Automated Image Analysis Pipeline Development to Monitor Disease Progression in Muscular Dystrophy Using Cell Profiler

Abstract

Muscular dystrophies are inherited disorders that are characterized by progressive muscle degeneration. These disorders are caused by mutations in the genes encoding structural elements within the muscle, which leads to increased vulnerability to mechanical stress and sarcolemma damage. Although myofibers have the capacity to regenerate, the newly formed myofibers still harbor genetic mutation, which induces continuous cycles of muscle fiber death and regeneration. This repeated cycling is accompanied by an inflammatory response which eventually provokes excessive fibrotic deposition. The histopathological changes in skeletal muscle tissue are central to the disease pathogenesis. Analysis of muscle histopathology is the gold standard for monitoring muscle health and disease progression. However, manual, or semi-manual quantification methods, are not only immensely tedious but can be subjective. Here, we present four image analysis pipelines built in CellProfiler which enable users without a background in computer vision or programming to quantitatively analyze biological images. These image analysis pipelines are designed to quantify skeletal muscle histopathological staining for membrane damage, the abundance and size distribution of regenerating muscle fibers, inflammation via quantification of CD68+ M1 macrophages, and collagen deposition. Additionally, we discuss methods to address common errors associated with the quantification of microscopy images. These automated tools can not only improve workflow efficiency but can provide a better understanding of the histopathological progression of muscular dystrophy.

Keywords: Automated Histopathology Analysis; CellProfiler; Evan’s Blue Dye; Fibrosis; Muscle Regeneration; Muscular Dystrophy.

Figure 1:

Processing and quantification of EBD+ myofibers by CellProfiler. A) Representative images of gastrocnemius muscle cross-sections from 4–6-month-old B10 and mdx mice immunolabeled with laminin α (green), and EBD (red). Scale bar = 100μm. B) Quantification analysis of EBD-positive muscle fiber expressed as the percentage of the total number of muscle fibers in WT and mdx mice (n = 5 mice/group). **p<0.01. C) Overview of the image analysis workflow. D) A sample mdx image was fed into the EBD quantification pipeline, which first inverts the original image. The inverted image is then converted into a grayscale image with the red and green channels combined. The pixels are then squared to mask the EBD+ cells. The “not” EBD cells are then identified and counted. E) The original image is retrieved and split into separate grayscale images representing each of the color channels (red, green, and blue). F) The green channel grayscale image is then inverted and used for subsequent identification of all muscle cells.

Figure 2:

Examples of correct and incorrect image thresholding. Improper thresholding in the “IdentifyPrimaryObjects” module can impair object detection. A) The inverted green channel (grayscale) image used for muscle cell identification. B) Poor muscle cell identification. C) Improved muscle cell identification was achieved by adjusting the upper and lower bounds of the threshold.

Figure 3:

Processing and quantification of eMyHC+ myofibers by CellProfiler. A) Representative images of gastrocnemius muscle cross-sections from 4–6-month-old B10 and mdx mice immunolabeled with laminin α1 (red), and eMyHC (green). Scale bar = 100 μm. B) Boxplot displaying quantification of the sample eMyHC-positive muscle fiber expressed as the percentage of the total number of muscle fibers in WT and mdx mice (n = 5 mice/group). **p<0.01. C) Histogram showing the size distribution (in pixels) of the eMyHC+ fibers obtained from the sample dataset. D) Overview of the image analysis workflow. E) A sample mdx image was fed into the pipeline and converted into separate grayscale images representing each of the color channels. Here, only the red and green channels were used for subsequent processing. F) The red channel grayscale image was inverted and used for the G) identification of the muscle cells. H) Outlines of the identified eMyHC+ cells.

Figure 4:

Processing and quantification of inflammation via detection of CD68 deposition by CellProfiler. A) Representative images of gastrocnemius muscle cross-sections from 4–6-month-old B10 and mdx mice immunolabeled with laminin α1 (red), and CD68 (green). Scale bar = 100μm. B) Boxplot illustrating the quantification of the sample CD68 dataset images showing the percent area of CD68 deposition in each gastrocnemius muscle of WT and mdx mice (n = 5 mice/group). **p< 0.01. C) Overview of the image analysis workflow. D) A sample mdx image was fed into the CD68 quantification pipeline which first converts the color image into separate grayscale images representing each color channel. The green channel grayscale image was used for further processing. E) Outlines of the identified CD68. F) The identified objects (CD68) were then converted into a binary black and white image for easy quantification of CD68 deposition (% area) within the muscle cross-section.

Figure 5:

Processing and quantification of collagen deposition by CellProfiler. A) Representative images of gastrocnemius muscle cross-sections from 4–6-month-old B10 and mdx mice stained with Picrosirius red. Scale bar = 100μm. B) Boxplot illustrating quantification of percent area of collagen deposition in each muscle section (n = 5 mice/group). **p<0.01. C) Overview of image analysis workflow. D) A sample mdx image was fed into the collagen quantification pipeline, which first converts the color image into separate grayscale images representing each of the color channels. The green channel grayscale image was selected for subsequent processing because the collagen had a prominent appearance in this channel. E) The green channel image was inverted and used for F) identification of the collagen. G) The identified collagen was then converted into a binary black and white image for simple quantification of collagen deposition (% area) within the muscle cross-section.