(#16) Quantitative analysis of pancreatic T cell infiltration in type 1 diabetes: a numbers game

PRESENTED BY: Teresa Rodriguez-Calvo

First NameLast NameAffiliation/Institution
PaolaApaolazaHelmholtz Zentrum Munich
DianaBalcaceanHelmholtz Zentrum Munich
JoseZapardiel-GonzaloHelmholtz Zentrum Munich
TeresaRodriguez-CalvoHelmholtz Zentrum Munich


Islet infiltration is considered a hallmark of type 1 diabetes (T1D). Several definitions of insulitis have been established over the years based on multiple immune cell populations, which has made data difficult to standardize. Current consensus defines insulitis as a minimum of three islets infiltrated by ≥ 15 CD45+ cells or by ≥ 6 CD3+ cells. While these are useful references, none of these definitions accurately consider the size of the islets. Thus, we aimed to quantitatively analyze the extension and density of T cell infiltration in the pancreas of non-diabetic (ND), autoantibody positive (AAb+) and T1D donors.


T cell infiltration was quantified in all the islets from 11 ND (n=4797), 2 single and 5 double AAb+ (n=1526) and 10 T1D donors (n=2235, 0-2 years of disease duration). Frozen sections were stained for insulin, glucagon, CD3 and CD8 by immunofluorescence. CD3+ and CD8+ cells were identified as areas of staining above background and optimized threshold values for intensity and cell size by using QuPath (University of Edinburgh). All images were manually checked to identify insulin containing islets (ICIs) or insulin deficient islets (IDIs), and corrected for any possible errors.

Summary of Results

The proportion of infiltrated islets (≥ 1 CD3+ cells inside or immediately adjacent to the islet) was 17.2% in ND, 28.7% in AAb+ and 40.7% in T1D donors. When the threshold for insulitis was applied (≥ 6 CD3+), 0.4% in ND, 3.6% in AAb+ and 8.2% of the islets in T1D donors were accounted for. Similar values were found for CD8+ cells using 1 or 6 cells as threshold. Interestingly, the proportion of islets with at least 1 CD4+ cell was significantly higher in both, AAb+ (16.1%) and T1D donors (21.7%) compared to ND donors (7%). As expected, in T1D donors, ICIs were preferentially infiltrated independently of the T cell population.
Then, CD3+ density was calculated for all the islets. It was significantly higher in T1D (74.8 CD3+ cells/mm2) compared to AAb+ (34.2 CD3+ cells/mm2) and ND (16.1 CD3+ cells/mm2) donors. When the analysis was restricted to islets that contained ≥ 1 CD3+ cells, cell density remained significantly higher in T1D donors. However, when the threshold for insulitis was used, mean cell density per se was not statistically different between the groups, regardless of the cell population. Next, cell density was compared between ICIs and IDIs from the same individuals within the T1D group. When all the islets were analyzed (without any threshold), there was a tendency to increased T cell density in ICIs compared to IDIs. When the threshold of ≥ 1 CD3+ cells was applied, T cell density was comparable between both islet types, whereas in insulitic islets, it tended to be higher in IDIs than in ICIs regardless of the T cell population. Lastly, we aimed to define a threshold value of islet T cell density able to distinguish ND from T1D donors. The optimal threshold was found to be 33.5 CD3+ cells/mm2 (average 93% specificity and 87% sensitivity) when a minimum of 40 islets per donor were analyzed, and was validated in an external image dataset obtained from nPOD.


Using novel image analysis tools, we show that a high proportion of infiltrated islets and high islet T cell density are defining features in T1D. Our data suggest that it is the proportion of infiltrated islets rather than the number of infiltrating cells what dramatically changes during the course of T1D. This indicates that a “mild wave “of infiltrating cells slowly extends in the pancreas as disease progresses, whereas large accumulations of cells are rare. Furthermore, our data shows that, in T1D donors, while the proportion of infiltrated ICIs is significantly high, T cell density values do not statistically differ between ICIs and IDIs. This indicates that the number of cells that infiltrate an islet (in relation to the islet size) is similar in the presence or absence of insulin, but a larger proportion of ICIs are actually infiltrated at any given time. Lastly, we believe that the use of cell density values, which account for magnitude of infiltration and size of the islets, rather than absolute cell numbers per islet should be implemented. Therefore, we proposed a new threshold to define insulitis based on T cell density. Further analysis and establishment of a T cell density range to complement the current definition of insulitis will help to fulfill the need of a better understanding of the role of T cells in the course of T1D.