Stereological analyses of the whole human pancreas

Abstract

The large size of human tissues requires a practical stereological approach to perform a comprehensive analysis of the whole organ. We have developed a method to quantitatively analyze the whole human pancreas, as one of the challenging organs to study, in which endocrine cells form various sizes of islets that are scattered unevenly throughout the exocrine pancreas. Furthermore, the human pancreas possesses intrinsic characteristics of intra-individual variability, i.e. regional differences in endocrine cell/islet distribution, and marked inter-individual heterogeneity regardless of age, sex and disease conditions including obesity and diabetes. The method is built based on large-scale image capture, computer-assisted unbiased image analysis and quantification, and further mathematical analyses, using widely-used software such as Fiji/ImageJ and MATLAB. The present study includes detailed protocols of every procedure as well as all the custom-written computer scripts, which can be modified according to specific experimental plans and specimens of interest.

Figure 1. Computer-assisted large-scale image analysis.

(A) Virtual slice view of a human pancreatic section (female, 2.2-y old, non-diabetic; nPOD #6107 Tail 02) immunostained for insulin (green), glucagon (red), somatostatin (white) and nuclei (blue). A series of contiguous optical panels of a specimen is collected and merged into a single image montage. A composite is made by merging four overlapping virtual slice images. (B) Views of each channel showing cellular composition of an islet cluster from the virtual slice in (A). (a) Beta-cells. (b) Alpha-cells. (c) Delta-cells. (d) Nuclei. (e) A composite of all three endocrine cells and nuclei. (f) Total endocrine cell area shown as a converted 8-bit mask after automatic thresholding. (g) Nuclei of endocrine cells. (h) Total islet area that includes unstained fractions such as intraislet capillary. Note that each islet including small clusters is designated with an identification number. Table summarizing data obtained through the computer-assisted large-scale analysis.

Figure 2. Measurements of spatial distribution of islets and each endocrine cell type within an islet.

(Aa) Examples of measured centroids of islets and small clusters. (b) Reconstructed endocrine cell distribution within each islet (beta-cells in green, alpha-cells in red and delta-cells in blue). (B) Sequential analysis by the Image analysis macro. (a) Watershed segmentation applied to an 8-bit image converted from DAPI fluorescent signals. (b) Nuclei assigned with each specific identification number that corresponds to individual spatial coordinates of the nuclei. (c) Enlarged view showing a defined perimeter of each nucleus (a yellow line) with its own identification number. (d) Beta-cells. (e) Alpha-cells. (f) Delta-cells. (C) Automated identification of each cell type. A case of a single beta-cell is shown as an example. The perimeter of an identified nucleus (outlined as a yellow line) is expanded by 1 pixel (~1 µm) to detect the cytoplasmic region immediately surrounding the nucleus, which will determine its specific cell type.

Figure 3. Schematic workflow.

Starting with multi-fluorescent immunohistochemistry, 4-channel imaging is performed in a virtual-slice mode on a microscope. Captured images (1–2 GB each) are first transferred to Server (e.g. with storage capacity of 30 TB. Note that the Server serves for both image processing and storage of large data sets). Processed images are retrieved at Workstation for the subsequent image analysis. Our labeling scheme for processed images is shown on the right together with approximate processing time at each step. First, “Background subtraction” is applied to all channel images using Fiji/ImageJ. On the DAPI image, “the pancreas area” is manually contoured and is saved as a “.roi” file. The determined pancreas area is applied to images that have been background-subtracted (“BS.**.tif”) and the resulting image files are saved as “Cropped.**.tif”. Then by running the “Image processing macro”, manually set the threshold of optimal intensity in gray scale by following the instruction embedded in the macro. The macro automatically creates a set of image files that are ready for quantification. These files in the “Final” folder are manually uploaded to the Server for the quantification of various parameters. Run the “Image analysis macro” and Excel files will be automatically created that contain results with selected parameters. In the bottom box, examples of various analyses are listed with reference to each figure that are shown in the present study.

Figure 4. Selection of islet-rich areas in pancreas sections.

(A) Simulation 1 using a pancreatic tissue section from the head region that contains PP-cell rich and poor area. (a) Pancreas area is contoured in light blue. PP-cell rich area is encircled in yellow. Inset: the first boxed area captured using a 10x objective shown in red. Optical panels were sequentially selected as enumerated in yellow boxes. (b) Selected panels were systematically added up to a total of 153 panels. (c) The cumulative averaged value of each quantification is plotted. An asterisk depicts the value obtained by our large-scale quantification. (B) Measurement of endocrine cell area on every 50^th section within a block. (a) Whole pancreas analysis of endocrine cell area (i.e. PP, beta and alpha-cells) from head, body to tail region (from left to right). (b) Block #4 was selected because it showed a similar endocrine cell mass to Block #6. The entire block of #4 was cut in 5 μm thickness, which yielded >300 sections. Total endocrine cell area was measured in every 50^th section by applying our large-scale analysis. (C) Simulation 2 using the nPOD specimen shown in Fig. 1. (a) To examine the commonly used selection of islet-rich areas, optical panels were sequentially selected up to 30 panels, as enumerated in yellow boxes. Note that each area of an optical panel is equivalent to that of using a 20x objective. (b) The range of overestimation resulting from the selection is shown compared to the large-scale analysis.

Figure 5. Comparison to the widely used point counting morphometry.

(A) An example of comparison using a grayscale panel of insulin-positive beta-cells, overlaid with a regular square grid with vertices spaced at 25 µm. The extent of beta-cell mass classified is outlined in yellow. Vertices registered as positive are highlighted in green. (B) Simulation using the same pancreatic tissue section from the head region that contains PP-cell rich and poor area shown in Fig. 2A. (C) Comparison in each optical panel. Differences are normalized to the measurement of our method and shown as a ratio. Underestimated values are shown in red bars. (D) A representative view of panels that resulted in underestimation. (E) An example of quantification in mouse and human islets when point counting morphometry is applied.

Figure 6. Islet size distribution, cellular composition.

Specimens from head, body and tail region of nPOD #6107 were analyzed. (A) Frequency of islet size (gray bar) and ratios of beta-cells (green), alpha-cells (red) and delta-dells (blue) within islets are plotted against islet size; means ± SEM. Note that islet size is presented as a logarithmic scale considering the high number of small islets and the low number of large islets. In addition, islet area is divided by the single-cell area (178 μm²31) to make them as dimensionless values representing the number of cells in a given islet area. See the conversion between logarithmic islet area (logarithmic) and effective diameter (μm). (B) Fraction of islet size distribution (gray bar) and total islet area (red line). (C) 3D visualization of islet size and shape distribution. Each dot represents a single islet/cluster with reference to size (area) and shape (circularity and Feret’s diameter). The density of islets is color-coded from sparse to dense. (D) 3D scatter plots highlighting PP-containing islets.

Figure 7. Whole pancreas analysis.

(A–E) non-diabetic subjects. (F–J) patients with T2D. X-axis: Block number from head, body to tail region (from left to right).

Figure 8. Simulation of tissue sampling.

(A) Sampling scheme. The boundary of the head region was anatomically determined at the neck, and the remaining portion was divided in half for body and tail regions. (B) Fold changes in islet cell mass from 5 types of simulations compared to the whole pancreas analysis. (C) Fold changes in ratios of islet cells (i.e. beta/alpha, beta-delta and alpha/delta cells) from selecting 100 large islets compared to the whole pancreas analysis.