Assessment of pancreas cell type composition is crucial to the understanding of the genesis of diabetes. Current approaches use immunodetection of protein markers, for example, insulin as a marker of β-cells. A major limitation of these methods is that protein content varies in physiological and pathological conditions, complicating the extrapolation to actual cell number. Here, we demonstrate the use of cell type–specific DNA methylation markers for determining the fraction of specific cell types in human islet and pancreas specimens. We identified genomic loci that are uniquely demethylated in specific pancreatic cell types and applied targeted PCR to assess the methylation status of these loci in tissue samples, enabling inference of cell type composition. In islet preparations, normalization of insulin secretion to β-cell DNA revealed similar β-cell function in pre–type 1 diabetes (T1D), T1D, and type 2 diabetes (T2D), which was significantly lower than in donors without diabetes. In histological pancreas specimens from recent-onset T1D, this assay showed β-cell fraction within the normal range, suggesting a significant contribution of β-cell dysfunction. In T2D pancreata, we observed increased α-cell fraction and normal β-cell fraction. Methylation-based analysis provides an accurate molecular alternative to immune detection of cell types in the human pancreas, with utility in the interpretation of insulin secretion assays and the assessment of pancreas cell composition in health and disease.

Article Highlights
  • Quantification of pancreas cell type composition is based on protein biomarkers, which are prone to artifacts due to altered protein content per cell.

  • We hypothesized that cell type–specific DNA methylation patterns allow for more accurate measurement of pancreas cell composition.

  • We developed methylation markers for human pancreas cell types and measured them in human islet and pancreas specimens.

  • Methylation-based analysis allows quantification of cell type composition in the human pancreas, with implications for understanding tissue dynamics and function in diabetes.

Cell type quantification in tissue samples is an important research tool. In the pancreas, such quantification is critical for the understanding of normal tissue dynamics and perturbations in pathologies such as diabetes, hyperinsulinism, pancreatitis, and pancreatic cancer. One striking example for the importance of defining tissue composition is in the study of recent-onset type 1 diabetes (T1D). Early work suggested that the near-total loss of detectable serum insulin or C-peptide levels at disease diagnosis is due to near-total elimination of β-cells. However, histological analyses revealed that at diagnosis, nearly one-half the normal mass of β-cells is still present (1,2) and that even adult patients with medium duration T1D retain significant β-cell mass (3). This indicates that surviving β-cells are dysfunctional, providing a new understanding for disease pathogenesis (4). The potential involvement of altered β-cell mass was also examined in type 2 diabetes (T2D), where studies by Butler et al (5) and more recent multiplexed imaging mass cytometry (6) revealed reduced β-cell mass in patients. As another example, glucagon-secreting α-cells are not thought to play a direct causal role in the pathogenesis of diabetes, but alterations in α-cell mass or function have been observed and may provide insight into islet endocrine cell dynamics in patients with diabetes (7,8). In the context of pancreatic cancer, the abundance of stromal cells, often composing the majority of cells within tumors, is an important characteristic that impacts tumor biology (9).

Quantifying pancreatic cell type is important for studying the relative contribution of cell failure and death in the development of diabetes. Immunostaining for islet hormones is currently the gold standard, but it has several drawbacks. The process can show variable results, depending on technical issues due to tissue fixation and storage conditions, and this can impact accuracy and reproducibility of measurements. Additionally, metabolic and cytokine stress may cause β-cell degranulation and insulin loss, potentially skewing insulin secretion assays that normalize for insulin content. Furthermore, recent evidence has suggested the presence of empty β-cells in the pancreas of patients with T1D, namely cells that have lost insulin immunoreactivity but can still be identified using other hallmark features of β-cells (10). The development of a reliable assay that does not rely on insulin protein detection is therefore necessary to accurately measure β-cell mass and function.

DNA methylation annotates the genome to regulate cell type–specific gene expression. It is a stable covalent marker that is conserved among cells of the same type and among individuals and is only minimally changed in disease (11,12). Thus, DNA methylation patterns can serve as a unique biomarker for DNA molecules derived from a given cell type. To identify genomic loci whose methylation patterns can serve as a cell type–specific biomarker, we have recently generated a comprehensive atlas of human cell type DNA methylation profiles (13). The atlas was generated using whole-genome bisulfite sequencing of multiple cell types sorted from surgical or cadaveric organ donors. Comparative analysis of the atlas revealed extensive cell type–specific methylation signatures from which it is possible to generate methylation biomarkers. While the original purpose of the atlas was to facilitate the analysis of DNA fragments circulating in plasma (liquid biopsies), the same information can be used to determine the cellular composition of any DNA mixture.

Herein, we describe the derivation of DNA methylation biomarkers that can distinguish DNA originating in specific pancreatic cells: α-, β-, and δ-cells; acinar cells; ductal epithelial cells; and vascular endothelial cells. In addition, we report a method to quantify the fraction of DNA derived from each cell type in a mixture, using either PCR followed by next-generation sequencing or digital droplet PCR (ddPCR). We then applied the method to assess cell composition in islet preparations and in formalin-fixed, paraffin-embedded (FFPE) sections obtained from the Human Pancreas Analysis Program (HPAP) and the Network for Pancreatic Organ Donors with Diabetes (nPOD).

Study Design and Cohorts

HPAP Cohort

Human pancreata were isolated from deceased multiorgan donors, and islets were isolated for perifusion secretion assays as previously described (14,15). After perifusion assays, DNA was extracted and used for quantification of cell type fractions. The number of samples for each group of donors were as follows: nondiabetic (ND), nondiabetic autoantibody negative (ND Aab−), n = 45; ND Aab+, n = 15; T2D, n = 26; T1D, n = 10.

nPOD Cohort

Two consecutive sections from each FFPE pancreas were obtained from the ND Aab− (n = 24), T2D (n = 15), and T1D (n = 18) donor groups. One section was used for DNA extraction, while the sequential section was subjected to immunostaining.

Donor details for both cohorts are provided in Supplementary Table 1. Sex was registered for each donor, but it was not considered as a factor in the statistical analysis.

Islet Cell Sorting

Human islet samples were obtained from the Integrated Islet Donor Program (16) or from the Alberta Islet Distribution Program. Donor details for both cohorts are provided in Supplementary Table 2. Islets were cultured in Prodo Islet Media containing 5 mmol/L glucose at 37°C in a 5% CO2 humidified incubator for 24 h. Islets were then dissociated using Accumax (Sigma-Aldrich) with DNase I (Roche), fixed with Cytofix/Cytoperm (BD Biosciences), permeabilized with Perm/Wash (BD Biosciences), and fluorescently immunostained for C-peptide, glucagon, and somatostatin for the purification of β-, α-, and δ-cells, respectively. Acinar, ductal, and endothelial cells from islet preparations of low purity were labeled with antibodies to cell surface proteins in PBS/0.5% BSA (Tamar) and FACS sorted (17,18) (Supplementary Table 3). DAPI staining (0.2 μg/mL) was used to exclude dead cells (ND, n = 33; T2D, n = 5). All sorted cells were used to validate specificity and linearity of markers chosen.

Perifusion of Human Islets

Islet perifusion was performed as previously described (7). Briefly, HPAP islet preparations from the four groups (ND Aab−, n = 45; ND Aab+, n = 15; T2D, n = 26; T1D, n = 10) were preperifused with substrate-free medium. Then, islets were incubated with 4 mmol/L amino acid mixture to stimulate glucagon secretion. Next, islets were incubated in low (3 mmol/L) and high (16.7 mmol/L) glucose for insulin secretion and glucagon inhibition. After 20 min of stimulation with high glucose, 0.1 mmol/L 3-isobutyl-1-methylxanthine (IBMX) was added for maximal secretion based on intracellular cAMP levels. Another step of substrate-free medium was performed to remove all stimulants’ effects. Finally, 30 mmol/L KCl was added to release all readily releasable granules.

Immunostaining

Staining was performed as previously described (19). Briefly, paraffin sections (5 μm thick) obtained from nPOD were rehydrated, and antigen retrieval was performed using a pressure cooker. Then, sections were immunostained for glucagon and insulin for the assessment of α- and β-cell fractions, respectively. Nuclear costaining was conducted with DAPI (Invitrogen). A list of primary and secondary antibodies is provided in Supplementary Table 3.

Immunofluorescent images were acquired on a PANORAMIC 250 Flash III machine, 20×. Slide images were converted into Tag Image File Format using CaseViewer with SlideConverter software. Images were processed for glucagon and insulin area, and total area was inferred from DAPI staining using ImageJ software.

DNA Extraction and Bisulfite Treatment

DNA from islets and sorted cells was extracted using the DNeasy Blood and Tissue kit (#69504; QIAGEN) according to the manufacturer’s instructions. DNA from FFPE material was extracted using QIAamp DNA FFPE Tissue Kit (#56404; QIAGEN). DNA concentration was determined using a Qubit double-strand molecular probes kit (Invitrogen). DNA was then treated with sodium bisulfite using EZ DNA Methylation-Gold (Zymo Research) and eluted in 24 μL elution buffer.

Selection of Pancreatic Cell Type Markers

The methylomes of pancreas cell types were described previously (13,20). Specific markers for each pancreatic cell type were selected as previously described (21). Briefly, we compared the whole-genome methylation pattern for each cell type with all other pancreatic cell types to identify uniquely methylated or demethylated blocks using the find markers function in wgbstools version 0.1.0 (https://github.com/nloyfer/wgbs_tools). We searched for blocks that cover four or more CpG sites, with amplicon length of ∼150 base pairs.

Methylation Analysis

Multiplex PCR of bisulfite-treated DNA followed by next-generation sequencing was performed as previously described (21). Supplementary Table 4 lists genomic coordinates of markers and primer sequences. A detailed working protocol for both PCR sequencing and ddPCR is provided in the Supplementary Material.

Data and Resource Availability

The raw sequencing data generated during the current study are available from the corresponding author upon reasonable request. The analyzed data in full are presented in Supplementary Table 5. No applicable resources were generated or analyzed during the current study.

Differentially Methylated Regions That Distinguish Pancreatic Cell Types

The methylation pattern of the human insulin gene has been used in multiple studies (2124). We have previously analyzed the methylation status of five genomic regions spanning the insulin promoter and gene (Fig. 1A). We found that region 5 within intron 2 of the insulin gene (positions +1079 to +1201 downstream of the transcription start site), which contains 10 CpG sites, exhibited the greatest specificity in distinguishing β-cells from acinar cells (demethylated in β-cells and methylated in acinar cells) (21,25). Using our comprehensive human DNA methylation atlas (13), we now analyzed the methylation pattern of these regions in all pancreatic and blood cell types included in the atlas. This analysis confirmed that methylation of insulin region 5 presents the highest specificity in distinguishing β-cells from all other cell types and thus, can be used as a biomarker to identify β-cell DNA (Fig. 1B). We selected 19 additional genomic loci, each containing a cluster of CpG sites with a unique methylation signature that distinguishes six types of pancreatic cells: α-, β-, δ-, acinar, ductal, and pancreatic endothelial cells (Fig. 1C). Most loci are demethylated in the cell type of interest while methylated elsewhere, consistent with observations across the entire atlas (13). One interesting exception is a locus hypermethylated specifically in α-cells, which maps adjacent to the PDX1 gene and is presumably involved in PDX1 repression in α-cells (Fig. 1C, top left). Next, we designed 20 primer pairs, each flanking a selected CpG cluster (including region 5 of the insulin gene). We used a PCR multiplex protocol (21) to coamplify all 20 markers from DNA isolated from pancreatic islets, sorted islet cells, or leukocytes after bisulfite treatment. The amplicons were subjected to next-generation sequencing to assess methylation patterns (Fig. 1D). We previously established that given the regional nature of DNA methylation, maximal cell type specificity is obtained when considering molecules that show a consistent methylation pattern in all adjacent cytosines (26). Accordingly, we scored only molecules in which all CpG sites in a molecule (four to seven in the current markers) exhibited the cell-specific methylation pattern. Using this analysis, we confirmed marker specificity and validated the ability to distinguish pancreatic cell types using DNA samples from the pancreas and isolated cell types. Figure 1E and F shows marker specificity for α- and β-cells; Supplementary Fig. 1AD shows marker specificity for the other cell types studied.

Figure 1

Pancreas cell type–specific DNA methylation markers. A: Schematic of the human insulin gene, highlighting distinct clusters of CpG sites. Red line indicates the CpG sites targeted by ddPCR. B: A heat map showing the methylation status of each insulin gene CpG cluster in the DNA of pancreas and blood cell types based on our published methylome atlas (13). Only region 5 (within intron 2 of insulin) is unmethylated uniquely in β-cells. C: A heat map showing methylation status of 20 loci that are differentially methylated in the indicated pancreas cell types, as inferred from the methylome atlas. In both B and C, for each sorted cell type, the average was calculated for three to four different donors, except for CD3 (n = 2) and islet endothelium (n = 1). D: Schematic of the method used for targeted analysis of methylation markers on DNA extracted from tissue or from histological sections. E: Specificity of α-cell markers. Bars show the percentage of molecules carrying the α-cell–specific methylation pattern (unmethylated for CCDC73 and PELI2, methylated for PDX1) in DNA from the indicated sources. F: Specificity of the β-cell markers. In both E and F, sorted α-cells, n = 6; sorted β-cells, n = 4; sorted δ-cells, n = 6; sorted acinar cells, n = 5; sorted ductal cells, n = 5; and sorted endothelial cells and leukocytes, n = 1. GJ: Sensitivity and linearity of the assay. α- or β-cell DNA was mixed in the indicated proportions with the non–α- or non–β-fraction of dissociated and sorted islets, and methylation markers were used to determine the percentage of α- or β-cell DNA in each mixture. G and H show assay performance across the entire range of concentrations, and α- or β-cell DNA mixed in leukocytes and HEK-293 DNA in I and J show performance where the cell type of interest comprises just 0–3% of the mixture. Circles and triangles represent the average of all markers used for α- or β-cells, respectively. NK, natural killer.

Figure 1

Pancreas cell type–specific DNA methylation markers. A: Schematic of the human insulin gene, highlighting distinct clusters of CpG sites. Red line indicates the CpG sites targeted by ddPCR. B: A heat map showing the methylation status of each insulin gene CpG cluster in the DNA of pancreas and blood cell types based on our published methylome atlas (13). Only region 5 (within intron 2 of insulin) is unmethylated uniquely in β-cells. C: A heat map showing methylation status of 20 loci that are differentially methylated in the indicated pancreas cell types, as inferred from the methylome atlas. In both B and C, for each sorted cell type, the average was calculated for three to four different donors, except for CD3 (n = 2) and islet endothelium (n = 1). D: Schematic of the method used for targeted analysis of methylation markers on DNA extracted from tissue or from histological sections. E: Specificity of α-cell markers. Bars show the percentage of molecules carrying the α-cell–specific methylation pattern (unmethylated for CCDC73 and PELI2, methylated for PDX1) in DNA from the indicated sources. F: Specificity of the β-cell markers. In both E and F, sorted α-cells, n = 6; sorted β-cells, n = 4; sorted δ-cells, n = 6; sorted acinar cells, n = 5; sorted ductal cells, n = 5; and sorted endothelial cells and leukocytes, n = 1. GJ: Sensitivity and linearity of the assay. α- or β-cell DNA was mixed in the indicated proportions with the non–α- or non–β-fraction of dissociated and sorted islets, and methylation markers were used to determine the percentage of α- or β-cell DNA in each mixture. G and H show assay performance across the entire range of concentrations, and α- or β-cell DNA mixed in leukocytes and HEK-293 DNA in I and J show performance where the cell type of interest comprises just 0–3% of the mixture. Circles and triangles represent the average of all markers used for α- or β-cells, respectively. NK, natural killer.

Close modal

We tested the effects of loosening the methylation criterion on assay performance. As expected, relaxing the requirement for full demethylation increased sensitivity as a higher proportion of DNA molecules from β-cells was identified but at the cost of specificity (a small proportion of non–β-islet cells showed a signal) (Supplementary Fig. 1E). Importantly, the methylation pattern of endocrine cell markers was maintained in cells from donors with T2D (Supplementary Fig. 1F and G). To evaluate the accuracy and reliability of the assay for each cell type, we conducted spike-in experiments and assessed standard curve linearity. We FACS sorted each cell type from dissociated immunostained human islet preparations and sorted separately the remaining islet cells. We then spiked increasing fractions of cell type–specific DNA into islet cell DNA from the same donor that did not contain the cell type being analyzed or with DNA from the HEK-293 human embryonic kidney cell line or leukocytes. We then used the methylation assay to assess the proportion of DNA from the spiked cell type. The assay was able to quantitatively detect <1% of specific islet cell type DNA, demonstrating sensitivity and accuracy in measuring small amounts of target DNA (Fig. 1G–J and Supplementary Fig. 1HK).

Assessment of Cell Composition Using ddPCR

We have previously demonstrated that cell type–specific methylation patterns can be assessed using ddPCR, with a performance similar to that of PCR followed by next-generation sequencing (2729). ddPCR provides more limited information than sequencing-based approaches (e.g., current machines have only two to six fluorescent channels), but is faster, cheaper, and simpler to implement. As proof of concept, we developed a ddPCR-based assay for bisulfite-treated DNA, specifically targeting the insulin gene region 5. The assay focused on three CpG sites within this region and used two different probes. One probe encompassed three specifically unmethylated CpG sites, while the other lacked CpG sites and allowed us to count all the DNA molecules, regardless of their methylation status (Fig. 2A). ddPCR analysis of sorted primary β-, α-, δ-, acinar, and ductal cell DNA, as well as HEK-293 and the EndoC human β-cell line, revealed that the assay is specific: non–β-cells have a very low baseline signal (<5% unmethylated molecules) compared with β-cells, EndoC cells, and human islets (90, 80, and average 25% unmethylated molecules, respectively) (Fig. 2B). The β-cell DNA spike-in samples tested in the sequencing experiment (Fig. 1H) were assayed by ddPCR. This experiment confirmed that the ddPCR assay is accurate enough to measure β-cell DNA in human islets (Fig. 2C).

Figure 2

ddPCR for detection of human β-cell DNA. A: Schematic of the method. One pair of primers is used to amplify a region in the insulin gene after bisulfite treatment. One TaqMan probe (hexachlorofluorescein [HEX] fluorophore) reports on the total number of template molecules (i.e., the number of positive droplets), while another probe (fluorescein amidite [FAM] fluorophore) reports on the number of unmethylated insulin molecules. B: Percentage of β-cell DNA as determined using ddPCR on DNA from the indicated sources. The ddPCR TaqMan probe for β-cell–specific methylation targets the second intron of the human insulin gene and is specific for demethylated CpG sites in positions +1169, +1172, and +1180. Islets, n = 6; sorted β-cells, n = 9; sorted β-cells T2D, n = 3; EndoC, n = 1; sorted α-cells, n = 8; sorted δ-cells, n = 6; sorted acinar cells, n = 4; sorted ductal cells, n = 4; and HEK-293 cells, n = 1. C: Comparison of the performance of β-cell methylation markers using ddPCR and next-generation sequencing (NGS). DNA from sorted human β-cells (0–20 ng) was mixed with DNA from sorted human islet non–β-cells (total DNA 20 ng) from the same donor. Analysis of DNA was performed using NGS targeting three specific β-cell markers, ddPCR targeting three CpG sites within the insulin gene, and NGS targeting the same three CpG sites as ddPCR.

Figure 2

ddPCR for detection of human β-cell DNA. A: Schematic of the method. One pair of primers is used to amplify a region in the insulin gene after bisulfite treatment. One TaqMan probe (hexachlorofluorescein [HEX] fluorophore) reports on the total number of template molecules (i.e., the number of positive droplets), while another probe (fluorescein amidite [FAM] fluorophore) reports on the number of unmethylated insulin molecules. B: Percentage of β-cell DNA as determined using ddPCR on DNA from the indicated sources. The ddPCR TaqMan probe for β-cell–specific methylation targets the second intron of the human insulin gene and is specific for demethylated CpG sites in positions +1169, +1172, and +1180. Islets, n = 6; sorted β-cells, n = 9; sorted β-cells T2D, n = 3; EndoC, n = 1; sorted α-cells, n = 8; sorted δ-cells, n = 6; sorted acinar cells, n = 4; sorted ductal cells, n = 4; and HEK-293 cells, n = 1. C: Comparison of the performance of β-cell methylation markers using ddPCR and next-generation sequencing (NGS). DNA from sorted human β-cells (0–20 ng) was mixed with DNA from sorted human islet non–β-cells (total DNA 20 ng) from the same donor. Analysis of DNA was performed using NGS targeting three specific β-cell markers, ddPCR targeting three CpG sites within the insulin gene, and NGS targeting the same three CpG sites as ddPCR.

Close modal

Assessment of Islet Cell Composition and Insulin Secretion

We next turned to assessing cell numbers in pancreas preparations using methylation markers and next-generation sequencing as outlined in Fig. 1C and compared the results with standard methods. We selected settings where accurate measurement of cell numbers is important for the understanding of key biological or pathological processes. Measuring glucose-stimulated insulin secretion in cadaveric human islets is the gold standard for assaying β-cell function ex vivo. Insulin secreted is typically normalized to total islet insulin content. Such normalization might be inadequate in cases where insulin content is altered, such as during prolonged hyperglycemia. An alternative is to measure insulin secretion on a per-islet basis or normalize to total DNA. The caveat with this method is that islet cell composition is highly heterogeneous (30), and the β-cell fraction can be reduced in pathological conditions such as diabetes, thus leading to erroneous estimation of insulin secretion per β-cell (in the case of reduced β-cell number, an underestimation of secretion per cell). Similarly, measurements of insulin content are important (e.g., when assessing stem cell differentiation to β-cells) but also require proper normalization, ideally for the number of β-cells. We reasoned that normalization to β-cell DNA, which is a stable and quantitative feature of β-cell number, would be optimal. We obtained DNA and insulin secretion data from islet preparations from donors who were ND (with or without Aab), had T2D, or had T1D, all processed through the HPAP. Using the DNA methylation–based assay, we measured for each islet preparation the fractions of β-, α-, and contaminating acinar cells. Total DNA content was similar in all groups (Fig. 3A). Insulin content per islet was similar in the ND Aab−, Aab+ and T2D groups and, as expected, was lower in the T1D preparations (Fig. 3B). In theory, the latter observation results from the near absence of β-cells in the T1D islet preparations. Indeed, insulin content per β-cell, as determined using the methylation assay, was normal (Fig. 3C). Glucagon content was normal when normalized to either islet number or α-cell number (Fig. 3D and E).

Figure 3

HPAP islet cell composition and normalized β-cell function. A: Total DNA content extracted from the indicated preparations, in nanograms, normalized per 100 islets. B: Insulin content as measured by radioimmunoassay and normalized per 100 islets. C: Insulin content as measured by radioimmunoassay and normalized to β-cell number in 100 islets. D: Glucagon content as measured by radioimmunoassay per 100 islets. E: Glucagon content as measured by radioimmunoassay and normalized to α-cell number in 100 islets. F: Islet cell composition. Shown is the percentage of methylation markers of α-, β-, and acinar cells, cumulatively. Each bar represents the average of all samples in each group. GI: Dynamic insulin secretion from perifused human islet preparations in different conditions, normalized to 100 islets (G), to insulin content in 100 islets (H), and to β-cell number in 100 islets, as determined using methylation markers (I). J: Insulin secretion in high glucose as a function of β-cell number in the same islet preparation. ND Aab−, n = 39–45; ND Aab+, n = 12–15; T2D, n = 25–27; T1D, n = 9–10. Black open circles represent donors who were ND with more than one autoantibody. *P < 0.05. AAM, amino acid mixture.

Figure 3

HPAP islet cell composition and normalized β-cell function. A: Total DNA content extracted from the indicated preparations, in nanograms, normalized per 100 islets. B: Insulin content as measured by radioimmunoassay and normalized per 100 islets. C: Insulin content as measured by radioimmunoassay and normalized to β-cell number in 100 islets. D: Glucagon content as measured by radioimmunoassay per 100 islets. E: Glucagon content as measured by radioimmunoassay and normalized to α-cell number in 100 islets. F: Islet cell composition. Shown is the percentage of methylation markers of α-, β-, and acinar cells, cumulatively. Each bar represents the average of all samples in each group. GI: Dynamic insulin secretion from perifused human islet preparations in different conditions, normalized to 100 islets (G), to insulin content in 100 islets (H), and to β-cell number in 100 islets, as determined using methylation markers (I). J: Insulin secretion in high glucose as a function of β-cell number in the same islet preparation. ND Aab−, n = 39–45; ND Aab+, n = 12–15; T2D, n = 25–27; T1D, n = 9–10. Black open circles represent donors who were ND with more than one autoantibody. *P < 0.05. AAM, amino acid mixture.

Close modal

Islet cell composition was also strikingly different in T1D preparations. While α- and β-cells constituted the major fraction of most islet preparations from the ND or T2D groups, islets from the T1D group were predominantly contaminated with acinar cells (Fig. 3F and Supplementary Fig. 2AC). Consistently, insulin secretion per islet was extremely low in T1D islets (Fig. 3G and Supplementary Fig. 2D). In contrast, insulin secretion normalized to insulin content was higher than normal in T1D islets (Fig. 3H and Supplementary Fig. 2E). Strikingly, normalization of glucose-stimulated insulin secretion to β-cell DNA content revealed that insulin secretion from T1D islets was similar to Aab+ and T2D islets, and all three groups secreted distinctively less than islets from the ND Aab− group (Fig. 3I and Supplementary Fig. 2F). Overall, we observed a positive correlation between the absolute quantity of β-cell DNA in an islet preparation and the amount of secreted insulin (r = 0.55) (Fig. 3J and Supplementary Fig. 2G). These findings demonstrate how measurements of β-cell number may better inform how diabetes alters basic β-cell features, such as insulin content and secretion.

Assessment of α- and β-Cell Fractions in Pancreatic Tissue Sections

Measuring β- and α-cell mass in histological pancreas specimens obtained from cadaveric donors can help to determine the relative contribution of deficient mass versus deficient function of β- and α-cells in diabetes. Standard assessment of α- and β-cells is performed using immunostaining for glucagon and insulin on multiple histological sections, scanning entire slides, and quantifying the fraction of tissue stained. Multiplying the fraction of a cell type averaged over multiple sections by the total weight of the pancreas then provides an estimate for the total mass of α- or β-cells. Although widely used, this method is prone to errors due to technical variations in staining quality or biological variations in hormone expression levels. In addition, since different cell types have different sizes, the fraction of stained tissue would overestimate the proportion of large cells and underestimate the proportion of small cells.

To examine the utility of methylation-based assessment of cell numbers, we obtained FFPE pancreatic sections from donors who were ND (n = 24) and donors with T2D (n = 15) and T1D (n = 18) through nPOD. To determine the fraction of α- and β-cells in these specimens, we stained slides for insulin and glucagon using immunofluorescence, scanned the entire slide in high resolution, and measured the fraction of stained tissue. We then used a serial section for DNA extraction and determined the fraction of α- and β-cell DNA using the methylation markers. We observed a strong correlation between the two methods for both α- and β-cells across the entire set of samples (Spearman correlation r = 0.63–0.8) (Fig. 4A and B). We then examined α- and β-cell fraction and mass in relation to disease status. We derived rough estimates for β- and α-cell mass by analyzing a single randomly selected histological section. Note that all sections used were from the pancreas tail. Pancreas weight (measured at nPOD when obtaining the organ) was higher than normal in the donors with T2D (explained by the higher BMI of these donors) (Supplementary Table 1) and was significantly lower in the donors with T1D, as previously reported (3133) (Fig. 4C).

Figure 4

Quantification of α- and β-cell fractions in DNA extracted from sections of FFPE pancreata. A and B: α-Cell and β-cell DNA, as a function of the percentage of glucagon- and insulin-stained area in the same slide. ND Aab−, n = 24; T2D, n = 15; and T1D, n = 18. In all panels, black triangles represent patients with recently diagnosed T1D. C: Pancreas weight in the indicated donors. ND Aab−, n = 20; T2D, n = 14; and T1D, n = 16. D and E: Percentage of β-cell DNA and insulin-stained area in pancreatic slides from the indicated donors. Black triangles indicate patients with recently diagnosed T1D (0 duration of disease). F and G: β-Cell mass calculated by multiplying pancreas mass by the fraction of β-cells inferred from methylation markers (F) and from insulin immunostaining (G). ND Aab−, n = 20; T2D, n = 14; and T1D, n = 16. H and I: Percentage of α-cell DNA and glucagon-stained area in pancreatic slides from the indicated donors. J and K: α-Cell mass calculated by multiplying pancreas mass by the fraction of α-cells inferred from methylation markers (J) and from glucagon immunostaining (K). *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Figure 4

Quantification of α- and β-cell fractions in DNA extracted from sections of FFPE pancreata. A and B: α-Cell and β-cell DNA, as a function of the percentage of glucagon- and insulin-stained area in the same slide. ND Aab−, n = 24; T2D, n = 15; and T1D, n = 18. In all panels, black triangles represent patients with recently diagnosed T1D. C: Pancreas weight in the indicated donors. ND Aab−, n = 20; T2D, n = 14; and T1D, n = 16. D and E: Percentage of β-cell DNA and insulin-stained area in pancreatic slides from the indicated donors. Black triangles indicate patients with recently diagnosed T1D (0 duration of disease). F and G: β-Cell mass calculated by multiplying pancreas mass by the fraction of β-cells inferred from methylation markers (F) and from insulin immunostaining (G). ND Aab−, n = 20; T2D, n = 14; and T1D, n = 16. H and I: Percentage of α-cell DNA and glucagon-stained area in pancreatic slides from the indicated donors. J and K: α-Cell mass calculated by multiplying pancreas mass by the fraction of α-cells inferred from methylation markers (J) and from glucagon immunostaining (K). *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Close modal

Both immunostaining and methylation analyses revealed a substantially lower β-cell fraction and mass in sections from donors with T1D but not with T2D (Fig. 4D–G). Among the T1D samples, those with higher β-cell fraction and mass were from recently diagnosed donors, indicating that in these cases, β-cell destruction is accompanied or preceded by β-cell dysfunction (and that both processes contribute to hyperglycemia). All cases of long-standing T1D had extremely low β-cell fraction in both the immunostaining and DNA measurements, arguing against a massive process of dedifferentiation that eliminates insulin expression but retains viable empty β-cells. Consistently with this idea, the fraction of β-cells in donors with T1D as measured by methylation was negatively correlated with the duration of disease and positively correlated with the age of onset (Supplementary Fig. 3). Interestingly, the fraction of β-cells measured using DNA methylation was consistently higher than the fraction measured using immunostaining (Fig. 4B, D, and E), potentially because of the relatively small size of β-cells compared with acinar cells.

Both immunostaining and DNA methylation demonstrated a higher-than-normal fraction and total mass of α-cells in T2D sections (Fig. 4H–K), consistent with previous reports (6). The α-cell fraction was higher than normal in the T1D sections, but adjustment for pancreas weight revealed that there was no significant difference in α-cell mass between T1D and ND samples (Fig. 4J and K).

The method reported here allows for objective quantification of the DNA from multiple important cell types in the human pancreas, with applications for assessment of cell function and tissue composition. We have developed the method using our standard PCR sequencing platform, but we also provide a simplified, cheaper, and faster approach based on ddPCR, a technology that is more readily available in many laboratories. While we believe that the method can be useful already in its current form, there are obviously ways to develop it further. For example, the ligation of unique molecular identifiers to DNA molecules prior to PCR amplification may greatly reduce intra-assay variation resulting from PCR duplications.

We propose that there are multiple settings in which this method may provide unique information not available through standard immunodetection of cellular markers. Even if equally accurate, digital methylation-based analysis is simpler and faster than immunostaining, particularly in applications that involve image analysis. For example, the assessment of β- and α-cell mass in human xenografts placed under the mouse kidney is challenging when using immunostaining, and DNA-based assessment may provide a convenient alternative. In addition, the same starting material can be used for determining the proportion of any other cell type in a specimen via additional tissue-specific methylation markers. Another example of potential utility is in the assessment of β-cell mass. It has been proposed that the large variation observed among individuals may underlie the risk for diabetes (34). It is possible that at least part of the perceived variation results from different insulin expression levels or staining quality. DNA-based assays provide an opportunity to independently assess interindividual variation of β-cell mass.

An obvious caveat in methylation-based analysis of tissue composition is the loss of spatial information. Given the rapid development of spatial omics technology, we predict that this limitation will be overcome soon and will allow us to analyze tissue sections using antibody-independent, methylation-based cell type–specific markers. Another limitation concerns the applicability to other species. Methylation markers for specific human cell types will typically not work in mice, given species variation in genomic sequence.

This study was primarily aimed at developing the method. However, the analysis of material from healthy donors and donors with diabetes already suggests some biological insights, which invite further experiments with different approaches. Most interestingly, insulin secretion per β-cell was similar in cultured islets from the T1D group and from the ND Aab+ and T2D groups, although it was significantly lower than in the ND group, which indicates β-cell dysfunction in all three groups. Notably, there are mixed reports regarding β-cell function in T1D (7,8,3537). Regardless, β-cell function in isolated islets from donors with T1D contrasts with the near-total loss of β-cell function in recently diagnosed patients in vivo, even though some of these patients have up to 40% of normal β-cell mass remaining. This suggests that the β-cell dysfunction in T1D is imposed by extrinsic factors, potentially cytokines and glucose load, and is reversed when islets are put into culture. We also note that our analysis did not reveal reduced β-cell mass in histological material from patients with T2D (Fig. 4F and G), in contrast to dogma. However, since we have used just one random paraffin section from each pancreas, and given the small sample size, these measurements are insufficient to conclude that β-cell mass is normal in T2D.

In conclusion, DNA methylation-based measurements of the fraction of β-cells and other pancreatic cell types in mixtures are feasible, accurate, and convenient. Furthermore, they may provide unique information on functionality and total cell mass in health and disease.

This article contains supplementary material online at https://doi.org/10.2337/figshare.25044500.

This article is featured in podcasts available at diabetesjournals.org/diabetes/pages/diabetesbio.

Funding. This work was supported by grants from Grail, JDRF, and the Human Islet Research Network, National Institute of Diabetes and Digestive and Kidney Diseases, Alex U. Soyka Pancreatic Cancer Fund, Israel Science Foundation, and Diabetes Onderzoek Nederland (DON) Foundation (to Y.D.) and by Division of Diabetes, Endocrinology, and Metabolic Diseases grants DK133442-01A1 and DK135001-01. Y.D. holds the Walter and Greta Stiel chair and research grant in heart studies. This manuscript used data acquired from the HPAP (Research Resource Identifier [RRID]: SCR_016202) Database (https://hpap.pmacs.upenn.edu); Human Islet Research Network (RRID: SCR_014393) Consortium grants UC4-DK-112217, U01-DK-123594, UC4-DK-112232, and U01-DK-123716; nPOD (RRID: SCR_014641), a collaborative T1D research project supported by JDRF grant 5-SRA-2018-557-Q-R; and The Leona M. & Harry B. Helmsley Charitable Trust grants 2018PG-T1D053 and G-2108-04793.

The content and views expressed are the responsibility of the authors and do not necessarily reflect the official view of nPOD. Organ Procurement Organizations partnering with nPOD to provide research resources are listed at https://npod.org/for-partners/npod-partners.

Duality of Interest. D.N., J.M., T.K., R.S., B.G., A.K., and Y.D. have filed patents on DNA methylation biomarkers and methods. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. Z.D., D.N., O.F., A.P., J.M., A.V.R., N.M.D., and J.S. researched data. Z.D. and Y.D. wrote the manuscript. D.A.St., K.H.K., D.A.Sc., C.W., M.C.-T., T.K., R.S., B.G., A.K., and Y.D. contributed to the discussion and reviewed and edited manuscript. Y.D. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this article were presented in abstract and poster form at the Human Islet Research Network 2022 Annual Investigator Meeting, Washington, DC, 14–16 September 2022, and as a talk at the Human Islet Research Network 2023 Annual Investigator Meeting, Washington, DC, 19–21 September 2023.

1.
Campbell-Thompson
M
,
Fu
A
,
Kaddis
JS
, et al
.
Insulitis and β-cell mass in the natural history of type 1 diabetes
.
Diabetes
2016
;
65
:
719
731
2.
Leete
P
,
Willcox
A
,
Krogvold
L
, et al
.
Differential insulitic profiles determine the extent of β-cell destruction and the age at onset of type 1 diabetes
.
Diabetes
2016
;
65
:
1362
1369
3.
Wang
YJ
,
Traum
D
,
Schug
J
, et al;
HPAP Consortium
.
Multiplexed in situ imaging mass cytometry analysis of the human endocrine pancreas and immune system in type 1 diabetes
.
Cell Metab
2019
;
29
:
769
783.e4
4.
Aly
H
,
Gottlieb
P.
The honeymoon phase: intersection of metabolism and immunology
.
Curr Opin Endocrinol Diabetes Obes
2009
;
16
:
286
292
5.
Butler
AE
,
Janson
J
,
Bonner-Weir
S
,
Ritzel
R
,
Rizza
RA
,
Butler
PC.
Beta-cell deficit and increased beta-cell apoptosis in humans with type 2 diabetes
.
Diabetes
2003
;
52
:
102
110
6.
Wu
M
,
Lee
MYY
,
Bahl
V
, et al;
HPAP Consortium
.
Single-cell analysis of the human pancreas in type 2 diabetes using multi-spectral imaging mass cytometry
.
Cell Rep
2021
;
37
:
109919
7.
Doliba
NM
,
Rozo
AV
,
Roman
J
, et al;
HPAP Consortium
.
α Cell dysfunction in islets from nondiabetic, glutamic acid decarboxylase autoantibody-positive individuals
.
J Clin Invest
2022
;
132
:
e156243
8.
Brissova
M
,
Haliyur
R
,
Saunders
D
, et al
.
α Cell function and gene expression are compromised in type 1 diabetes
.
Cell Rep
2018
;
22
:
2667
2676
9.
Hosein
AN
,
Brekken
RA
,
Maitra
A.
Pancreatic cancer stroma: an update on therapeutic targeting strategies
.
Nat Rev Gastroenterol Hepatol
2020
;
17
:
487
505
10.
Wasserfall
C
,
Nick
HS
,
Campbell-Thompson
M
, et al
.
Persistence of pancreatic insulin mRNA expression and proinsulin protein in type 1 diabetes pancreata
.
Cell Metab
2017
;
26
:
568
575.e3
11.
Dor
Y
,
Cedar
H.
Principles of DNA methylation and their implications for biology and medicine
.
Lancet
2018
;
392
:
777
786
12.
Cedar
H
,
Bergman
Y.
Linking DNA methylation and histone modification: patterns and paradigms
.
Nat Rev Genet
2009
;
10
:
295
304
13.
Loyfer
N
,
Magenheim
J
,
Peretz
A
, et al
.
A DNA methylation atlas of normal human cell types
.
Nature
2023
;
613
:
355
364
14.
Shapira
SN
,
Naji
A
,
Atkinson
MA
,
Powers
AC
,
Kaestner
KH.
Understanding islet dysfunction in type 2 diabetes through multidimensional pancreatic phenotyping: the Human Pancreas Analysis Program
.
Cell Metab
2022
;
34
:
1906
1913
15.
Kaestner
KH
,
Powers
AC
,
Naji
A
;
HPAP Consortium
; Atkinson MA.
NIH initiative to improve understanding of the pancreas, islet, and autoimmunity in type 1 diabetes: the human pancreas analysis program (HPAP)
.
Diabetes
2019
;
68
:
1394
1402
16.
Brissova
M
,
Niland
JC
,
Cravens
J
,
Olack
B
,
Sowinski
J
,
Evans-Molina
C.
The integrated islet distribution program answers the call for improved human islet phenotyping and reporting of human islet characteristics in research articles
.
Diabetes
2019
;
68
:
1363
1365
17.
Dorrell
C
,
Schug
J
,
Lin
CF
, et al
.
Transcriptomes of the major human pancreatic cell types
.
Diabetologia
2011
;
54
:
2832
2844
18.
Dorrell
C
,
Abraham
SL
,
Lanxon-Cookson
KM
,
Canaday
PS
,
Streeter
PR
,
Grompe
M.
Isolation of major pancreatic cell types and long-term culture-initiating cells using novel human surface markers
.
Stem Cell Res (Amst)
2008
;
1
:
183
194
19.
Swisa
A
,
Avrahami
D
,
Eden
N
, et al
.
PAX6 maintains β cell identity by repressing genes of alternative islet cell types
.
J Clin Invest
2017
;
127
:
230
243
20.
Moss
J
,
Magenheim
J
,
Neiman
D
, et al
.
Comprehensive human cell-type methylation atlas reveals origins of circulating cell-free DNA in health and disease
.
Nat Commun
2018
;
9
:
5068
21.
Neiman
D
,
Gillis
D
,
Piyanzin
S
, et al
.
Multiplexing DNA methylation markers to detect circulating cell-free DNA derived from human pancreatic β cells
.
JCI Insight
2020
;
5
:
e136579
22.
Husseiny
MI
,
Kaye
A
,
Zebadua
E
,
Kandeel
F
,
Ferreri
K.
Tissue-specific methylation of human insulin gene and PCR assay for monitoring beta cell death
.
PLoS One
2014
;
9
:
e94591
23.
Fisher
MM
,
Watkins
RA
,
Blum
J
, et al
.
Elevations in circulating methylated and unmethylated preproinsulin DNA in new-onset type 1 diabetes
.
Diabetes
2015
;
64
:
3867
3872
24.
Herold
KC
,
Usmani-Brown
S
,
Ghazi
T
, et al;
Type 1 Diabetes TrialNet Study Group
.
β Cell death and dysfunction during type 1 diabetes development in at-risk individuals
.
J Clin Invest
2015
;
125
:
1163
1173
25.
Neiman
D
,
Moss
J
,
Hecht
M
, et al
.
Islet cells share promoter hypomethylation independently of expression, but exhibit cell-type-specific methylation in enhancers
.
Proc Natl Acad Sci U S A
2017
;
114
:
13525
13530
26.
Lehmann-Werman
R
,
Neiman
D
,
Zemmour
H
, et al
.
Identification of tissue-specific cell death using methylation patterns of circulating DNA
.
Proc Natl Acad Sci U S A
2016
;
113
:
E1826
E1834
27.
Lehmann-Werman
R
,
Magenheim
J
,
Moss
J
, et al
.
Monitoring liver damage using hepatocyte-specific methylation markers in cell-free circulating DNA
.
JCI Insight
2018
;
3
:
e120687
28.
Zemmour
H
,
Planer
D
,
Magenheim
J
, et al
.
Non-invasive detection of human cardiomyocyte death using methylation patterns of circulating DNA
.
Nat Commun
2018
;
9
:
1443
29.
Shemer
R
,
Magenheim
J
,
Dor
Y.
Digital droplet PCR for monitoring tissue-specific cell death using DNA methylation patterns of circulating cell-free DNA
.
Curr Protoc Mol Biol
2019
;
127
:
e90
30.
Wang
YJ
,
Golson
ML
,
Schug
J
, et al
.
Single-cell mass cytometry analysis of the human endocrine pancreas
.
Cell Metab
2016
;
24
:
616
626
31.
Virostko
J
,
Williams
J
,
Hilmes
M
, et al
.
Pancreas volume declines during the first year after diagnosis of type 1 diabetes and exhibits altered diffusion at disease onset
.
Diabetes Care
2019
;
42
:
248
257
32.
Campbell-Thompson
ML
,
Kaddis
JS
,
Wasserfall
C
, et al
.
The influence of type 1 diabetes on pancreatic weight
.
Diabetologia
2016
;
59
:
217
221
33.
Campbell-Thompson
ML
,
Filipp
SL
,
Grajo
JR
, et al
.
Relative pancreas volume is reduced in first-degree relatives of patients with type 1 diabetes
.
Diabetes Care
2019
;
42
:
281
287
34.
Weir
GC
,
Gaglia
J
,
Bonner-Weir
S.
Inadequate β-cell mass is essential for the pathogenesis of type 2 diabetes
.
Lancet Diabetes Endocrinol
2020
;
8
:
249
256
35.
Krogvold
L
,
Skog
O
,
Sundström
G
, et al
.
Function of isolated pancreatic islets from patients at onset of type 1 diabetes: insulin secretion can be restored after some days in a nondiabetogenic environment in vitro: results from the DiViD study
.
Diabetes
2015
;
64
:
2506
2512
36.
Panzer
JK
,
Hiller
H
,
Cohrs
CM
, et al
.
Pancreas tissue slices from organ donors enable in situ analysis of type 1 diabetes pathogenesis
.
JCI Insight
2020
;
5
:
e134525
37.
Marchetti
P
,
Dotta
F
,
Ling
Z
, et al
.
Function of pancreatic islets isolated from a type 1 diabetic patient
.
Diabetes Care
2000
;
23
:
701
703
Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. More information is available at https://www.diabetesjournals.org/journals/pages/license.