Differential Insulitic Profiles Determine the Extent of β-Cell Destruction and the Age at Onset of Type 1 Diabetes

Type 1 diabetes (T1D) is widely perceived as an autoimmune condition arising from the selective destruction of pancreatic β-cells by influent immune cells targeted to the islets of Langerhans (1–4). This concept has been examined extensively in animal models, but many fewer studies have been undertaken in humans. Indeed, In’t Veld (5,6) has noted that as few as 150 T1D case patients with insulitis have been scrutinized worldwide since the beginning of the last century. This is unfortunate because the extent of insulitis observed in animal models is more fulminant than in most human patients, implying that additional studies are required to improve our understanding of the process in humans (5). Accordingly, we have exploited a large collection of autopsy pancreas samples, recovered from patients who died soon after the diagnosis of T1D, to investigate the immune infiltrate in inflamed islets (2,7,8). This has revealed aspects of the temporal changes that accompany β-cell destruction and has shown that the principal cells involved in the destructive process are likely to be CD8⁺ cytotoxic T cells (2,7,8). However, the studies also yielded a more unexpected outcome in that they revealed a potential role for CD20⁺ B cells in some patients based on the finding that these cells infiltrate islets in large numbers, in parallel with the influx of CD8⁺ T cells (7).

Building on this background, we have also highlighted an unexpected dichotomy in the composition of the immune cell infiltrates present in the inflamed islets of patients with recent-onset T1D (8). Thus, it was found that, although significant numbers of CD20⁺ B cells infiltrate the islets in many individuals, this is not universally true. Rather, other patients could be found in whose inflamed islets very few CD20⁺ cells were present. To denote this difference, the terms CD20Hi and CD20Lo were coined to differentiate the two profiles of insulitis, and it was shown that each was present among a group of recent-onset patients in whom T1D had been diagnosed before the age of 20 years (8). Importantly, it was also noted that the variation in B-cell numbers occurs independently of the presence of CD8⁺ T cells. Despite this, it was also clear that the total population of infiltrating immune cells is least abundant in cases where CD20⁺ cell numbers are depleted. Accordingly, the two states might also be considered as relatively hyperimmune (CD20Hi) or pauci-immune (CD20Lo), respectively.

In order to develop a greater understanding of the functional significance of these two insulitic profiles in human T1D, we now report a more complete characterization of the phenomenon in the following three different pancreas collections: those harvested postmortem (U.K.) (9), biopsy samples from living donors (Diabetes Virus Detection [DiViD] study; Norway) (10), and samples from organ donors (Network for Pancreatic Organ Donors with Diabetes [nPOD]; U.S.) (11,12). The studies confirm the existence of two distinct profiles of insulitis and imply that these patterns exert a profound influence on the rate of disease progression and the extent of β-cell destruction in humans. As such, these new results suggest very strongly that correct patient stratification will be essential when testing the efficacy of targeted immunotherapeutic agents in human populations and imply that caution should be exercised when interpreting the outcomes of earlier trials where this was not achieved.

Insulitis was examined in 20 human pancreas sections recovered at the time of postmortem examination from patients selected from among a U.K. cohort of patients with T1D who had died within 3 months of diagnosis (7,9). These comprised 14 females and 6 males with ages ranging from 1 to 19 years (mean age 10.5 ± 1.3 years). For studies of the extent of residual insulin positivity, sections from all 151 patients within this collection were evaluated, irrespective of disease duration or the presence of insulitis (63 males; 88 females).

In addition, insulitis was also studied in pancreas sections recovered by laparoscopic minimal tail resection from six living adult donors who had participated in the DiViD study (10), as well as in two organ donors with recent-onset T1D from the nPOD cohort (Supplementary Table 1). Participants in the DiViD study were between 25 and 35 years of age, and their biopsy specimens were obtained between 3 and 9 weeks after initial diagnosis. The two nPOD donors were 4.75 and 11 years old at the time of death, and their disease durations were 120 and 0 days, respectively. The collection of pancreatic tissue in the DiViD study was approved by The Norwegian Government’s Regional Ethics Committee, and all samples were studied with ethical approval.

Serial sections (4 µm) were cut from each pancreas and mounted on glass slides. Samples from the U.K. cohort were processed and labeled using a standard immunoperoxidase technique for paraffin sections, as described previously (7,8). All antigens, except insulin and glucagon, were unmasked by heat-induced epitope retrieval. In brief, sections to be labeled with anti-CD20⁺ antiserum (M0755; Dako) were heated in 10 mmol/L citrate buffer, pH 6, for 20 min in a microwave oven at full power (800 W). Those to be labeled with anti-CD4⁺ (ab846; Abcam) or anti-CD8⁺ (ab4055; Abcam) were heated in 10 mmol/L Tris, 1 mmol/L EDTA, pH 9. Primary antibodies were applied for 1 h at room temperature, with the exception of anti-CD4, which was incubated at 4°C overnight (Supplementary Table 2). Antigen visualization was achieved via the Dako REAL Envision Detection System. Some slides were processed in the absence of primary antibody or with isotype control antisera to confirm the specificity of labeling.

Pancreas sections from the DiViD and nPOD collections were examined by a combination of immunofluorescence and immunohistochemistry using horseradish peroxidase and alkaline phosphatase staining, as appropriate. Where possible, and to minimize sample usage, at least two antigens per section were examined in parallel. Antigens were unmasked by heat-induced epitope retrieval (20 min at full power in a microwave oven; 800 W) in 10 mmol/L citrate buffer (pH 6). Combined staining was achieved by using guinea pig anti-insulin (A0564; Dako) in combination with mouse anti-glucagon (ab10988; Abcam) and visualized using a Vector Red Substrate kit for detection of alkaline phosphatase activity (AK-5100 and SK-5100; Vector Laboratories, Peterborough, U.K.). Binding of rabbit anti-CD8 and mouse anti-CD20 sera was visualized using secondary antibodies tagged with appropriate fluorescent probes (Alexa Fluor anti-mouse 488 [A11029; Invitrogen, Paisley, U.K.] or anti-rabbit 568 [A1136; Invitrogen]). Cell nuclei were stained with DAPI.

Bright-field image acquisition and analysis were performed on a Nikon 50i Microscope fitted with a DS-Fi camera and a DSL2 camera control unit. Images were captured and analyzed using the ImageJ platform. Immunofluorescence image collection and processing were achieved using a Leica DM4000 B LED upright fluorescence microscope and Leica Image analysis software (LAS AF).

Ten islets were selected randomly from within each stained section among the U.K. cohort, and the number of CD8⁺, CD4⁺, and CD20⁺ cells within the islet area and in the endocrine-exocrine interface in close contact with the islet boundary were recorded. The mean number ±SEM of each immune cell subtype was calculated for each patient as an average of the 10 individual values from each islet. The insulin-positive area of each islet was assessed and expressed as a percentage of the total area, as described previously (7), and was scored as +++, ++, or +, respectively. In the DiViD and nPOD cohorts, immune cells were counted from ∼30 islets per patient section and were expressed as described above. Statistical significance was calculated using the Kruskal-Wallis with Dunn multiple-comparison test, Student t test, or χ² analysis, as appropriate.

Detailed analysis was undertaken of the numbers of CD8⁺, CD4⁺, and CD20⁺ cells within, or in close proximity to, 10 randomly selected inflamed islets across multiple case patients from among the U.K. collection of recent-onset T1D patients. This confirmed the existence of the two independent profiles (CD20Hi and CD20Lo) (Fig. 1A and B). The results also revealed that these profiles are patient specific, in that all inflamed islets within any given individual display the same phenotype.

A more dynamic evaluation of the profile of immune cell infiltration in relation to β-cell loss was also undertaken in patients from each group (Fig. 1C–F). This analysis confirmed that CD8⁺ T cells always comprise the majority population within inflamed islets, irrespective of the number of CD20⁺ cells present. This was evident both when the absolute number of each cell type was examined (Fig. 1C and D) and when the relative proportions of each were calculated (Fig. 1E and F). It was also clear that the number of CD8⁺ T cells varied according to the progression of β-cell loss (Fig. 1C and D). CD4⁺ T-cell numbers also tended to rise as the extent of β-cell loss neared totality, but their absolute number (and relative proportion) was always much lower than the number (or proportion) of CD8⁺ cells. Importantly, the most marked difference between the two phenotypes was in the proportion of CD20⁺ B cells found within, or near to, inflamed islets. Among one group (CD20Hi), these were abundant, and they increased in parallel with the CD8⁺ cells as β-cell death progressed (Fig. 1C and E). By contrast, in the second group (CD20Lo individuals) they were the least abundant cell type at all stages of the disease process (Fig. 1D and F).

In addition to the variation in the absolute number of CD20⁺ cells, the ratio of CD20⁺ to CD4⁺ cells also varied consistently between the two phenotypes (Fig. 1). Thus, in CD20Hi case patients, the ratio of CD20⁺ to CD4⁺ cells in inflamed islets was >1.0 at all stages of β-cell loss, whereas this ratio was consistently <1.0 in CD20Lo patients. On this basis, it seemed that estimation of the CD20⁺/CD4⁺ cell ratio might offer greater utility as a means to subclassify patients. Accordingly, we focused on this ratio to define the CD20Hi (CD20⁺/CD4⁺ ratio >1.0) and CD20Lo (CD20⁺/CD4⁺ ratio <1.0) subgroups in subsequent studies.

Because the two profiles of insulitis may represent differentially aggressive phenotypes with variable rates of disease progression (8), the ratio of CD20⁺ to CD4⁺ cells was examined in patients from the U.K. at increasing age at diagnosis. This revealed a striking relationship despite the limitation that all patients studied had received a diagnosis at <20 years of age (Fig. 2). Notably, it was clear that CD20Hi patients tended to receive a diagnosis at younger ages than those defined as CD20Lo. Indeed, when patients within the U.K. cohort were stratified into those in whom T1D was diagnosed at between 1 and 6 years of age versus those in whom it was diagnosed at ≥13 years of age, the separation between the two phenotypes was absolute (Fig. 2). All individuals within the younger group were classified as CD20Hi, whereas those who received a diagnosis at the older age were CD20Lo. Patients in whom T1D developed between the ages of 7 and 13 years fell into either group.

To verify these conclusions, we next examined the profile of insulitis in pancreas samples from among the nPOD collection and from the DiViD cohort of living organ donors. In practice, only two patients who met our criteria for recent-onset T1D (within 120 days of diagnosis) were available from the nPOD cohort (case nos. 6209 and 6228), whereas all six of the DiViD patients conformed to this criterion. Strikingly, the insulitic profile in each of the inflamed islets of the six DiViD patients (all of whom had received a diagnosis beyond the age of 20 years) was CD20Lo (Fig. 3). A similar profile was found in one of the two nPOD patients (nPOD case 6228; T1D diagnosed originally at 13 years of age). By contrast, the second nPOD case patient (nPOD case 6209) had received a diagnosis much earlier in life (<5 years of age) and was defined as CD20Hi.

In view of the finding that age at diagnosis is predictive of the profile of insulitis in patients with recent-onset T1D, and that individuals who have received a diagnosis at younger ages appear to display a more aggressive phenotype, we considered the possibility that the proportion of residual insulin-containing islets (ICIs) might also differ between these groups. Accordingly, we calculated the proportion of such islets in sections of pancreas from each patient. This was achieved by using glucagon immunoreactivity to define the total islet number, with simultaneous costaining for insulin to monitor the proportion of residual ICIs. Initially, this analysis was performed for the patients within the U.K. cohort, and these were stratified according to age, as follows: group 1, from 1 to 6 years of age at diagnosis; group 2, from 7 to 12 years of age at diagnosis; and group 3, ≥13 years of age at diagnosis. The results revealed a strongly positive relationship in which individuals who had received a diagnosis at the youngest ages retained a much lower proportion of ICIs than those who had received a diagnosis in the older age group (P < 0.001) (Fig. 4).

Extending this analysis to include pancreas samples from the two additional nPOD case patients, as well as those from the six DiViD patients, reinforced these conclusions. Thus, the DiViD patients retained a mean of 25% ICIs in the pancreas sections studied at diagnosis (Fig. 4), and this was also the situation in nPOD case patient 6228 (CD20Lo) (Fig. 4). By contrast, nPOD case patient 6209, whose insulitic profile was consistent with a more aggressive inflammatory infiltrate, retained only 10% ICIs at diagnosis.

Having examined patients soon after diagnosis, the analysis was extended to include U.K. case patients having a disease duration of between 1 and 10 years (Supplementary Fig. 1). Consistent with the data from recent-onset patients, individuals who had received a diagnosis at the younger ages retained the fewest ICIs.

In view of this striking outcome, we extended the data set further to examine the entire collection of T1D pancreas samples available within the U.K. cohort (a total of 151 patients), irrespective of their disease duration or the presence of insulitis (Fig. 5). The proportion of residual ICIs present in at least one pancreatic section (selected randomly from those available for study) from each block was calculated. When stratified into three groups according to the original age at diagnosis (1–6 years, 7–12 years, and >13 years of age), the proportion of residual ICIs increased from a mean of 14 ± 3% in those who received a diagnosis early in life to 39 ± 4% in those who received a diagnosis from their teens or beyond (P < 0.001). Indeed, among the latter group, 56 of 69 patients still retained residual ICIs in the pancreas sections studied. This was significantly more than the equivalent proportion among those in the youngest age group (19 of 43 patients; P < 0.004 by χ² analysis). Similar findings were made in the nPOD cohort when the proportion of case patients with residual ICIs was examined (Supplementary Fig. 2 and Supplementary Table 2).

It is often assumed that insulitis and β-cell loss proceed in a broadly consistent manner in the majority of patients in whom T1D develops and that the final descent into clinical disease occurs when a defined proportion of the initial β-cell mass has been lost. Estimates vary as to the precise extent, but it is frequently argued that a >80% loss may be required to precipitate the disease (13–16). However, in accord with Klinke (17), we find that this may be an oversimplification. Importantly, we note that β-cell numbers do not necessarily dwindle to very low levels before diabetes is manifest but that many patients have diabetes while still retaining significant numbers of these cells. We also find that the process and extent of β-cell loss varies with age such that individuals in whom T1D develops early in life (before the age of 7 years) lose β-cells to a much greater extent than those who receive a diagnosis of the illness in their teenage years or beyond. Similar conclusions that patients with an early disease onset may have a more fulminant form of the illness have also been reached previously (18,19), and we now attribute this variation to the existence of two distinct profiles of insulitis, which yield very different rates of β-cell loss (8).

Careful examination of multiple islets across a range of patients reveals that all of the inflamed islets within a given patient display the same insulitic profile but that this profile differs significantly between individuals. Importantly, this variation does not occur in a purely stochastic manner; rather, two identifiable forms of insulitis can be differentiated. Both are characterized by a preponderance of CD8⁺ T cells, suggesting that these are likely to be the key cytotoxic mediators in each subtype of the disease. By contrast, there is much greater variation in the proportion of CD20⁺ B cells; these are abundant in the inflamed islets of some individuals but are almost absent from the islets of others. As a result, the terms CD20Hi or CD20Lo were coined to denote these two profiles (8). This designation is intended not to imply that CD20⁺ cells are necessarily (or selectively) cytotoxic in one form of the disease but, rather, to emphasize that these cells may play an important accessory role in the destructive process and that this is exerted more effectively in one form than in the other. Comparison of the number (or proportion) of CD4⁺ cells within inflamed islets revealed no significant differences among this cell population but suggested that they could provide a useful means to distinguish the two profiles, based on measurement of the ratio of CD20 to CD4 cells. Thus, among CD20Hi patients this ratio was >1.0, whereas it was <1.0 in those defined as CD20Lo. Accordingly, we propose that monitoring of the ratio of CD20 to CD4 cells may provide a useful and robust means to distinguish the two subtypes of insulitis.

The justification for our conclusion that the two profiles of insulitis are functionally different lies in the striking finding that children in whom diabetes develops at <7 years of age uniformly display a CD20Hi profile, whereas the islets of those patients in whom diabetes is diagnosed beyond their teenage years always display a CD20Lo profile. This extraordinary level of differentiation was unexpected, and it was important to establish that it did not reflect a hidden variable arising from the study of individuals whose pancreata were harvested in a particular way or from one specific geographical region (the majority of our observations having been made using samples collected at autopsy in the U.K. from individuals who had died in ketoacidosis) (9). In noting this, it must also be emphasized that extending the analysis is not a trivial task because the number of case patients with recent-onset T1D available for study worldwide is very small. Nevertheless, we were able to confirm the findings by analysis of two organ donors with recent-onset T1D from the nPOD collection (patients 6209 and 6228) (11,12) and six adults with new diagnoses who had consented to undergo a pancreatic biopsy in the Norwegian DiViD study (10). These followed patterns identical to those seen in the larger group from the U.K. and allowed the compilation of a more extensive data set to verify the conclusions. This revealed statistically significant differences in several parameters assessed at the islet level between those who had received a diagnosis at younger ages (<7 years of age) and those who were older at disease onset (≥13 years of age). Most notably, all patients who received a diagnosis at >13 years of age displayed a CD20Lo islet profile, whereas young children had a CD20Hi profile. Not surprisingly, individuals who received a diagnosis of T1D between the ages of 7 and 12 years could fall into either category, as might be expected if two overlapping distribution profiles exist.

Analysis of the age-related variation in disease pathology revealed a second important feature, namely, that the proportion of ICIs remaining at the diagnosis of T1D varies significantly between groups. Consistent with the results of a previous meta-analysis (17), we found that those who received a diagnosis in their early years had a much lower proportion of ICIs than patients with an older age at onset. This difference was so pronounced that age at diagnosis could be used as a strong surrogate to predict the proportion of ICIs present at diagnosis and the profile of insulitis that would be found (whether CD20Hi or CD20Lo). Moreover, when this analysis was extended to include all 151 samples from within the U.K. cohort and 116 samples from nPOD (irrespective of disease duration) stratified according to age at diagnosis, these relationships still held. Younger patients (<7 years of age) had a significantly lower proportion of residual ICIs than those who presented with T1D in their teenage years or beyond (even when many years had passed between the initial diagnosis and the availability of the pancreas for analysis).

In drawing this conclusion, it is important to recognize that the present analysis was undertaken on only a small number of sections from any given pancreas. Indeed, in many cases the proportion of ICIs was estimated on a single section. This then raises an important question about the extent to which an individual set of data derived from one section can be considered representative of the entire pancreas. This is especially true given that the processes of insulitis and β-cell loss occur in a lobular manner across the gland (20–23). Such considerations inevitably mean that it would be dangerous to draw firm conclusions from a single pancreatic section taken from one individual patient studied in isolation. However, a key strength of the current study is that we were able to examine sections selected at random from multiple patients and, thereby, to counter the likelihood of systematic bias. On this basis, we consider that the cumulative results provide a representative view of the pathological mechanisms causing T1D and that wider generalization of the conclusions is valid. If this were not the case, then factors such as the proportion of ICIs would be expected to be distributed more randomly in the analysis rather than segregating in parallel with age at disease onset, as we find.

The discovery that T1D develops in many patients at a time when their β-cell mass remains relatively high suggests that, in these individuals, β-cell dysfunction rather than cell death may be a significant contributory factor leading to glucose intolerance. Indeed, this is supported directly by study of the insulin secretory capacity of the islets isolated from pancreas biopsy samples recovered from the DiViD cohort (24). These patients all displayed a CD20Lo profile of insulitis, and, in accord with this, they received a diagnosis beyond their teenage years and retained a large proportion of ICIs at disease onset. Importantly, their islets were largely unresponsive to a glucose stimulus upon isolation but gradually regained responsiveness during tissue culture under ex vivo conditions (24). This implies that the ICIs had become dysfunctional in vivo rather than experiencing an irreversible decline arising from complete β-cell demise. These results therefore offer the prospect that, by appropriate therapeutic intervention, the residual β-cells of patients with a CD20Lo insulitic profile might also be reactivated in vivo and that, in combination with an appropriate immunotherapeutic regimen, this could lead to a reduced requirement for exogenous insulin.

One important conclusion arising from the present data is that due care should be exercised when designing immunotherapeutic interventions in patients with T1D. In particular, it will be important to ensure that patients are stratified according to their insulitic profile (for which age at onset appears a strong surrogate) and that the regimen is tailored appropriately. One obvious example to illustrate this principle arises from the outcome of trials with the anti-CD20 monoclonal antibody rituximab (25,26). Because we find that infiltrating CD20⁺ cells are present preferentially in the islets of patients who received a diagnosis at <7 years of age, it is, perhaps, likely that such individuals may be most sensitive to the therapy, whereas older patients will be more refractory.

In conclusion, we confirm that insulitis occurs in two distinct profiles in patients with T1D. These differ principally in the proportion of CD20⁺ B cells (expressed relative to CD4⁺ cells) present within the infiltrate, and this appears to lead to a more or less aggressive phenotype, respectively. These phenotypes not only lead to differential rates of β-cell loss (as judged by marked differences in the age at onset between the groups) but also culminate in the loss of glucose tolerance when dramatically different numbers of β-cells remain in the pancreas. In the youngest patients, β-cell loss is rapid and extensive. By contrast, those receiving a diagnosis in their teenage years or beyond have less aggressive disease, and they experience clinical onset when a significant β-cell reserve is still measurable. These results emphasize that, in the future, it will be critical to stratify patients appropriately according to their insulitic profile and age at onset of disease in order to design and conduct trials of effective therapies for T1D.

Acknowledgments. The authors thank specialist nurse Trine Roald, Oslo University Hospital, Norway, whose invaluable efforts were essential to the success of the DiViD study.

Funding. This research received financial support from the European Union Seventh Framework Programme PEVNET (Persistent Virus Infection as a Cause of Pathogenic Inflammation in Type 1 Diabetes – An Innovative Research Program of Biobanks and Expertise) (FP7/2007-2013) under grant agreement number 261441. The participants of the PEVNET consortium are described at http://www.uta.fi/med/pevnet/publications.html. K.D.-J. has received grants from the South-Eastern Norway Regional Health Authority and the Novo Nordisk Foundation. Additional support came from a JDRF Career Development Award (5-CDA-2014-221-A-N) to S.J.R. The research was also performed with the support of the Network for Pancreatic Organ Donors with Diabetes (nPOD), a collaborative type 1 diabetes research project sponsored by the JDRF (25-2013-268), and with JDRF research grants awarded to the nPOD-V consortium (JDRF 25-2012-516 and JDRF 25-2012-770). Organ Procurement Organizations (OPO) partnering with nPOD to provide research resources are listed at www.jdrfnpod.org/our-partners.php.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. P.L. and A.W. performed data collection and analysis and reviewed the manuscript. L.K., K.D.-J., and A.K.F. collected patient material and revised the manuscript. S.J.R. and N.G.M. designed the study, performed data analysis, and wrote the manuscript. S.J.R. and N.G.M. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.