Data Processing and Analysis

Time requirements for the different imaging modalities were normalized to the number of acquisition sequences and the imaged sample volume according to the following equation:

Images were deconvolved using the LIGHTNING module in LAS X (Leica, https://www.leica-microsystems.com/science-lab/how-to-extract-image-information-by-adaptive-deconvolution/) and widefield images were furthermore processed with the Large Volume Computational Clearing module (LVCC, Leica). Computational Clearing is an opto-digital technology that removes the typical “blur”/”haze” of wide-field images. For that, it automatically takes all relevant optical parameters into account thus maintaining the real local dimensions of the sample. It allows widefield imaging into thick 3D specimens. LVCC parameters were automatically selected based on the respective objective used for imaging. Further Data Processing was performed on a Z640 Workstation (HP; Win10 Enterprise 64-bit; Intel Xeon CPU E5-2650 v3 @ 2.30GHz; 32.0 GB RAM; NVIDIA Quadro K2200 4 GB GDDR5 (DirectX 12.0)).

Contrast adjustment for display purposes and image analysis was performed using Imaris (Bitplane) version 9.5. We used the ‘Surface’ and ‘Spot Creation Wizard’s in Imaris (Bitlane) to translate fluorescence data into volumetric, representative surfaces or point-like spots. For each object, a variety of parameters is calculated. For surfaces, these parameters include for example the position (x, y, z), volume, sphericity, median fluorescence intensities for all channels, volume and the distance to the closest object of defined surfaces/spots. For the analysis of the distance of MHC-II cells to glomeruli, we asked for the distance of MHC-II surfaces to the closest glomeruli surface (“shortest distance”). We next exported selected parameters into FlowJo, analyzed the distribution of MHC-II cells towards their nearest glomerulus as dot plots and histograms, and defined populations according to their distribution (visible separate peaks or populations). Defined populations were verified by looking at the cellular positioning using the X and Y parameters in FlowJo dot plots.

All statistics relevant for analysis were exported as a collection of csv files and subsequently edited and concatenated into a single summary file with our open-access standalone Python application [https://gitlab.com/kepplerlab/imaris_statistics_converter; incorporates ‘pandas’ module (44, 45)] for compatibility with further analysis in FlowJo (BD Biosciences).

For comparative analysis of signal and background fluorescence intensities, we acquired overviews with identical imaging settings at 100 µm depth in z from the first frame filling image plane. Using Fiji (ImageJ) (46, 47) we measured the mean pixel grey values of marked areas of defined signal (glomeruli) or background, at 10 randomly selected positions each in the stitched but otherwise unprocessed images, to calculate the average signal or background intensity, respectively.

Descriptive size and shape analysis of glomeruli structures in images of histological stains was performed using Fiji (ImageJ) (46, 47). Radom areas throughout the entire section were marked with the elliptical selections tool and area and roundness (inverse aspect ratio) of the selections was measured.