Arivis vision 4d

Whole heads from Shh-GFP embryos at E11.5 E12.5, E13.5 and E14.5 were cleared using a modified CUBIC protocol (Susaki et al., 2014 (link)). In brief, embryos were fixed by using 4% PFA in PBS for 4 hr at four degrees before incubating in CUBIC one solution (25% urea, 25% N,N,N′,N′-tetrakis-(2-hydroxypropyl) ethylenediamine and 15% Triton X-100) at 37°C under shaking conditions for 3 days. Subsequently, the samples were washed in PBS at RT. Next, samples were immersed in CUBIC two solution (50% sucrose, 25% urea, 10% 2,2′,2″-nitrilotriethanol, and 0.1% Triton X-100) and left shaking at RT for an additional 2–3 days before image acquisition.
Whole embryo head (E11.5–E14.5) GFP fluorescence images were acquired on a Light sheet Z.1 microscope (Zeiss) using a × 5 (EC Plan Neofluar 5×/0.16) detection objective,×5/0.1 illumination optics, and laser excitation at 488 nm. Samples were imaged in CUBIC two solution with a measured refractory index of 1.45. Each plane was illuminated from a single side of the sample. Whole images were obtained through tile scanning. 3D-rendered images were visualized with Arivis Vision4D for Zeiss (v. 2.11) or Imaris (v. 7.4.2, Bitplane).
Bitplane IMARIS software was subsequently used for 3D visualization and analysis of the light sheet tiles. By using the surface option in IMARIS the different parts of Shh-GFP have been highlighted.

3D measurements were performed using Arivis Vision4D (Zeiss) while ImageJ was used for all 2D measurements. The details of each analysis pipeline used in this study are listed below:

Lightsheet microscopy 3D images were acquired with a Zeiss Z.1 lightsheet microscope using the 10×/0.2 NA excitation and 20×/1.0 water immersion detection objective lenses, 488 nm and 561 nm excitation and the respective green (505–545 nm) and red (575–615 nm) emission. Image processing and visualisation were performed in ZEN (Zeiss, GmbH, Aalen, Germany) and Arivis Vision 4D (Zeiss, GmbH, Aalen, Germany) software, respectively.

ER, mitochondria, lipid droplets, nucleus and cell borders were segmented using the 3dEMtrace platform of ariadne.ai (https://ariadne.ai/) by using a customized convolutional neural network (CNN) architecture based on 3D U-Net. Binary tiff masks (for ER and nucleus) and instance-based segmentation (for mitochondria, LD and cell borders) were generated, to assign a unique identifier to each organelle. The raw data (from Lean-Fed and Obese-Fed datasets) used to extract mitochondria instance segmentation were published previously^{10 (link)}. The mitochondria data from these two and other new datasets are extracted and re-segmented individually and used to quantify inter-organelle interactions. The false positive connections between different mitochondria were improved by the instance-based segmentation. Data were analyzed in Arivis Vision 4D (Zeiss) software by creating objects from segmented data. After generating the 3D objects (full connectivity in x, y, z), proof-reading was done to eliminate the unspecific ER objects less than 100,000 voxels in all datasets. Minimum mitochondria object voxel volumes are as follows per dataset: Lean Fed: 21696 Lean Fasted: 92243, Obese Fed: 50382, Obese Fasted: 80293, Obese LacZ: 50437, Obese RRBP1: 50234. Organelles were then re-segmented by the cell that they belong (5 hepatocyte volumes per condition).

Image acquisition was done in ECi using a Lavision Ultramicroscope II (Lavision Biotec, Germany) equipped with an Olympus MVPLAPO 2× (NA 0.50) objective lens and DBE-corrected LV OM DCC20 dipping cap. Images were recorded with a Neo sCMOS camera (Andor), using a 2× objective and a digital zoom of 0.63×. Samples were optically sectioned using a z-step size of 10 µm, resulting in 5.08 µm × 5.08 µm × 10 µm voxel size (16 bit per pixel). A 488 nm laser and 561 nm laser were used in combination with a 525/50 nm and 620/60 nm emission filter, respectively. Sagittal optical sections were acquired in a mosaic of four tiles using the left and right light sheets separately, respectively, for the left and right parts of the mosaic. The resulting image sets were then imported into Arivis Vision 4D software (Zeiss, Germany; RRID SCR_018000) for stitching and acquisition of movies, images and three-dimensional renderings.