Enhanced CD4+ and CD8+ T cell infiltrate within convex hull defined pancreatic islet borders as autoimmune diabetes progresses

Alexander J Dwyer,Mohannad Alkhatib,Brian T Fife,Jason S Mitchell,Christopher G Tucker,Tijana Martinov,Jacob M Ritz,Nubia Silva

doi:10.1038/s41598-021-96327-2

Alexander J Dwyer, Mohannad Alkhatib + Show 6 more

Open Access

https://doi.org/10.1038/s41598-021-96327-2

Copy DOI

Abstract

The notion that T cell insulitis increases as type 1 diabetes (T1D) develops is unsurprising, however, the quantitative analysis of CD4+ and CD8+ T cells within the islet mass is complex and limited with standard approaches. Optical microscopy is an important and widely used method to evaluate immune cell infiltration into pancreatic islets of Langerhans for the study of disease progression or therapeutic efficacy in murine T1D. However, the accuracy of this approach is often limited by subjective and potentially biased qualitative assessment of immune cell subsets. In addition, attempts at quantitative measurements require significant time for manual analysis and often involve sophisticated and expensive imaging software. In this study, we developed and illustrate here a streamlined analytical strategy for the rapid, automated and unbiased investigation of islet area and immune cell infiltration within (insulitis) and around (peri-insulitis) pancreatic islets. To this end, we demonstrate swift and accurate detection of islet borders by modeling cross-sectional islet areas with convex polygons (convex hulls) surrounding islet-associated insulin-producing β cell and glucagon-producing α cell fluorescent signals. To accomplish this, we used a macro produced with the freeware software ImageJ equipped with the Fiji Is Just ImageJ (FIJI) image processing package. Our image analysis procedure allows for direct quantification and statistical determination of islet area and infiltration in a reproducible manner, with location-specific data that more accurately reflect islet areas as insulitis proceeds throughout T1D. Using this approach, we quantified the islet area infiltrated with CD4+ and CD8+ T cells allowing statistical comparison between different age groups of non-obese diabetic (NOD) mice progressing towards T1D. We found significantly more CD4+ and CD8+ T cells infiltrating the convex hull-defined islet mass of 13-week-old non-diabetic and 17-week-old diabetic NOD mice compared to 4-week-old NOD mice. We also determined a significant and measurable loss of islet mass in mice that developed T1D. This approach will be helpful for the location-dependent quantitative calculation of islet mass and cellular infiltration during T1D pathogenesis and can be combined with other markers of inflammation or activation in future studies.

Highlights

Background correctionTo eliminate user bias influenced by treatment conditions when performing background correction, image randomization occurs prior to any alterations of the raw data
We demonstrate that non-infiltrated islets can be precisely modeled by overlaying minimal area convex polygons that entirely capture the sum of insulin (β cell) and glucagon (α cell) signals to distinguish individual contributions of immune cell subtypes during inflammation either within or outside the islet border
Before demonstrating application of our convex hull approach to determine inflammation in non-obese diabetic (NOD) islets through type 1 diabetes (T1D) progression, we first present an overview of the macro to show its accessibility

Summary

Introduction

To eliminate user bias influenced by treatment conditions when performing background correction, image randomization occurs prior to any alterations of the raw data. A blinded user first marks each islet in the image series using an arrow to ensure background correction is optimized to the XY position of the islet of interest. This step is necessary to properly identify the islet of interest in images that contain multiple islets. Image randomization proceeds using the Fisher–Yates shuffle statistical randomization algorithm to decrease any user bias for treatment groups or experimental conditions between tissues. Identifying an ROI with bright intensity of only background signal is critical; an ROI containing only dim background signal will fail to remove background regions of brighter intensity (false positives) while an ROI containing any foreground signal will remove all foreground regions of similar or lower brightness (false negatives)

Objectives

Methods

Results

Conclusion