Poster R6 - Enhancing Islet Cell Transplantation Procedures Through Automated Quantification of Islet Morphology and Insulin Expression

Introduction: Islet cell function and viability are critical factors in determining successful outcomes of islet transplantation procedures to reverse diabetes. During isolation procedures, islets are subjected to mechanical, chemical, hypoxic, and oxidative stress. Assessing the quality of islets poses limitations by conventional methods that are restricted to brightfield microscopy. Moreover, insulin content cannot be evaluated prior to transplantation, as it necessitates a fixing and staining process. We have thus established an imaging protocol using confocal microscopy and created a computational pipeline to assess fluorescence images of islets and quantify their viability. Utilizing image processing techniques, overall insulin content and islet morphology can be analyzed and evaluated in relation to the islet’s corresponding brightfield image.

Materials and

Methods: Pancreatectomies were performed of wild type mice. Pancreatic tissue was digested in 5 mg/ml collagenase at 37°C in a water bath for approximately 10 min and washed with HBSS supplemented with 10% BSA and penicillin-streptomycin (10,000 U/mL). Islets were isolated through a gradient using a Biocoll-Separating Solution (density 1.10 g/mL and 1.077 g/mL) and centrifugation, handpicked using a micropipette, and seeded into glass bottom tissue culture dishes in RPMI supplemented with 10% FBS, 10 mM HEPES, penicillin-streptomycin (10,000 U/mL), and 100 mM sodium pyruvate for 3 days. The cells were then washed and fixed with 4% formaldehyde for 15 min, permeabilized with 0.1% Triton X-100 solution for 15 min, and blocking was performed with 2% BSA solution for 60 min. The islets were immunostained using Alexa Fluor 488-congugated insulin antibody and imaged on a Nikon A1R HD25 laser scanning confocal microscope.

Results, Conclusions, and Discussions: A series of algorithms were constructed within a Jupyter Notebook to facilitate the automated analysis of the confocal images. Z-stacks of the islets are loaded into the framework and undergo multiple processing steps to create a binary mask of the largest identified area along the z-axis (Fig. 1). This mask is then overlayed onto each frame where total fluorescence intensity of the insulin is calculated. The automated pipeline measures multiple morphological features of the islet, including the area, roundness, eccentricity, and smoothness. These morphological parameters are consistent with the assessment guidelines for islet quality evaluation. Instead of relying on manual scoring methods, we performed initial analysis of the confocal Z-stacks of islet images using the automated methodology. As shown in Figure 2, the eccentricity of islet is found to correlate with the insulin content. The automated analysis is therefore demonstrated to not only calculate quantifiable parameters of the islets but also correlate with the islet quality (e.g., insulin content).

In summary, imaging processing and automated analysis of islets provide rapid assessment of islet morphology and correlation of insulin expression to enhance or supplement conventional methods for determining islet cell quality. With the implementation of these image processing and automated analysis techniques, quantifiable morphology characteristics of islets can be measured and correlated with the corresponding brightfield image. This is expected to enable clinicians to utilize a data-based approach to selecting highest quality islets for transplantation. Moreover, as this dataset increases, other parameters associated with islet quality can be quantified. Additional future work will include the development of a deep learning model which will ultimately be able to predict an islet’s population of beta cells and insulin expression throughout the islet. The aim of this work is consistent with FDA-issued guideline that an appropriate assay must be established to measure potency of the islet product to satisfy the regulatory requirements, such as the accuracy, sensitivity, specificity, and reproducibility of the potency assay. It is interesting to note that the FDA guidelines categorize many of the currently utilized biological assays as not rapid enough and recommend exploring the development of rapid analytical assays.

Acknowledgements (Optional): Authors acknowledge generous support from the Alfred R. and Janet H. Potvin Endowment (MC), the Baylor University Medical Center (MCL), the Cancer Prevention and Research Institute of Texas grant (RR220015, JML), and the National Heart, Lung, and Blood Institute T32 Grant (HL134613). Confocal microscopy was made available through a NIH Instrument Grant (S10 OD025230).