Supervised Machine Learning for Semi-Quantification of Extracellular DNA in Glomerulonephritis

Glomerulonephritis causes pathological glomerular injury with considerable cell death and extracellular DNA release. This protocol enables the visualization of extracellular DNA at the site of release during pathological injury. The advantage of this technique is that it measures extracellular DNA from cell death in the kidney.

As opposed to serum or urine, extracellular DNA measurement is a surrogate marker for pathological damage. The significance of this method is that both the extracellular DNA staining and image analysis are easily transferrable to other autoimmune diseases such as rheumatoid arthritis and lupus. This method requires a basic knowledge of kidney histology.

To enable the accurate detection of glomerular extracellular DNA, the researcher must be able to correctly identify glomeruli. For extracellular DNA staining, first heat the samples of interest in a slide rack in 60 degrees Celsius for 60 minutes before immersing the slides in two 40-minute changes of solvent in a fume hood. After the second immersion, rehydrate the samples two times in 100%ethanol and one time in 70%ethanol for five minutes per treatment.

After the last treatment, use tongs to place the slides into boiling antigen retrieval solution in a pressure cooker on high for 10 minutes. After unmasking the antigen epitopes, transfer the pressure cooker from heat into a sink and immediately run cold tap water over the lid. Allow the slides to equilibrate for 20 minutes in the antigen retrieval solution before washing the samples two times in 0.01 molar PBS on an orbital shaker for five minutes per wash.

After the second wash, use a hydrophobic pen to draw circles around the kidney tissue samples and block any nonspecific staining with 60 microliters of 10%chicken serum in 5%bovine serum albumin for 30 minutes at room temperature. At the end of the incubation, carefully replace the blocking solution with 60 microliters of the primary antibody of interest in a humidity chamber overnight at four degrees Celsius. The next morning, wash the slides on an orbital shaker in PBS for two minutes per wash before adding 60 microliters of an appropriate fluorophore conjugated secondary antibody to each slide for a 40-minute incubation at room temperature.

At the end of the incubation, wash the slides in PBS as demonstrated, followed by treatment in 0.3%Sudan black in 70%ethanol to quench any potential formalin-induced autofluorescence. After 30 minutes, wash the slides in tap water to remove any precipitate and immerse the slides in PBS for 10 minutes to prevent any further Sudan black precipitate formation. At the end of the incubation, mount the slides onto confocal glass coverslips with three 60 microliter drops of mounting solution supplemented with DAPI and seal the coverslips with nail polish.

Allow the slides to cure for 24 hours at room temperature. Store them at four degrees Celsius protected from light until imaging by confocal laser scanning microscopy according to standard protocols. Be sure to seal the coverslips with nail polish to prevent any movement during imaging and to clean the coverslips of ethanol to enhance the imaging resolution.

After imaging 20 glomeruli per slide, open the trainable Weka Segmentation plugin in ImageJ and drop the first image onto the ImageJ toolbar. Click image, color, and split channels. An image showing the primary antibody staining will appear.

To create a merged file of the single images, click color and merge channels. A popup box will appear asking for a color to be assigned to each channel. Check the create composite and keep source images boxes and click OK.A merged composite image will appear.

Using the primary antibody staining as a guide, draw a region of interest around the glomerular tuft. To perform a Gaussian blur on the single blue DAPI image, click process, filters, and Gaussian blur. Set the blur at one to two.

The image will become a little blurry, but the nuclei will appear smooth, the background will be softened. Next, click plugins, segmentation, and trainable Weka Segmentation, and use the free hand tool to trace the intact nuclei. When all of the nuclei of interest have been added, the add class one box, use the line tool to delineate the areas of background to the class two box.

Click create new class and use the line tool to outline the DAPI-stained extranuclear DNA. Click train classifier and get probability. Toggle the mouse to obtain a black and white image of all the components with the object of selection highlighted in white.

To duplicate this image, click image and duplicate. Under image, adjust and threshold. Use the mouse button to adjust the threshold until only the extracellular DNA is observed.

After thresholding, click process, binary, and make binary, and copy the previously generated glomerular region of interest. Press Control Shift E to select edit, selections, and restore to restore the selection. To measure only the extracellular DNA particles within the glomerulus, click analyze particles and OK.A result sheet showing the number of particles, area, average pixels, and percentage of glomeruli containing extracellular DNA will be generated.

To save the classifier as an extracellular DNA classifier, click save classifier and save the data. To reapply the model to subsequent images, repeat the initial image processing as demonstrated before selecting load classifier and the saved classifier file. The log menu will pop up and run the model.

Once the model has run, click load data and select the extracellularDNA.arff. Then click create result. If the resulting image has failed to pick up all the nuclei, background, or extracellular DNA, click retrain classifier and add more classifiers as necessary.

Here, single channel images showing representative DNA and beta actin staining are shown. Merging the two images allows a region of interest to be drawn around the glomerular area for measurement of the extracellular DNA within the mouse kidney glomeruli using trainable Weka Segmentation as demonstrated. This model can be adopted to identify extracellular DNA in kidney biopsy specimens from a control patient or comparison to that of a kidney biopsy from a patient with myeloperoxidase ANCA-associated vasculitis.

In addition, this program can be translated to other stains within kidney tissue samples, for example, to stain for neutrophil extracellular traps and the deposition of extracellular myeloperoxidase in ANCA-associated vasculitis. Importantly, no significant differences are observed when the same data is analyzed by multiple users. The most critical step of the analysis is providing enough examples of background, nuclei, and extracellular DNA with glomeruli as classifiers for the Weka Segmentation program to learn from.

This method is easily transferrable. For example, further analysis of the different types of cell death can be investigated through staining for specific antibodies against markers for apoptosis or necrosis. Our technique can be used with other inflammatory processes such as assessing TLR expression in autoimmune kidney disease or extracellular myeloperoxidase expression in human kidney biopsies.