Revisiting type 1 diabetes progression in the non-obese diabetic mouse

Abstract

As conventional diagnostic criteria in non-obese diabetic (NOD) mice may not reflect the early phase of type 1 diabetes, the applicability of NOD mouse models for early type 1 diabetes in humans can be limited. We therefore aimed to assess stages of beta cell loss in early type 1 diabetes at blood glucose (BG) levels of 80–200 mg/dl in age-stratified NOD mice. Islet composition of five pancreas sections each from 38 female NOD mice was determined using multiplexed-immunohistochemistry staining and Cell2Grid/rule-based automated islet identification. Visual analysis of fm-IHC images led to classification of islet stages. Average islet stage per animal (Islet Score) was correlated with average BG to identify NOD mice subgroups of disease progression. We categorized 3324 islets into islet stages, describing beta cell loss from 0 to 4.The proportion of insulin-deficient islets increased from BG > 126 mg/dl onwards. Three disease progression subgroups in NOD mice were identified: Non-diabetic (Islet Score < 1.0, BG < 126 mg/dl), pre-diabetic (Islet Score > 1.0, BG < 126 mg/dl) and early-diabetic (Islet Score > 1.0, BG > 126 mg/dl and diabetes progression model was established. The revised classification of type 1 diabetes in NOD mice and the resulting type 1 diabetes progression model improves adaptation of the NOD model to human pathophysiology.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-35483-9.

Introduction

The non-obese diabetic mouse (NOD mouse) is considered the main model for type 1 diabetes because it is genetically very susceptible and has a similar pathogenesis to human type 1 diabetes, and particularly because of the accelerated timeframe in which studies can be conducted in these animals ^1–3. Type 1 diabetes in NOD mice and humans is characterised by the autoimmune destruction of insulin-producing beta cells, possibly triggered by a combination of environmental, genetic, and metabolic factors. The pathological hallmark of type 1 diabetes is insulitis, an inflammatory infiltration around and within islets of Langerhans containing mainly CD4 + T helper cells and CD8 + cytotoxic T cells ^4–6.

NOD mice are commonly classified as diabetic when BG levels exceed 200 mg/dl (11.1 mmol/l). However, other thresholds (ranging from 175 mg/dl to 250 mg/dl; 9.7 to 13.9 mmol/l) have also been used ^7,8. Notably, massive islet destruction can occur already at much lower BG levels ^9–11, and thus the use of conventional BG thresholds does not allow classification of early disease stages in NOD mice in contrast to human disease diagnosis at fasting BGs above 126 mg/dl ^12,13.

In NOD mice, early diabetes development is currently assessed using age cohorts in which insulitis can be detected as early as 4 to 5 weeks of age. Beginning insulitis signifies the recognition of autoantigens by CD4 + and CD8 + T-cells in islets and insulitis is sometimes used to assess the disease state of the mice ^14–16. At 12 to 14 weeks of age, the first cases of diabetes can be diagnosed in NOD mice. The majority (about 80%) of female and 20% to 40% of male NOD mice are expected to have developed diabetes after week 20, however some animals remain disease-free for reasons which are not yet understood ^4,8,9,17,18.

To refine the assessment of disease progression in NOD mice, insulitis grading systems using haematoxylin and eosin (H&E) staining have been applied which assess the degree of insulitis next to an islet ^19–21. However, results from these insulitis grading systems lack information about the presence of insulin in islets. To overcome this shortcoming the concept of insulin-containing and insulin-deficient islets has been applied in human pancreas sections ^22,23. Nevertheless, this concept has not yet been implemented in NOD mice and there is still a lack of concepts that can reliably assess the stages of early disease progression in NOD mice in order to make the NOD mouse model a suitable model for early type 1 diabetes progression in humans.

The aim of this study was therefore to investigate the early phase of type 1 diabetes progression by assessing the stages of beta cell loss in NOD mice at BG levels below the conventionally applied BG threshold (200 mg/dl), and to introduce a novel classification of type 1 diabetes progression in an age-stratified NOD mouse cohort by integrating the concept of insulin-containing and insulin-deficient islets.

Research design and methods

Animal husbandry

NOD/ShiLtJ and C57BL/6J (BL6) strains (breeding stock purchased from Charles River Laboratories, Freiburg, Germany and Jackson Laboratories, Bar Harbor, ME, USA) were inbred in specific pathogen free (SPF) conditions at the animal facility at the Medical University of Graz. They were housed in a 12 h light/12 h dark cycle in individually ventilated cages including nest building material with access to standard food and water ad libitum. Animal experiments were approved by the Federal Ministry of Education, Science and Research (BMWFW-66.010-WF-V-3b-2016, BMWFW-66.010/0142-WF/V/3b/2017). The study followed the ARRIVE guidelines. It also considered the 3R principle. Staff regularly assessed health status of animals. NOD mice with last BG values below 200 mg/dl were included and stratified by age from five weeks to 40 weeks. We employed broad age stratification to capture different stages of disease development, as the NOD model exhibits heterogeneous onset and progression patterns that vary significantly across age. Thirty-four female NOD mice at around five (n = 4), ten (n = 3), 20 (n = 7), 30 (n = 6), 35 (n = 8) and 40 (n = 6) weeks of age were included. Further, two BL6 mice aged ~ 15 weeks (negative controls) and two ~ 16-week-old female NOD mice with their last BG measurements exceeding 600 mg/dl (late-diabetic controls) were used. Mice were anaesthetised by intraperitoneal injection of a ketamine/xylazine mixture (ketamine hydrochloride, Ketasol 100 mg/kg body weight, aniMedica, Senden, Germany; xylazine hydrochloride, Rompun 16 mg/kg body weight, Bayer, Leverkusen, Germany) into the lower left abdomen. Following deep anaesthesia, blood samples were collected via cardiac puncture, and euthanasia was achieved through exsanguination. All methods were performed in accordance with relevant guidelines and regulations.

Blood for BG measurements was obtained via tail puncture and the average of the BG level over the last two weeks (four values) was used for all BG analyses. Non-fasting BG was measured twice a week with a glucometer (Accu Check by Roche, Mannheim, Germany) starting at ten weeks of age. Due to the lack of BG measurements in NOD mice aged five weeks they were only included in the initial islet stage analysis.

Histology and fluorescent multiplexed immunohistochemistry staining

Animals were sacrificed at the respective ages, pancreas was extracted, immediately fixed in 10% formalin (pH 6.90–7.10) and embedded in paraffin as described in ²⁴. Sections of formalin-fixed, paraffin-embedded tissue (FFPE) with a thickness of 2.5 μm were mounted on Superfrost Plus glass slides (J1800AMNZ from Fisher Scientific, Vienna, Austria). Five sections per mouse, each approximately 70 μm apart, were subsequently stained with H&E and fluorescent multiplexed immunohistochemistry (fm-IHC). Three representative sections were additionally stained with in-situ hybridization using padlock probe technology. Device, reagent and protocol specifications are given in Supplemental methods (Supplemental table T1, T2). The staining protocol and antibodies were optimized and validated by staining pancreatic tissue sections from BL6 and NOD mice. Islet and immune cell staining patterns were in accordance with the literature ^25,26.

Cells were segmented based on DAPI signals and marker-positive cells were detected by setting an optimized dye cytoplasm-positive threshold using High-Plex FL v2.0 module in the software Halo^® (Indica Labs, Albuquerque, NM, USA). An islet was defined by containing at least three cells stained positive for either insulin or glucagon. Beta cells were defined by their production of insulin, and alpha cells by their production of glucagon in exclusion of all other markers. Bihormonal cells were defined by staining positive for insulin and glucagon. Cells inside the islet border that did not stain for any of the assessed antibodies were referred to as unstained islet cells. The positivity of cells was validated for each marker separately through visual inspection of at least three islets and immune cell areas per section.

Rule-based, automated processing of digital slides

A low-resolution spatial data representation for cell segmentation data (Cell2Grid)^27,28 was combined with rule-based automated islet classification to identify the islet stages with respect to their cellular composition (Supplemental Figure S1, S2). Islet cores were defined by cell areas containing insulin and glucagon. Islet core boundaries were annotated by using computational image-morphological operations ^27,28. Areas containing the immune cell markers CD45, CD4, and CD8 were defined as islet immune environments if located within 15 μm of an islet. Automated primary islet staging was performed through filtering of the obtained image area objects (area, minimum number of cells, fraction of islet core boundary covered by islet immune environment) in accordance with the staging system (Table 1; Fig. 1). All results obtained from the Cell2Grid/rule-based automated islet identification were visually double-checked and manually restaged, if necessary, by relying on the four-eyes principle.

Table 1.

Criteria to define islet stages from 0 to 4.

Islet stage	Description of endocrine core of the islet	Description of immune environment of the islet	Rule-based automated staging criteria
0	At least one insulin-positive cell is present. Glucagon-positive cells can be present.	Less than three T-cells (CD4 + or CD8+) are found adjacent to islet or within endocrine area.	Beta cell fraction: 5 to 100% Coverage of islet circumference by T-cells: 0%
1	At least one insulin-positive cell is present. Glucagon-positive cells can be present.	More than three T-cells (CD4 + or CD8+) are found adjacent to islet.	Beta cell fraction: 5 to 100% Coverage of islet circumference by T-cells: 5% − 90%
2	At least one insulin-positive cell is present. Glucagon-positive cells can be present.	Abundant T-cells (CD4 + or CD8+) that surround the islet.	Beta cell fraction: 5 to 100% Coverage of islet circumference by T-cells: > 90%
3	Less than or equal to one insulin-positive cell. Glucagon-positive cells are the most abundant endocrine cell type.	More than three T-cells (CD4 + or CD8+) are found adjacent to islet.	Beta cell fraction: < 5% Coverage of islet circumference by T-cells: 5–100%
4	Less than or equal to one insulin-positive cell. Glucagon-positive cells are the most abundant endocrine cell type.	Less than three T-cells (CD4 + or CD8+) are found adjacent to islet.	Beta cell fraction: < 5% Coverage of islet circumference by T-cells: 0%

Fig. 1.

Islet stages displaying representative islets of Langerhans before, during and after insulitis during type 1 diabetes progression in NOD mice. (A): H&E images depicting islets of Langerhans (light purple), T-cells appear as small, dark purple cells. (B): fm-IHC staining of endocrine and immune cell epitopes: insulin (green), glucagon (red), CD4 (magenta) and CD8 (cyan) including nuclear counterstain with DAPI (blue). (C): in-situ hybridization using padlock probe technology for insulin (green) and glucagon (red) mRNAs, in islet stages 3 and 4 signals are highlighted with small red arrows.

In-situ hybridisation using padlock probe technology

From the fm-IHC images, three representative pancreatic tissue sections with around 50 islets were chosen and stained with in-situ hybridization using padlock probe technology to assess and validate the newly defined islet stages. In-situ hybridization was used to localize Ins1, Ins2, and Gcg mRNA transcripts. mRNA detection procedures were performed as described in ²⁹. For reference sequences see Supplemental table T3. Padlock probes were designed using an open-source Python software package (https://github.com/Moldia/multi_padlock_design) as described by Gyllborg et al. 2020 ³⁰. The padlock probes were validated on tissue samples of BL6 and NOD mice. Imaging was performed with a digital slide scanner (Olympus Evident slideview VS200, Tokyo, Japan), equipped with an LED source (Excelitas Technologies, X-Cite Xylis, Göttingen, Germany) and filter cubes. Visual analysis was accomplished with OlyVIA 2.4 (Olympus Soft Imaging Solutions, Tokyo, Japan).

Analysis and statistics

Data were analysed and Fig. 3 was visualized using Python 3.10.4. Statistical tests were performed using GraphPad Prism 9.5.1. In bar plots, mean values and SD are depicted unless stated otherwise. Datasets were tested for normal distribution with Shapiro-Wilk test and by inspection of histograms. Islet numbers per mm² tissue of the non-diabetic subgroup could not be confirmed as normally distributed, thus the Kruskal-Wallis test with Dunn’s multiple comparisons correction was applied. For cell-based datasets, normal distribution could not be determined for all subgroups, hence Kruskal Wallis test with Dunn’s multiple comparisons adjustment were applied. Sections were treated as technical replicates. Results were considered significant for p < 0.05. Data is available from the corresponding author upon request.

Fig. 3.

Correlation of average BG levels with the proportion of islets in each islet stage in (A): the NOD mouse cohort and (B): BL6 (left) and late-diabetic controls (right). Islets markedly shift towards islet stages 3 and 4 at an average BG around 126 mg/dl.

Results

Islet staining and definition of islet stages

Fm-IHC and in-situ hybridization staining were successfully established in NOD and BL6 mice. Islets obtained from an initial subset of eight NOD mice underwent manual categorization, leading to the identification of distinct islet stages. By integrating the level of insulitis with insulin-containing and insulin-deficient islets, the process of beta cell loss was classified into islet stages from 0 to 4 (Table 1; Fig. 1).

Subsequently, the parameters characterizing each islet stage were systematically extracted and integrated into Cell2Grid/rule-based automated islet identification which then accurately identified and characterized 82% of the islets. Classification errors were primarily attributable to staining artefacts or their classification as fringe cases within the staging system. A total of 3324 islets in all mice were identified and classified into islet stages. Islets of islet stage 0 were predominant and accounted for about 62% of total islets. Islet stage 1 accounted for around 13%, islet stage 2 for about 4%, islet stage 3 for approximately 6%, and islet stage 4 represented about 16% of all identified islets.

Islets in stage 0 were positive staining for insulin, can contain glucagon (Fig. 1B and C) and showed less than three T-cells per islet (Fig. 1, column 0). In islet stage 1, more than three T cells were observed, typically at one point on the edge of the islet core (Fig. 1, column 1). In islet stage 2 over 95% of the islet circumference was surrounded by T cells. In islet stage 2, insulin staining was lower than in islet stages 0 and 1 and at least 5% insulin-positive cells were present (Fig. 1, column 2). In islet stage 3 less than 5% of cells were insulin-positive and most cells were staining for glucagon, while the number of T cells varied (Fig. 1, column 3). In islet stage 4, T-cells were entirely absent, and the presence of insulin-positive staining was less than 5% of cells (Fig. 1, column 4). The presence of insulin and glucagon transcripts confirmed the findings from fm-IHC staining (Fig. 1C).

Cellular composition of islet stages

When characterizing the islet stages regarding their cellular composition, the percentage of beta cells was more than 50% in islet stages 0 and 1, about 40% in islet stage 2 and then decreased to less than 5% in stage 3 islets (Fig. 2A).

Fig. 2.

Cellular composition of islets in islet stage 0 to 4 in (A): NOD mice; and (B): BL6 and late-diabetic (diab) controls (Ctrl). Only the endocrine islet core was analysed, and insulitis and exocrine tissue was excluded. Data are plotted as mean of each cell type within the islet core. (n: number of islets).

In islet stage 4, beta cells were almost completely absent. The proportion of alpha cells was nearly constant (about 10% of all cells) in islet stages 0, 1, and 2. It increased to about 50% in islet stage 3, and to about 60% in islet stage 4. Bihormonal cells accounted for around 10% in all islet stages. The proportion of T-cells inside the islets was low at all islet stages except for islet stage 2 where a peak of 18% T cells of all cells was observed. The proportion of unstained islet cells increased steadily during beta cell destruction from islet stage 0 to 4. The cellular composition of islets from BL6 controls was comparable to those of stage 0 islets of NOD mice (Fig. 2B). Also, the cellular composition of islets derived from late-diabetic controls and those of islet stage 4 in NOD mice was similar.

Correlation of islet stages with average BG

Subsequently, we correlated the islet stages to the average BG of each animal. NOD mice exhibited major changes in their islet stages at an average BG between 115 and 135 mg/dl (6.4 and 7.5 mmol/l; Fig. 3A).

Below an average BG of 115 mg/dl (6.4 mmol/l), the endocrine pancreas predominantly consisted of islet stage 0, some animals also exhibiting islet stages 1 and 2. Above an average BG of 135 mg/dl the predominant islet stages were islet stages 3 and 4. Between average BG of 121 and 135 mg/dl (6.7 and 7.5 mmol/l) a transition phase was observed with a pronounced shift starting at 126 mg/dl (7.0 mmol/l) towards a loss of islets (reflected in increasing abundance of islet stages 3 and 4). Only three mice with low average BG levels of 115 mg/dl, 118 mg/dl and 121 mg/dl (6.4, 6.5 and 6.7 mmol/l) showed an increased proportion of islet stages 3 and 4 (between 20% and 50%; Fig. 3A). In BL6 controls (average BG below 148 mg/dl or 8.2 mmol/l) more than 98% of islets exhibited islet stage 0, and in late-diabetic controls (average BG: about 550 mg/dl or 30.5 mmol/l) more than 90% of the islets exhibited islet stage 3 and 4 (Fig. 3B). Sorted by the investigated age cohorts, islet stages appear very heterogeneous (Supplemental Figure S3).

Identification of disease subgroups in NOD mice

To accurately characterize the level of beta cell destruction in each animal, we introduced the Islet Score which is reflecting the average islet stage per mouse. For this purpose, the numerical values of the individual islet stage were summed up and divided by the total number of islets per animal. An Islet Score below 1.0 indicated that less than 50% of islets are stages 0 and 1 islets, whereas at an Islet Score above 1.0 more than 50% of the islets belonged to islet stages 2, 3 and 4. Correlation of average BG levels with the Islet Score allowed the identification of three disease subgroups in NOD mice with BG levels under 200 mg/dl, namely the non-diabetic (Islet Score < 1.0, BG < 126 mg/dl or 7.0 mmol/l), pre-diabetic (Islet Score > 1.0, BG < 126 mg/dl) and early-diabetic subgroup (Islet Score > 1.0, BG > 126 mg/dl; Fig. 4).

Fig. 4.

Identification of three different subgroups based on the average BG level and the Islet Score in the NOD mouse cohort. Lower left quadrant: non-diabetic subgroup; Upper left quadrant: pre-diabetic subgroup; Upper right quadrant: early-diabetic subgroup. Each point represents one NOD mouse. Data points follow a sigmoid curve.

Animals of the non-diabetic subgroup contained predominantly islet stage 0 and 1 and their average BG levels were low (under 126 mg/dl). The three NOD mice belonging to the pre-diabetic subgroup showed an increased amount of islet stage 2, 3 and 4 (Islet Score above 1.0) despite an average BG level below 126 mg/dl (Figs. 3A and 4), indicating a progressed disease state compared to NOD mice in the non-diabetic subgroup. The animals in the early-diabetic subgroup exhibited mostly stage 3 and 4 islets and had higher BG levels (above 126 mg/dl).

Number of Islets per mm² of tissue

In the next step, we analysed the number of islets per mm² of tissue for each section in the NOD mouse cohort. The median number of islets per mm² significantly decreased from the non-diabetic to the early-diabetic subgroup (Fig. 5).

Fig. 5.

Violin plot of islet number per mm² tissue of all sections. The subgroups are following type 1 diabetes progression, represented by the three identified subgroups in our NOD mouse cohort as well as BL6 and late-diabetic controls. Each dot represents one section. Dotted lines represent quartiles and dashed lines the median (* p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001).

The non-diabetic subgroup showed the broadest range in their number of islets, with a maximum of 1.83 and a minimum of 0.11 islets/mm² and a median islet number of 0.65 per mm². Intriguingly, the highest cumulative islet numbers in the non-diabetic animals, depicted in Supplemental Figure S4, all belong to the youngest animals where we could obtain BG values (age: 10 weeks). Pre-diabetic NOD mice exhibited a median 0.33 islets/mm². The median islets/mm² in pre-, early- and late-diabetic controls were not significantly different from each other (0.65, 0.33, 0.28). The differences in islets/mm² were statistically significant when the BL6 or the non-diabetic NOD subgroup was included (Fig. 5).

Composition of Islets in the non-, pre-, early and late-diabetic subgroups

The mean number of beta cells per µm² endocrine area (islet core) continuously and significantly decreased from 4899 in the non-diabetic subgroup, to 2422 in the pre-diabetic subgroup, to 836.8 in the early-diabetic subgroup, and to 85.2 in the late-diabetic subgroup (Fig. 6).

Fig. 6.

Composition of islets in the non-, pre-, early and in late-diabetic subgroups in NOD mice and BL6 controls. Bars represent the mean and SD per µm² of islet core (IC) for beta cells (green), alpha cells (red), bihormonal cells (orange), and unstained islet cells (grey) in each subgroup and each control group. (* p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001).

The mean number of alpha cells per µm² increased continuously from the non- to the early-diabetic subgroup and was more than four times higher in early- and late-diabetic subgroups (4719 and 5806 cells/µm²) than in non-diabetic subgroup and BL6 controls (1190 and 930.7 cells/µm2). A similar, but less pronounced trend is observable in endocrine cells not stained in our panel (e.g. delta, gamma, epsilon, referred to as other cells). The mean number of other cells per µm² increased from 2176 in the non-diabetic to 3908 cells/µm² in the pre-diabetic subgroup and remained high in the early- and late-diabetic subgroup (3701 and 4397 cells/µm² respectively). Bihormonal cells per µm² were highest in the pre-diabetic subgroup (1653 cells/µm²) and lowest in the late-diabetic subgroup (151.6 cells/µm²).

BL6 controls exhibited similar values in all cell types to the non-diabetic subgroup, and no difference between these groups was statistically significant. Late-diabetic controls contained the fewest insulin-staining cells (85.2 beta cells/µm² and 151.6 bihormonal cells/µm²), which is consistent with the insulin-deficient islet stages shown in Fig. 3B. The mean cellular composition of islets irrespective of endocrine area follows similar trends and is depicted in Supplemental Figure S5.

Discussion

By revisiting type 1 diabetes progression in the NOD mouse, a classification system for islets describing type 1 diabetes progression and beta cell loss in NOD mice was developed (Fig. 7).

Fig. 7.

Model of type 1 diabetes progression in NOD mice. Schematic progression of type 1 diabetes, as described by the percentage of functional islets (Islet Score) and the average BG (126 mg/dl) in NOD mice. Late diabetic animals were characterized by the classical convention of diabetes diagnosis in NOD mice. Figure was created in Biorender.com.

We identified three subgroups of type 1 diabetes progression using the Islet Score and a BG threshold of 126 mg/dl.

We implemented fm-IHC staining to analyse the changes in islets of NOD mice. This method allowed for the first time the distinction of insulin-containing islets from insulin-deficient islets in parallel with insulitis assessment and it allowed detection of low-expressed proteins due to signal amplification and staining of up to six markers in the same tissue section ^23,31,32. As opposed to H&E staining, with fm-HC staining it was possible to discern islet stages 1 and 2 from islet stage 3, and islet stage 0 from islet stage 4 islets. We were able to use consecutive sections for fm-IHC and for in-situ hybridization and therefore we could topically assign staining of insulin and glucagon peptides in endocrine cells to their respective mRNA transcripts. We analysed 3324 islets and findings are in good agreement with those observed in human pancreatic tissue ²³.

In several mice with BG levels below 115 mg/dl, we found 20 to 40% of islets classified into islet stage 1 or above, indicating that insulitis and the consecutive loss of beta cells occurred far below BG levels of 200 mg/dl, in line with results from a previously performed study in NOD mice ¹⁸. We observed a highly dynamic process of insulitis on the one hand and beta cell loss on the other hand in several animals with BG levels between 115 and 135 mg/dl, with a turning point around 126 mg/dl. Of note, the conventionally used threshold for diabetes diagnosis in NOD mice (i.e., 200 mg/dl or above) misses this dynamic process. The BG threshold of 126 mg/dl is also the threshold currently used for human type 1 diabetes diagnosis ¹³.

The transition from insulin-containing to insulin-deficient islets is underrepresented in age-stratified NOD mouse cohorts with BG levels under 200 mg/dl, and these age groups can be very heterogeneous in their states of beta cell loss (Supplemental Figure S3). We suggest future studies use BG criteria rather than age to reduce heterogeneity when investigating non-diabetic NOD mice.

In BL6 controls 98% of islets were in islet stage 0, in line with the lack of autoimmunity against beta cells in these animals. Interestingly, BG levels in this model were 140 and 148 mg/dl (7.8 and 8.2 mmol/l) which corresponds to results from Amrani et al. who reported that NOD mice exhibit lower glycaemia than BL6 mice ¹⁷. In NOD mice with a BG above 500 mg/dl almost all islets exhibited islet stages 3 and 4. These islet stages were also dominant in NOD mice with BG levels above 140 mg/dl, underlining the importance of the suggested BG threshold around 126 mg/dl.

The novel classification of insulitis and beta cell loss in whole animals proposed here reflects the disease stages suggested for type 1 diabetes in humans ^33–35. Animals with blood glucose levels above 126 mg/dl were associated with islet scores of 2.5 and higher, which reflects the decreasing insulin content in islets and these animals were categorised as early-diabetic. This contrasts with BG values used in conventional diabetes diagnosis in NOD mice. Interestingly, pre-diabetic animals’ BG levels cluster close to the threshold of 126 mg/dl, in contrast to the animals assigned to the non-diabetic group.

Two observations in our study indicated a fast progression from the non-diabetic to the diabetic state in NOD mice: First, the low number of animals observed in the pre-diabetic group, despite a large range of age groups studied and second the rare incidence of islets belonging to islet stages 1 and 2 that was leading to an overall sigmoid curve of disease progression. A similar phenomenon, the so-called immune flare around the clinical disease manifestation of type 1 diabetes in humans has been described ³⁴.

We observed a constant decline in islet numbers with disease progression. It is currently not clear if NOD mice have a reduced number of islets in general or if the loss of islets starts at a very early age. The animals in our age-stratified cohort of 10 weeks showed islet numbers close to those found in BL6 controls.

In-depth analysis of islets revealed that, in parallel to the expected loss of beta cells, numbers of glucagon-positive cells were four-fold in stage 4 islets and unstained islet cells doubled their number (Supplemental Figure S6). These findings could indicate that in contrast to the common understanding, islets after onset of diabetes do not solely consist of the residual alpha cells, but the lost beta cell mass seems to be replaced by cells staining for glucagon or by unstained islet cells. The observed increase in alpha and other endocrine cell numbers is substantiated by systematic quantification using absolute cell counts, appropriate controls, and statistical analysis. The possibility that these findings are due to changes in cell proportions or islet size alone is effectively ruled out by the study’s methodology. Bihormonal cells appeared consistently in islets, except in the late-diabetic disease stage, where the amount was markedly lowered. Currently, it is unclear if alpha cells and unstained islet cells are proliferating or, even more intriguingly, if beta cells de- or transdifferentiate during insulitis. This process could be triggered by some unknown molecular signal possibly produced by beta cells during their destruction or maybe even by the immune infiltrate via interferons ³⁶. This leads to the speculation that the fate of endocrine cells is considerably more dynamic and susceptible to influence than previously anticipated, which may be of relevance for future studies. A comprehensive characterization of this phenomenon is currently ongoing.

One limitation of this study is the lack of assays directly assessing the function and proliferation of endocrine cells. Another limitation of the study is the small number of animals, particularly those under 30 weeks of age, and BL6 and late-diabetic controls. However, islet composition of both control groups was very homogeneous and consistent with previous results. The study design was the result of careful planning, and we kept the number of animals small in consideration of the animal welfare act (Federal Act on Animal Welfare, 2004/118). While these preliminary findings provide valuable insight into the pathophysiology of type 1 diabetes, we recommend interpreting results from groups with low animal numbers (BL6, late-diabetic and pre-diabetic subgroup) cautiously and emphasize the need for validation in larger cohorts. Despite this, we were able to show valuable trends and also significant differences in this study. Comprehensive characterisation of the described phenomena is ongoing and will require further studies.

The strengths of this study lie in the power of here established fm-IHC staining method which made it possible to differentiate between insulin-containing and insulin-deficient islets parallel to insulitis assessment. In addition, this method allowed a differentiation between islet stages 1,2 and 3 as well as between islet stage 0 and 4. Another strength of the here presented study is the enormous number of islets analysed.

In conclusion, we proposed a classification system for islets describing type 1 diabetes progression in NOD mice. We identified three subgroups of disease progression using the Islet Score and the BG threshold of 126 mg/dl. The revised classification of type 1 diabetes in NOD mice and the resulting model of type 1 diabetes progression in NOD mice facilitates the alignment of NOD mouse model with the disease progression described in humans. By reducing heterogeneity in the progression of type 1 diabetes in NOD mouse cohorts, these results have great potential to advance the investigation of interventions in the early stages of the disease.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Conceptualization: BE, BB, CK, A E-H, VH, PK, BP, TRP, Methodology: BE, LH, CK, LB, KS, Software: BE, LH, LB, KS, Validation: BE, LH, CK, KB, JF, VH, CH, Formal analysis: LH, BE, Investigation: BE, CK, KB, JF, LB, KS, VH, CH, Data interpretation and selection: BE, BB, BP, TRP, Resources: BO, CK, AE-H CH, BP, TRP, Data curation: BE, Writing – original draft: BE, BB, TRP. All authors reviewed the manuscript and approved it for publication.

Funding

This project was supported by a research grant from by BioTechMed Graz (Flagship Project 2020). The study sponsor/funder was not involved in the design of the study; the collection, analysis, and interpretation of data; writing the report; and did not impose any restrictions regarding the publication of the report.

Data availability

Data is available from the corresponding author upon request.

Declarations

Competing interests

The authors declare no competing interests.