on θ and sample thickness (t, arrow indicates increasing t). Bottom right: MIPL dependence on the sample width and thickness: magenta and cyan highlight LSTM < LSM and LSTM > LSM respectively, assuming θ = 60° (see Additional file 6: Figure S4 for full θ range). c Effective planar illumination thickness can be approximated as b/sin(θ), where b is the actual light sheet thickness. The right graph plots the effective light sheet thickness as a function of θ and b (arrow points to increasing b). d Comparison of redundant illumination in LSTM and LSM for imaging of a single plane (top row) and an image stack (bottom row). e Ratios of total illumination energy loads (LSTM/LSM) as a function of sample width (w), angular configuration (θ), sample thickness (t), and objective magnification (10× and 25×). Illumination energy load is higher in LSTM for smaller samples and similar to LSM for larger samples. The energy load ratio also decreases with increased angular separation (60° is marked) and the magnification of detection objective. Additional file 7: Figure S5 provides details. f The average signal of tiles in the order of acquisitions. Note that no significant photo-bleaching trend is observed Total energy load ratio LSTM / LSM Normalized mean int. Eff. LS Thickness (µm) Illumination depth (mm) LSM LSTM Migliori et al. BMC Biology (2018) 16:57 Page 9 of 19 ab LSTM image x-z projection LSTM imaging of thick human brain section exp=0.1ms exp=0.2ms exp=0.3ms exp=0.4ms exp=0.5ms ~ 10.5mm x 14.1mm x 3mm DAPI exp=0.6ms exp=0.7ms exp=0.8ms exp=0.9ms exp=1.0ms 1 mm Fig. 4 LSTM optical sectioning. a x-z maximum intensity projections of an image stack acquired from human brain tissue, shown in (b), stained with DAPI. The camera rolling shutter exposure time determines the effective slit (rolling shutter) width (0.1–1 ms, i.e., 66–665 μm on the sCMOS sensor and 6–60 μm on the sample. The images were acquired using two different scanning modes: LSTM 1-axis scan (1-AS) and LSTM 2-axis scan (2-AS, default). Total frame exposure was 20 ms for all the images. As evident, the 2-AS mode allows for uniform planar illumination for achieving quantitative imaging, and the axial resolution decreases with increased rolling shutter exposure. All scale bars are 100 μm. b LSTM imaging of a large thick section of cleared human brain tissue (~ 10.5 mm × 14.1 mm × 3 mm) stained with DAPI. We used 0.5-ms rolling shutter exposure set- tings and 20 ms for entire frame exposure to acquire this dataset. Scale bar is 1 mm rolling shutter exposure parameter). As expected, the which are typically the rate-limiting step in the entire axial resolution is better for smaller rolling shutter imaging procedure. For example, as reported previously widths (Fig. 4a), analogous to the effect of the pinhole , a typical sample stage takes more than 50 ms for a diameter in confocal microscopy. 5-μm z-step sample motion, resulting in ~ 60–70 ms Finally, a major advantage of LSM is the high imaging acquisition time per z-plane, i.e., 10–15 Hz imaging speed. Similar to standard LSM implementations (e.g., speed for LSTM as well as state-of-the-art LSM COLM [5, 14]), the LSTM imaging procedure involves implementations for large sample imaging. Alterna- synchronization of the illumination line with the rolling tively, instead of using a step-wise motion, the sample shutter detection of the sCMOS camera. Therefore, a can also be continuously scanned at a small uniform single image acquisition takes 20 ms (which is a property speed to allow overall higher imaging speeds, al- of the camera imaging speed in rolling shutter mode), though at the cost of significant shearing artifacts. yielding 50 full frames (2048 × 2048) per second. The We would also like to highlight that, unlike LSM (e.g. LSTM also allows the synchronization of the two ,the COLM system ), LSTM imaging did not re- illumination line profiles independently with the bi- quire any pre-calibration/adaptive parameter correc- directional readout mode of the sCMOS sensors (10 ms tion steps because of the overall smaller illumination for full frame). However, the imaging of large volumes path lengths, resulting in a higher effective (about requires the use of long travel-range motorized stages, twofold) imaging speed. 1-Axis Scan Sim. 2-Axes Scan 1-Axis Scan Sim. 2-Axes Scan Migliori et al. BMC Biology (2018) 16:57 Page 10 of 19 LSTM enables rapid quantitative imaging of large suitability of LSTM for rapid high-resolution imaging of samples with uniform high resolution very large samples of different shapes. Unlike LSM, LSTM To assess the performance of LSTM in real-world experi- provides high uniform imaging quality even for the interior mental scenarios, we performed high-resolution imaging parts of the samples. In summary, these examples clearly of cleared samples of various sizes and shapes. First, we demonstrate the complementary advantages of LSTM over imaged a large thick section of cleared DAPI-stained hu- LSM for rapid high-resolution imaging of large samples. man brain (~ 10.5 mm × 14.1 mm × 3 mm) and a cleared intact central nervous system (11.8 mm × 27.6 mm × 5. LSTM enables rapid imaging of nervous system-wide 2mm) of a Thy1-eYFP transgenic mouse using 405 nm neuronal dynamics of freely motile animal and 488 nm illumination wavelengths respectively. As Finally, we demonstrate the compatibility of LSTM in shown in Figs. 4b and 5a and Additional file 8:Video 3 capturing the nervous system dynamics of a highly , LSTM enabled rapid high-resolution imaging of motile animal. Live samples often undergo substantial these large samples with uniform imaging quality through- rearrangements in their body shape and cellular density out, even for a sample with highly curved surfaces. We fur- which significantly alter their local optical properties. ther imaged a thick (~ 9.6 mm × 13.5 mm × 5.34 mm; the Although LSM-based imaging methods have been effect- sample expanded ~ 1.5- to 2-fold due to the immersion in ive in capturing the cellular dynamics of developing em- glycerol solution ) coronal slice of a CLARITY-cleared bryos and the neuronal activity of immobilized zebrafish Thy1-eYFP transgenic mouse brain, with 10×/0.6NA/ larvae, LSM remains susceptible to large changes in 8mmWD (Fig. 5b, Additional file 9:Video 4) as well as shape and density of motile samples, mainly because of 25×/1.0NA/8mmWD (Fig. 5c) objectives. As demonstrated the use of orthogonal illumination. This limitation has by zoom-in views at various locations in the samples, been partly addressed by utilizing a sophisticated array LSTM provides high uniform quality throughout the sam- of hardware and software components that facilitate ple. To directly compare the performance of LSTM with real-time adaptation of light sheet parameters . LSM, we also imaged a very wide (~ 1.5 cm) and thick (> LSTM, with its non-orthogonal illumination, provides a 5 mm, i.e., as thick as mouse brain) slice of highly cleared simpler and highly effective solution. We tested this hy- (see Fig. 6a) rat brain, which was stained for uniform dis- pothesis by performing rapid volumetric calcium im- tributed blood vessels (Fig. 6a; using tomato lectin, excita- aging of a highly motile Hydra, which has been recently tion wavelength 488 nm). Previous attempts of using LSM established as an effective model for exploring the role for the imaging of rat brain resulted in very poor image of neuronal circuit activity on behavior [31, 32]. We quality in most of the internal parts of the brain be- found that, indeed, LSTM enables aberration-free cal- cause of the heavy scattering of illumination light sheets. As cium imaging of freely behaving Hydra undergoing shown in Fig. 6 and Additional file 10:Video 5, LSTM drastic changes in body shape and cellular density in the allowed uniform high-resolution imaging of the entire tis- recordings (Fig. 7a and Additional file 12: Video 7, sue, whereas LSM resulted in progressively poor image Additional file 13: Video 8 ). In a way, the large and quality towards the center of the sample, similar to the pre- non-isomorphic body deformation of Hydra represents vious report . the worst-case scenario for tracking the activity of neu- To further demonstrate the suitability of LSTM for rons during behavior. We validated LSTM datasets by imaging of very large samples, we performed rapid extracting and comparing neuronal traces with previous high-resolution imaging of a very large (3.32 cm × 1. observations, finding excellent agreement . Note 93 cm × 1 mm) uniformly expanded (using the ExM that, for this demonstration, we used the relatively slow approach ) brain slice of a transgenic Thy1-eYFP process of step-wise motion of the sample stage to ac- mouse using 488 nm excitation wavelength (Fig. 6c, d, quire the image stacks. The LSTM mechanism can be Additional file 11:Video 6). The advent of tissue straightforwardly combined with piezo motor-based syn- expansion approaches (ExM) is enabling higher effective chronous rapid scanning of the detection objective and imaging resolution, however at the cost of hugely also with extended detection depth of field . increased imaging time and data sizes. For example, the im- aging of this sample with state-of-the-art confocal or two- Discussion photon microscopy will likely take several weeks of con- We reported the development of LSTM, which addresses tinuous imaging, whereas LSTM took only ~ 22 h (using the lateral size limitation of state-of-the-art LSM approaches. 10×/0.6NA), yielding 723,200 full frame (2048 × 2048 LSTM employs two symmetrically arranged oblique static pixels) images. The resulting dataset reveals the finest light sheets generated using independent illumination details of brain neuronal architecture (e.g., dendritic objectives, and their scanning using simultaneous two- spines, Fig. 6c, d and Additional file 11:Video 6). Taken dimensional (2D) translation along (using an electrically together, these imaging examples clearly demonstrate the tunable lens (ETL)) and perpendicular (using galvo scanners) Migliori et al. BMC Biology (2018) 16:57 Page 11 of 19 Z X 13.5 mm 5.34 mm 9.6 mm Fig. 5 (See legend on next page.) Migliori et al. BMC Biology (2018) 16:57 Page 12 of 19 (See figure on previous page.) Fig. 5 Rapid uniform high-resolution imaging of mouse central nervous system. a A CLARITY-cleared Thy1-eYFP transgenic mouse brain with attached spinal cord was imaged with LSTM microscopy using 10×/0.6NA/8mmWD detection objective (correction collar adjusted to 1.45 refractive index). A rolling shutter exposure of 0.5 ms and a full frame exposure of 20 ms were used. High-resolution 3D rendering was generated after 2 × 2-fold down-sampling. The bounding boxes are 11.8 mm × 27.6 mm × 5.2 mm for the whole sample and 5.1 mm × 3.1 mm × 3.5 mm for the subvolume (magenta). Images were acquired with 5-μm z-spacing using an effective light sheet thickness of ~ 5 μm. Lateral pixel sampling was 0.585 × 0.585 μm. A detailed volume rendering is shown in Additional file 8: Video 3 . b A large thick coronal slice of a Thy1-eYFP transgenic mouse brain was imaged with LSTM using 488 nm excitation wavelength. A rolling shutter exposure window of 0.5 ms and a full frame exposure of 20 ms were used. The volume rendering was performed using 4 × 4 fold down-sampled data. The bounding box is 9.6 mm × 13.5 mm × 5.34 mm. Images were acquired with 5-μm z-spacing using an effective light sheet thickness of ~ 5 μm. Lateral pixel sampling was 0.585 × 0.585 μm. Additional file 9: Video 4  shows volumetric rendering. c The same sample as shown in b was imaged with a high-NA 25×/1.0NA/8mmWD objective. A rolling shutter exposure window of 0.4 ms and a full frame exposure of 20 ms were used. The volume rendering was performed after 2 × 2-fold down-sampling. The bounding box is 6 mm × 9.6 mm × 0.5 mm. Images were acquired with 5-μm z-spacing using an effective light sheet thickness of ~ 3 μm. Lateral pixel sampling was 0.234 × 0.234 μm to their propagation directions, resulting in uniform illumin- shown to easily achieve  > 30 Hz for the full-range ation and detection (using synchronized rolling shutter de- coverage without any distortions, and they can achieve tection of sCMOS cameras) of thin optical sections. This even higher speeds if run continuously using sinusoidal optical configuration eliminates the fundamental restrictions waveforms (given that the light sheets in LSTM have a of LSM on the lateral dimensions of the imaging volumes significant confocal parameter, the synchronization of si- while ensuring high imaging speed and resolution and the nusoidal ETL waveforms with linear row-by-row detec- utilization of the entire available WD of high-NA detection tion is feasible). The ETLs can also be easily replaced objectives. The use of two light sheets (as opposed to one) with faster acoustic tunable lenses (e.g., TAG lenses from ensures better quality (e.g., if one of the sheets is obstructed TAG Optics) for applications requiring much higher by opaque objects) but is not necessary. To minimize optical volumetric imaging speed (e.g., for the calcium imaging aberrations, we used refractive index optimized detection ob- of functioning nervous systems). Hence, the introduction jectives and used 3D-printed caps (with quartz coverslips) of ETLs in the LSTM imaging procedure has no conse- for illumination objectives (air, low NA, and long WD) to en- quences for the overall imaging speed vis-à-vis LSM for sure a perpendicular incidence of light sheets to the mount- high-resolution imaging of large samples. We also found ing media. that, unlike LSM, LSTM does not require a pre-imaging The enhanced performance of LSTM entails an in- calibration step for the estimation of sample-position- creased overall energy load for imaging of smaller sam- dependent alignment parameters, resulting in overall ples, such as embryos, but remains comparably as low as faster imaging speeds. LSM for high-resolution imaging of large samples, as We demonstrated the LSTM performance by rapid high- supported by the simulations (Fig. 3d, e) and empirical resolution imaging of large samples of various sizes and calculations (Fig. 3f). Moreover, the use of only the shapes, including the entire intact mouse central nervous thinnest part of the light sheets enables quantitatively system, thick coronal sections of mouse and rat brains, a uniform illumination of the entire detection plane, which large chunk of human brain, uniformly expanded brain tis- is a foremost requirement for 3D quantitative imaging. sue, and a highly motile Hydra. Also, due to geometric ad- Therefore, LSTM is most suitable for rapid high-resolution vantage, LSTM is expected to enable volumetric calcium imaging of very large samples, whereas LSM provides better imaging in live rodent brains, similar to SCAPE , espe- performance for smaller samples. cially by combining with rapid de-focusing/focusing (e.g., The use of ETLs in the illumination arm for translat- using ETLs in the detection arm). We also performed dir- ing light sheets along their propagation direction intro- ect comparative imaging of the same thick rat brain cor- duces an additional component which needs to match onal section using LSTM and LSM (Fig. 6)to demonstrate the camera acquisition speed. The high-resolution im- that LSTM indeed eliminates the limit on lateral dimen- aging of large samples essentially requires full camera sions of imaging volumes while providing high imaging frame acquisition (i.e., 20 ms or 50 Hz for current state- speed and uniform imaging resolution. Therefore, LSTM of-the-art sCMOS cameras in one-direction rolling shut- provides complementary advantages over LSM for rapid ter mode) and the use of long travel-range motorized high-resolution imaging of very large samples. These cap- sample stages for step-wise z-plane acquisition (typically abilities of LSTM are expected to significantly accelerate > 50 ms motion and settling time, in addition to the our understanding of healthy and diseased tissue architec- camera exposure) which is generally the rate-limiting tures. Future work will include integration of super- step. This results in a typical acquisition rate of 10–15 z- resolution approaches (such as structured illumination) planes per second as reported previously . ETLs are and simultaneous multi-view imaging. Migliori et al. BMC Biology (2018) 16:57 Page 13 of 19 a b LSTM Z = 1.0 mm Z = 2.5 mm 2 cm 1 mm LSM Z = 4.0 mm 1 mm LSTM LSM 0 5 10 15 Sampling coordinate (mm) cd 3.32 centimeters Fig. 6 LSTM enables rapid uniform high-resolution imaging of very large samples. a For an unbiased comparison of the imaging performance of LSTM and LSM a highly cleared large rat brain tissue (~ 2 cm wide and ~ 5 mm deep; vasculature stained with tomato lectin) was imaged using the exact same detection (10×/0.6NA/8mmWD, correction collar adjusted to 1.45 refractive index) and illumination objectives (4×/0.28NA/ 28.5WD). Maximum intensity projections are shown. The bottom graph profiles the mean intensity across the length of the specified (dashed rectangles) regions of interest. In LSM (cyan), the intensity signal is progressively degraded towards the interior of the sample, whereas LSTM (magenta) allows uniform quality across the entire sample. The scale bars are 1 mm. b An image stack from the sample shown in a. Maximum intensity projections (50 μm) are shown at three different depths (orange). The bounding box is 1 mm × 1 mm × 5 mm. The scale bars are 100 μm. A detailed volume rendering is shown in Additional file 10: Video 5 . c Uniformly expanded (~ 4-fold in all three dimensions) slice of Thy1-eYFP transgenic mouse was imaged using LSTM with 10×/0.6NA/8mmWD detection objective. A rolling shutter exposure window of 0.2 ms and a full frame exposure of 20 ms were used. The resulting dataset consists of 723,200 images (2048 × 2048 pixels) and required ~ 22 h of acquisition time. The volume rendering was performed with 8 × 8 fold down-sampled dataset. Zoomed-in images are marked. d An image stack from the dataset shown in c. The bounding box size is 1.2 mm × 1.2 mm × 1 mm. Note that the dendritic spines can be unambiguously identified. Detailed volume rendering in Additional file 11: Video 6  Normalized mean Int. 1.93 centimeters Migliori et al. BMC Biology (2018) 16:57 Page 14 of 19 a Non-isomorphic body deformation in living samples (Hydra) b Rapid volumetric calcium imaging of highly motile Hydra 100 sec Fig. 7 LSTM enables rapid volumetric imaging of highly motile animals. Live samples can undergo substantial non-isomorphic rearrangements in their body shape and cellular density, resulting in continuously changing local optical properties. LSM is particularly susceptible to misalignments and other aberrations because of the use of orthogonal light sheet illumination. LSTM is uniquely suitable for rapid volumetric live imaging of such difficult samples, as demonstrated by imaging of highly motile Hydra. a Hydra image is shown at different time points to highlight the non- isomorphic changes in freely moving animal. b LSTM was used to perform long-term (> 1 h demonstrated, Additional file 12: Video 7 ) high- resolution live imaging of an adult Hydra expressing GCaMP6s . Each volume consists of 17 z-planes. Manual tracking and analyses of calcium signaling were performed for the first ~ 500 s of recording. Maximum intensity projections covering the two halves are shown. Representative neuronal traces are shown for cells marked in corresponding colors. As shown in Additional file 13: Video 8 , the neuronal traces correlate with the rapid longitudinal contraction behavior of Hydra, and the other two traces are part of rhythmic potential circuits, in excellent agreement with the observations reported recently . Scale bars are 100 μm Conclusions in a practical time-frame and direct in situ transcriptomics We report the development of a distinct light sheet mi- of whole rodent brains. croscopy (LSM) approach, termed light sheet theta mi- croscopy (LSTM), which addresses the fundamental Methods limitation of LSM on the lateral dimensions of imaging LSTM implementation volumes due to the orthogonal illumination-detection The optical layout and physical implementation details geometry. We have presented extensive characterization are presented in Fig. 2, Additional file 1: Figure S1, and of the LSTM system properties and performed several Additional file 2: Figure S2. A complete parts list is pro- real-world high-resolution imaging experiments of very vided in Table 1. The illumination arms of the LSTM in- large samples, including mouse and rat brains, a large sec- strument consist of a laser source (Omicron-Laserage, tion of human brain, a highly expanded expansion micros- Rodgau, Germany; SOLE-6 engine containing four wave- copy sample, and rapid volumetric calcium imaging of a lengths: 405, 488, 561, and 647 nm), a cylindrical lens highly motile Hydra. These new imaging capabilities will (Thorlabs, Newton, NJ, USA; LJ1695RM-A, f = 50 mm), enable numerous novel applications, including imaging of a vertical slit (Thorlabs, VA100C), an iris (Thorlabs, an entire post-mortem human brain (thick slab-by-slab) CP20S), an electrically tunable lens (Optotune, Dietikon, Bottom half Top half Migliori et al. BMC Biology (2018) 16:57 Page 15 of 19 Switzerland; EL-16-40-TC), a galvo scanner (Thorlabs, The range of allowable angular positions was calculated GVS001), a scan lens (Thorlabs, CLS-SL), a tube lens by taking the effective working distances (WDs) and the (Thorlabs, ITL200), and the illumination objective objective diameters into account as shown in the sche- (Olympus Macro 4×/0.28NA/28.5 mmWD air). Since matics of Additional file 3: Figure S3, resulting in the fol- the illumination objectives were air objectives, we used a lowing relationships: 3D-printed cap (using Ultimaker 2 + extended, in D1 cosðÞ θi polylactic acid (PLA)) containing a 1-in. diameter quartz W2 sinðÞ θi ¼ þ D2 2 2 coverslip, to allow dipping in immersion oil, ensuring that the low-numerical aperture (NA) light rays enter the media sinðÞ θf W2 cosðÞ θf ¼ W1 þ D2 near perpendicularly. The detection arm is composed of a detection objective (Olympus, XLPLN10XSVMP/10× or XLSLPLN25XGMP/25×), a tube lens (Thorlabs, TTL200), a where W and W are the WDs of the detection and illumin- 1 2 multi-band emission filter (Semrock, FF01-432/515/595/ ation objectives respectively, D and D are the diameters of 1 2 730-50-D), and an sCMOS camera (Hamamatsu Orca Flash the detection and illumination objectives respectively, and θ 4.0 V3). The illumination arms were vertically mounted at and θ are the boundary angular positions. Since we used a an approximately 60° angle relative to the detection axis. To Macro 4×/0.28NA/29.5WD (Olympus) air objective for illu- facilitate the optical alignment adjustments, all three optical mination, the approximate effective WDs were calculated as arms were mounted on manual translation stages (Thorlabs, shown in Additional file 3: Figure S3. For most of the experi- TBB0606) attached to the breadboard. We used a 3D- ments we used a 10×/0.6NA/8mmWD objective (Olympus) printed (using Ultimaker 2 + extended in PLA) open-top in the detection arm with values of W =8, and D = 40. For 1 1 sample chamber that was filled with an immersion oil of re- this combination, we found the allowable angular range to ° ° fractive index 1.454 (Cargille Laboratories, Cedar Grove, NJ, be 43.3 to 62.3 , which served as our initial guide for USA). The CLARITY-cleared (refractive index = ~ 1.454) identifying the maximum possible angular positioning. We sample was mounted in a quartz cuvette (refractive index ~ used ~ 60 as the final angle separation. All calculations were 1.45, FireflySci, Staten Island, NY, USA) which was then performed in MATLAB. affixed to appropriate grooves at the base of the sample chamber (Additional file 1: Figure S1). The 3D model of the Illumination depth calculations LSTM microscope was made with Autodesk Inventor 2017. We used geometric calculations (Fig. 3b)toestimate the maximum illumination path lengths (MIPLs) of LSM imaging LSTM as t/cos(θ), where t is the sample thickness to All LSM imaging experiments were performed using the be imaged and θ is the angle between the illumin- COLM implementation as described previously . Briefly, ation propagation direction and the detection axis. the optical components were the same as described for TheMIPLinLSM wouldbethe same as the sample LSTM, i.e., detection objectives: Olympus 10×/0.6NA/ width (w). We calculated the ratio of these illumin- 8mmWD or Olympus 25/1.0NA/8mmWD with a correction ation path lengths, which was then converted into a collar for the refractive indices of water to oil; sCMOS cam- binary representation by thresholding at 1 and plotted era: Hamamatsu Orca Flash 4.0 V3; low-NA illumination ob- as a heat map, shown in Fig. 3b. The edge effects jective: Olympus Macro 4×/0.28 NA; tube lens: Thorlabs were approximated. TTL200; and scan lens: Thorlabs, CLS-SL. The dynamic light sheets are generated by rapid scanning of Gaussian LSTM effective light sheet thickness calculations beams (Thorlabs, GVS102). Similar to LSTM, COLM uses Due to the non-orthogonal incidence of the light sheets synchronized illumination-detection to improve the imaging on the detection plane, the effective light sheet thickness quality. Additionally, an automated-alignment parameter can be approximated as the projection of the original calibration (using linear adaptation) corrects for misalign- thickness onto the detection direction, resulting in b/ ment artifacts across the whole sample space. sin(θ), where b is the actual light sheet thickness at the most focused position, and θ is the angle of incidence LSTM geometric constraint calculations relative to the detection axis. The relationship is plotted The physical geometric constraints of the oblique arrange- in Fig. 3c. ment of illumination and detection objectives were ana- lyzed by calculating the upper and lower bounds (Fig. 3a, Imaging experiments Additional file 3: Figure S3) with the following criteria: the All the imaging experiments are summarized in Table 2.We illumination objective should not touch the detection ob- used the passive CLARITY method (as described previously jective (Fig. 3a top) and it should not extend below the ) for all the tissue clarification experiments. The hydrogel physical extent of the detection objective (Fig. 3a middle). monomer (HM) solution recipe consisted of 1–4% (wt/vol) Migliori et al. BMC Biology (2018) 16:57 Page 16 of 19 Table 2 Summary of imaging experiments reported in this study Samples Fig. No. Label Det. objective Illum. objective Imaging volume dimensions No. of images/raw data Imaging time Thick human brain tissue 4 DAPI 10×/0.6NA/8mmWD 4×/0.28NA/28.5WD ~ 10.5 mm × 14.1 mm × 3 mm 116,736/~ 0.97 TB ~ 2.7 h Mouse brain with attached 5a Thy1-eYFP 10×/0.6NA/8mmWD 4×/0.28NA/28.5WD 11.8 mm × 27.6 mm × 5.2 mm 388,687/~ 3.3 TB ~ 9 h spinal cord Thick mouse brain slice 5b Thy1-eYFP 10×/0.6NA/8mmWD 4×/0.28NA/28.5WD ~ 9.6 mm × 13.5 mm × 5.34 mm 256,560/ ~ 5.9 h ~ 2.1 TB Thick mouse brain slice 5c Thy1-eYFP 25×/1.0NA/8 mm/WD 4×/0.28NA/28.5WD 6 mm × 9.6 mm × 1.9 mm 211,616/ ~ 4.9 h ~ 1.8 TB Thick rat brain slice 6a, b Tomato lectin 10×/0.6NA/8mmWD 4×/0.28NA/28.5WD ~ 20 mm × 16.5 mm × 3.6 mm deep 285,821 slices/2.4 TB ~ 6.6 h expanded (~ 4×) mouse brain 6c Thy1-eYFP 10×/0.6NA/8mmWD 4×/0.28NA/28.5WD 33.2 mm × 19.3 mm × 2 mm 723,200/ ~22 h slice ~6 TB Hydra live imaging 7 GCaMP6s 10×/0.6NA/8mmWD 4×/0.28NA/28.5WD 1.2 mm × 1.2 mm × 0.136 mm 23,001/~ 193 GB ~ 1 h live imaging Summary of the datasets reported in this report. The image volume acquired was ~ 6 mm × 9.6 mm × 1.9 mm; however, due to constraints of high-quality volume rendering, a smaller (0.5-mm-thick) subset was used for the rendering shown in Fig. 5c The approximate imaging volume was ~ 20 mm × 16.5 mm × 3.6 mm, and a few ~ 5-mm-deep image stacks were acquired to demonstrate the imaging depth in Fig. 6b The imaging volume acquired was ~ 33.2 mm × 19.3 mm × 2 mm to ensure complete coverage of ~ 1-mm-thick expanded non-rigid tissue TB terabytes, GB gigabytes, h hours Migliori et al. BMC Biology (2018) 16:57 Page 17 of 19 acrylamide, 0.05% (wt/vol) bisacrylamide, 4% paraformalde- Illumination energy load calculations hyde (PFA), 1× phosphate-buffered saline (PBS), deionized The procedure is summarized in Additional file 7: water, and 0.25% thermal initiation VA-044 (Wako Figure S5. To calculate the total illumination energy Chemicals, Boston, MA, USA; NC0632395). All ani- load in LSTM, we performed a simulation of the step-wise mal procedures were followed according to Institu- scanning of the sample through the illuminating light sheet. tional Animal Care and Use Committee (IACUC) A horizontal plane across an entire sample can be imaged guidelines. For whole brain clearing, transcardiac per- with approximately non-overlapping thin sheets of light; fusion was performed with 20 mL HM solution, therefore, the total energy is calculated by step-wise scan- followed by overnight incubation at 4 °C. The rat ning of the sample through the illumination volume. All brain was perfused with 4% PFA, post-fixed for 16 h, voxels receiving illumination are incremented by 1. The and then frozen in isopentane for storage. The frozen final energy load is calculated as the total sum of accumu- brain was thawed at room temperature in PBS buffer, lated illuminations in all voxels. The procedure was imple- then sliced and incubated in HM solution overnight mented for a range of parameters and two detection at 4 °C. The human brain tissue was incubated in 4% objectives (10×/0.6NA/8mmWD and 25/1.0NA/8mmWD). PFA for ~ 2 days, followed by incubation in HM solu- For LSM calculations, each voxel is illuminated (w/f) tion overnight at 4 °C. All the perfused tissues were times, where w is the width of the sample, and f is de-gassed and then stored at 37 °C for 3–4hfor the field-of-view size of the detection arm. Therefore, hydrogel polymerization. The tissues were cleared by the total energy load is approximated as (w/f)*number incubating (with shaking) in clearing buffer (4% (wt/ of voxels. Note that the energy load in LSM as well vol) sodium dodecyl sulfate (SDS), 0.2 M boric acid, as LSTM scales up by the same constant factor, pH 8.5) at 37 °C until clear (2–3 weeks). Afterwards, which cancels out in the ratio. the tissues were washed with 0.2 M boric acid buffer (pH 8.5) with 0.1% Triton X-100 for up to 24 h. The Additional files cleared tissue was labeled with DAPI (1 μg/mL final concentration) and/or the blood vessel marker tomato Additional file 1: Figure S1. LSTM microscopy implementation. (a) Image of the physical LSTM setup. (b) 3D model of LSTM and the sample lectin (Vector Labs, Burlingame, CA, USA; FL-1171) mounting system. The 3D-printed sample chamber is designed to by incubating in the labeling solution for 3–4days. accommodate large biological samples of virtually any dimensions, while After washing with the buffered solution (0.2 M boric still allowing the objectives to be immersed in the immersion oil. Two transparent glass windows, located on the lateral sides, provide visual acid buffer, pH 7.5, 0.1% Triton X-100), the tissue was view of the sample for ease of positioning. An additional window is transferred into 85–87% glycerol solution in graded fashion realized at the bottom part of the chamber to allow the illumination light (i.e. 25%, 50%, 65%, and finally 87%) for final clearing and to pass through. An additional adapter was designed to allow mounting a prism mirror at about approximately 10° from the normal surface to imaging. For uniform tissue expansion (4–4.5× uniformly), facilitate the optical alignment of the system. (PDF 1623 kb) a Thy1-eYFP mouse brain slice (250 μm, perfused and fixed Additional file 2: Figure S2. Detailed annotation of LSTM optical path with 4% PFA and sliced with vibratome) was gelled and shown in Fig. 2a. (PDF 571 kb) digested following the protein retention expansion Additional file 3: Figure S3. Physical constraints of positioning microscopy (proExM) protocol . Thesamplewas stored illumination and detection objectives. (a) Schematics showing calculations of the elongated working distance (EWD) of the air in 1× PBS before changing the buffer to 65% glycerol (with illumination objective (Olympus Macro 4×/0.28NA/29.5WD Air) when 2.5 mg/mL 1,4-diazabicyclo[2.2.2]octane (DABCO)) for the used in immersion liquid (refractive index 1.454). Original working LSTM imaging. All imaging experiments were performed distance (OWD) is the working distance in air according to the objective specifications. A thin quartz coverslip and a 3D-printed cap were used to with an effective light sheet thickness of 2–5 μm. seal the illumination objectives. EWD was estimated to be 43.84 mm. (b) Geometric constraints calculation for the co-arrangement of the illumination Image analyses and detection objectives. The two boundary conditions are shown in blue and green shading of the illumination objective. For the upper bound limit A TeraStitcher -based pipeline  was used for the (blue), the relationship among different parameters is defined by the stitching of acquired image stack tiles of all the datasets. cosðθiÞ D1 equation W2 sinðθiÞ¼ þ D2 . For the lower bound limit 2 2 sinðθ fÞ Maximum intensity projections and other linear image (green), it is defined by W2 cosðθfÞ¼ W1 þ D2 . W and W are 1 2 the effective working distances of detection and illumination objectives contrast adjustments were performed using Fiji [36, 37] respectively. D and D are the diameters of the detection and illumination 1 2 and MATLAB. All volume renderings were performed objectives respectively. θ and θ are the angular positions of upper and i f using Amira (FEI, Lausanne, Switzerland). All the fluor- lower bounds respectively. For the 4×/0.28NA/29.5 mmWD (as illumination objective) and 10×/0.6NA/8mmWD (as detection objective), the calculated escent bead image analyses were performed using Fiji. θ and θ are 43.32° and 63.37° respectively. This range served as a starting i f To calculate the axial full width at half maximum point during the optical alignment of the system. (PDF 596 kb) (FWHM), x-z projections of the bead image stacks were Additional file 4: Video 1. 3D model of LSTM implementation. The 3D used. For individual beads a line intensity profile was modeling and rendering was performed using Autodesk Inventor 2017, and the animation was performed using Autodesk Fusion 360 2017 and calculated along the central axis, followed by manual MATLAB. The components labeled are LS (laser source), collimator, ND calculations of FWHM intensity values. Migliori et al. BMC Biology (2018) 16:57 Page 18 of 19 (neural density) filter mount, iris, ETL (electrically tunable lens), slit, CL and radial contractions are annotated. The scale bar is 100 μm. The (cylindrical lens), galvo scanner, SL (scan lens), iris and TL (tube lens). high-resolution video is available in the figshare repository at The high-resolution video is available in the figshare repository at https://doi.org/10.6084/m9.figshare.c.4072160. (MOV 66662 kb) https://doi.org/10.6084/m9.figshare.c.4072160. (MP4 82944 kb) Additional file 13: Video 8. Neuronal activity traces of representative Additional file 5: Video 2. Comparison of image volumes acquired neurons. A visualization of the neuronal traces shown in Fig. 7b.The with LSTM in 1-axis scan (1-AS) and 2-axis scan (2-AS) modes. The 3D high-resolution video is available in the figshare repository at rendering visualizes the image stacks acquired from the same sample https://doi.org/10.6084/m9.figshare.c.4072160. (MOV 11161 kb) (human brain section stained with DAPI) with LSTM in 1-AS and simultaneous 2-AS modes. The high-resolution video is available in the figshare repository at https://doi.org/10.6084/m9.figshare.c.4072160.(MOV Acknowledgements 104448 kb) We thank all the members of the Tomer Laboratory for helpful discussions and Additional file 6: Figure S4. Comparison of maximum illumination inputs. We are grateful to Darcy Kelley and Oliver Hobert for general advice and path lengths in LSTM and LSM. The graphs plot the binarized ratios reading of the manuscript draft. We would like to thank Serge Przedborski for (w=ð Þ) of maximum illumination path lengths required for complete providing rat brain tissue and Peter Canoll for the human brain tissue. cosðθÞ coverage of samples of various widths (w) and thicknesses (t) for different angular arrangements. Magenta and cyan regions mark the combinations Funding of w and t for which the illumination path lengths were smaller in LSTM This work is mainly supported by a Columbia University Arts and Sciences and LSM, respectively. (PDF 262 kb) startup grant to RT. BM and OH were supported by the Swedish Research Additional file 7: Figure S5. Total illumination energy load in LSTM vs. Council (VR-2015-02675) and the Swedish Childhood Cancer Foundation LSM. The schematic summarizes the calculations of total energy loads (BCF-PR2016-0129). RY and CD were supported by the National Eye Institute imparted in LSTM and LSM for imaging of a sample of specific (NEI, grant DP1EY024503), and the reported Hydra material is based upon dimensions, imaged with a specific detection objective. (a) In LSTM, a the work supported by the Defense Advanced Research Projects Agency horizontal plane across the entire sample is imaged with approximately (DARPA) under Contract No. HR0011-17-C-0026. In addition, the Hydra re- non-overlapping thin sheets of light. Therefore, total energy load can be search was supported in part by competitive fellowship funds from the H. calculated by step-wise scanning of the sample (for each plane) through Keffer Hartline, Edward F. MacNichol, Jr. Fellowship Fund, the E. E. Just the illuminating light. For each of the steps, all voxels that receive light Endowed Research Fellowship Fund, the Lucy B. Lemann Fellowship Fund, are incremented by 1. The procedure was implemented for a range of and the Frank R. Lillie Fellowship Fund Fellowship Fund of the Marine Bio- parameters and two detection objectives (10×/0.6NA/8mmWD and logical Laboratory in Woods Hole, MA, USA. 25/1.0NA/8mmWD). (b) In LSM a stack (or tile) is acquired by approximately non-overlapping thin sheets of light. The total energy load is calculated by summing up the illumination for all tiles in a row along Availability of data and materials the width. Note that the dwell time of illumination line profile is same for A complete computer-aided design (CAD) model of LSTM and other related both LSTM and LSM (scanned light sheet implementation, e.g., COLM). resources are detailed in the additional files. A complete parts list is included The energy load for tiles along the sample length scales up by the same as Table 1. High-resolution movies and figures will also be shared at the re- constant factor in LSTM and LSM; therefore, we only simulated one row source webpage: http://tomerlab.org/lstm/. The datasets generated and/or of tiles along the sample width. (PDF 556 kb) analyzed during the current study are available in the figshare repository Additional file 8: Video 3. High-resolution LSTM imaging of intact (https://doi.org/10.6084/m9.figshare.c.4072160). Thy1-eYFP mouse central nervous system. The bounding box for the entire sample is 11.8 mm × 27.6 mm × 5.2 mm, and for the subvolume Authors’ contributions shown is 5.1 mm × 3.1 mm × 3.5 mm. The raw data was down-sampled RT conceived the project and designed the microscopes. BM, MSD, and RT 2 × 2 fold (to make the volume rendering feasible) for the subvolume built the microscopes. MSD, BM, and RT performed the imaging experiments. rendering. The high-resolution video is available in the figshare repository MCA assisted with the experiments. CD prepared the Hydra samples for the at https://doi.org/10.6084/m9.figshare.c.4072160. (MOV 143360 kb) live imaging experiments. RY conceived of the Hydra project in general, and Additional file 9: Video 4. High-resolution LSTM imaging of a large supervised the Hydra sample preparations used in this work. SA, RG, and ESB tissue of Thy1-eYFP mouse brain. The bounding box is 9.6 mm × generated the expanded tissue. OH supported BM. RT, BM, and MSD analyzed 13.5 mm × 5.34 mm. The raw data was down-sampled 4 × 4 fold to make the data and wrote the paper with inputs from all authors. RT supervised all the volume rendering feasible. The high-resolution video is available in aspects of the work. All authors read and approved the final manuscript. the figshare repository at https://doi.org/10.6084/m9.figshare.c.4072160. (MOV 167936 kb) Additional file 10: Video 5. Visualization of an image stack of Ethics approval and consent to participate Not applicable. vasculature-stained rat brain tissue. This video visualizes an image stack acquired from a large rat brain slice (stained for vasculature with tomato lectin) using LSTM in 2-AS mode. The bounding box is 1 mm × 1 mm × Competing interests 5 mm. The high-resolution video is available in the figshare repository at Columbia University has filed a patent application for LSTM. https://doi.org/10.6084/m9.figshare.c.4072160. (MOV 143360 kb) Additional file 11: Video 6. High-resolution LSTM imaging of a large expanded section of Thy1-eYFP mouse brain. A thin (250 μm) coronal section was expanded ~ 4-fold using proExM procedure and imaged Publisher’sNote Springer Nature remains neutral with regard to jurisdictional claims in using LSTM with 10×/0.6NA detection objective. The resulting dataset published maps and institutional affiliations. (~ 6 TB) consisted of 723,300 full frame images (2048 × 2048). The data was down-sampled 8 × 8 fold to allow high-quality volumetric rendering. The high-resolution video is available in the figshare Author details Department of Biological Sciences, Columbia University, New York, NY, USA. repository at https://doi.org/10.6084/m9.figshare.c.4072160.(MOV 2 3 NeuroTechnology Center, Columbia University, New York, NY, USA. Data 439296 kb) Science Institute, Columbia University, New York, NY, USA. Department of Additional file 12: Video 7. Rapid volumetric calcium imaging of Neuroscience, Karolinska Institutet, Stockholm,, Sweden. MIT Media Lab and highly motile Hydra. GCaMP6s-expressing Hydra was imaged using McGovern Institute, Departments of Biological Engineering and Brain and LSTM with 10×/0.6NA objective. Maximum intensity projections are Cognitive Sciences, MIT, Cambridge, MA, USA. Pfizer Internal Medicine shown for the two halves of the volume. First occurrences of longitudinal Research Unit, Cambridge, MA 02139, USA. Migliori et al. BMC Biology (2018) 16:57 Page 19 of 19 Received: 17 April 2018 Accepted: 23 April 2018 22. Bouchard MB, Voleti V, Mendes CS, Lacefield C, Grueber WB, Mann RS, Bruno RM, Hillman EM. Swept confocally-aligned planar excitation (SCAPE) microscopy for high speed volumetric imaging of behaving organisms. Nat Photonics. 2015;9(2):113–9. 23. Dunsby C. Optically sectioned imaging by oblique plane microscopy. Opt References Express. 2008;16(25):20306–16. 1. Richardson DS, Lichtman JW. Clarifying tissue clearing. Cell. 2015; 24. Gao L. Extend the field of view of selective plan illumination microscopy by 162(2):246–57. tiling the excitation light sheet. Opt Express. 2015;23(5):6102–11. 2. Romanov RA, Zeisel A, Bakker J, Girach F, Hellysaz A, Tomer R, Alpar A, 25. Planchon TA, Gao L, Milkie DE, Davidson MW, Galbraith JA, Galbraith CG, Mulder J, Clotman F, Keimpema E, et al. Molecular interrogation of Betzig E. Rapid three-dimensional isotropic imaging of living cells using hypothalamic organization reveals distinct dopamine neuronal subtypes. Bessel beam plane illumination. Nat Methods. 2011;8(5):417–23. Nat Neurosci. 2017;20(2):176–88. 26. Dean KM, Roudot P, Welf ES, Danuser G, Fiolka R. Deconvolution-free 3. Lerner TN, Shilyansky C, Davidson TJ, Evans KE, Beier KT, Zalocusky KA, Crow subcellular imaging with axially swept light sheet microscopy. Biophys J. AK, Malenka RC, Luo L, Tomer R, et al. Intact-brain analyses reveal distinct 2015;108(12):2807–15. information carried by SNc dopamine subcircuits. Cell. 2015;162(3):635–47. 27. Fu Q, Martin BL, Matus DQ, Gao L. Imaging multicellular specimens with 4. Chung K, Wallace J, Kim SY, Kalyanasundaram S, Andalman AS, real-time optimized tiling light-sheet selective plane illumination Davidson TJ, Mirzabekov JJ, Zalocusky KA, Mattis J, Denisin AK, et al. microscopy. Nat Commun. 2016;7:11088. Structural and molecular interrogation of intact biological systems. 28. Vettenburg T, Dalgarno HI, Nylk J, Coll-Llado C, Ferrier DE, Cizmar T, Gunn- Nature. 2013;497(7449):332–7. Moore FJ, Dholakia K. Light-sheet microscopy using an Airy beam. Nat 5. Tomer R, Ye L, Hsueh B, Deisseroth K. Advanced CLARITY for rapid and Methods. 2014;11(5):541–4. high-resolution imaging of intact tissues. Nat Protoc. 2014;9(7):1682–97. 29. Migliori B; Datta MS; Dupre C; Apak MC; Asano S; Gao R; Boyden ES; 6. Tanaka N, Kanatani S, Tomer R, Sahlgren C, Kronqvist P, Kaczynska D, Hermanson O; Yuste R; Tomer R (2018): Video files relating to Louhivuori L, Kis L, Lindh C, Mitura P, et al. Whole-tissue biopsy technique, imaging and modelling of LSTM: light sheet theta phenotyping of three-dimensional tumours reveals patterns of cancer microscopy, for rapid high-resolution imaging of large biological heterogeneity. Nat Biomed Eng. 2017;1(10):796–806. samples. figshare. https://doi.org/10.6084/m9.figshare.c.4072160. 7. Chen F, Tillberg PW, Boyden ES. Optical imaging. Expansion microscopy. 30. Royer LA, Lemon WC, Chhetri RK, Wan Y, Coleman M, Myers EW, Keller PJ. Science. 2015;347(6221):543–8. Adaptive light-sheet microscopy for long-term, high-resolution imaging in 8. Chang JB, Chen F, Yoon YG, Jung EE, Babcock H, Kang JS, Asano S, living organisms. Nat Biotechnol. 2016;34(12):1267–78. Suk HJ, Pak N, Tillberg PW, et al. Iterative expansion microscopy. Nat 31. Dupre C, Yuste R. Non-overlapping neural networks in Hydra vulgaris. Curr Methods. 2017;14(6):593–9. Biol. 2017;27:1085–97. 9. Mei E, Fomitchov PA, Graves R, Campion M. A line scanning confocal 32. Han S,TaralovaE,Dupre C, YusteR.Comprehensive machinelearning fluorescent microscope using a CMOS rolling shutter as an adjustable analysis of Hydra behavior reveals a stable behavioral repertoire. elife. aperture. J Microsc. 2012;247(3):269–76. 2018;7:e32605. 10. Wolleschensky R, Zimmermann B, Kempe M. High-speed confocal 33. Fahrbach FO, Voigt FF, Schmid B, Helmchen F, Huisken J. Rapid 3D light- fluorescence imaging with a novel line scanning microscope. J Biomed Opt. sheet microscopy with a tunable lens. Opt Express. 2013;21(18):21010–26. 2006;11(6):064011. 34. Tillberg PW, Chen F, Piatkevich KD, Zhao Y, Yu CC, English BP, Gao L, 11. Weber M, Mickoleit M, Huisken J. Light sheet microscopy. Methods Cell Biol. Martorell A, Suk HJ, Yoshida F, et al. Protein-retention expansion microscopy 2014;123:193–215. of cells and tissues labeled using standard fluorescent proteins and 12. Stelzer EH. Light-sheet fluorescence microscopy for quantitative biology. antibodies. Nat Biotechnol. 2016;34(9):987–92. Nat Methods. 2015;12(1):23–6. 35. Bria A, Iannello G. TeraStitcher - a tool for fast automatic 3D-stitching of 13. Migliori B, Datta MS, Tomer R. Advanced light microscopy enables rapid teravoxel-sized microscopy images. BMC Bioinformatics. 2012;13:316. mapping of brain structure and function in high resolution. Laser Focus 36. Schindelin J, Arganda-Carreras I, Frise E, Kaynig V, Longair M, Pietzsch T, World. 2016;52(12):47–52. Preibisch S, Rueden C, Saalfeld S, Schmid B, et al. Fiji: an open-source 14. Tomer R, Lovett-Barron M, Kauvar I, Andalman A, Burns VM, Sankaran S, platform for biological-image analysis. Nat Methods. 2012;9(7):676–82. Grosenick L, Broxton M, Yang S, Deisseroth K. SPED light sheet 37. Schneider CA, Rasband WS, Eliceiri KW. NIH Image to ImageJ: 25 years of microscopy: fast mapping of biological system structure and function. image analysis. Nat Methods. 2012;9(7):671–5. Cell. 2015;163(7):1796–806. 15. Stefaniuk M, Gualda EJ, Pawlowska M, Legutko D, Matryba P, Koza P, Konopka W, Owczarek D, Wawrzyniak M, Loza-Alvarez P, et al. Light-sheet microscopy imaging of a whole cleared rat brain with Thy1-GFP transgene. Sci Rep. 2016;6:28209. 16. Holekamp TF, Turaga D, Holy TE. Fast three-dimensional fluorescence imaging of activity in neural populations by objective-coupled planar illumination microscopy. Neuron. 2008;57(5):661–72. 17. Wu Y, Ghitani A, Christensen R, Santella A, Du Z, Rondeau G, Bao Z, Colon-Ramos D, Shroff H. Inverted selective plane illumination microscopy (iSPIM) enables coupled cell identity lineaging and neurodevelopmental imaging in Caenorhabditis elegans. Proc Natl Acad Sci U S A. 2011;108(43):17708–13. 18. Wu Y, Wawrzusin P, Senseney J, Fischer RS, Christensen R, Santella A, York AG, Winter PW, Waterman CM, Bao Z, et al. Spatially isotropic four-dimensional imaging with dual-view plane illumination microscopy. Nat Biotechnol. 2013; 31(11):1032–8. 19. Strnad P, Gunther S, Reichmann J, Krzic U, Balazs B, de Medeiros G, Norlin N, Hiiragi T, Hufnagel L, Ellenberg J. Inverted light-sheet microscope for imaging mouse pre-implantation development. Nat Methods. 2016;13(2):139–42. 20. Glaser AK, Reder NP, Chen Y, McCarty EF,Yin C, Wei L, WangY,True LD, LiuJTC. Light-sheet microscopy for slide-free non-destructive pathology of large clinical specimens. Nat Biomed Eng. 2017;1:0084. 21. Wu Y, Chandris P, Winter PW, Kim EY, Jaumouille V, Kumar A, Guo M, Leung JM, Smith C, Rey-Suarez I, et al. Simultaneous multiview capture and fusion improves spatial resolution in wide-field and light-sheet microscopy. Optica. 2016;3(8):897–910.
BMC Biology – Springer Journals
Published: May 29, 2018
Access the full text.
Sign up today, get DeepDyve free for 14 days.