Exocrine Pancreas in Type 1 and Type 2 Diabetes: Different Patterns of Fibrosis, Metaplasia, Angiopathy, and Adiposity

Pancreatic islets, the clusters of endocrine cells responsible for glucose homeostasis, are surrounded by exocrine tissue that principally assists in nutrient digestion. The close structural relationship between the endocrine and exocrine compartments of the pancreas has functional and pathologic importance in health and disease, including in pancreatitis, pancreatic cancer, cystic fibrosis, type 1 diabetes (T1D), and type 2 diabetes (T2D). In contrast to extensive studies of islet structure and pathology, relatively little is known about the human exocrine pancreas or how it changes with age, obesity, and diabetes.

The endocrine and exocrine compartments are derived from a common pancreatic progenitor and may share vascular communication (1–3). Whether by vascular delivery or paracrine mechanisms, the exocrine tissue is exposed to higher concentrations of islet-derived hormones, such as insulin and somatostatin, relative to other tissues (1), which likely influences exocrine function (2). Conversely, exocrine pathologies are also known to affect endocrine function. For example, cystic fibrosis (3), chronic pancreatitis (4), and mutations in exocrine genes like CEL (5), GP2, and CTRB (6) may all lead to diabetes through mechanisms involving exocrine–endocrine communication. The relationship between pancreatic cancer and diabetes also suggests intercompartmental cross talk. Individuals with T2D are at the highest relative risk of liver and pancreatic cancers among the various cancers associated with T2D (7,8), which is speculated to be caused at least in part by increased exposure to insulin or other islet-derived signals (9). At the same time, new-onset T2D is sometimes diagnosed concurrently with pancreatic cancer (10) for unknown reasons.

Despite the relevance of the exocrine pancreas to diseases of the exocrine and endocrine systems, how the exocrine pancreas changes with age, obesity, and diabetes remains largely unknown. Although imaging studies have established differences in pancreas volume across a range of ages and BMIs (11), histologic analyses of human pancreatic exocrine tissue have historically been performed in relatively small cohorts of donors, with limited matching for age or sex (12), or using primarily autopsy specimens (13), which may be compromised because of tissue quality.

We leveraged access to well-preserved pancreatic specimens through the National Institutes of Health–supported Human Pancreas Analysis Program (HPAP; Research Resource Identifier SCR_016202), the Network for Pancreatic Organ Donors With Diabetes (nPOD; Research Resource Identifier SCR_014641), and the Vanderbilt Pancreas Biorepository (VPB) to characterize exocrine changes across age, obesity, T1D, and T2D compared with those in control donors with no diabetes (ND). Using classical tissue staining with hematoxylin-eosin (H-E) and Masson trichrome, in combination with multiplexed immunofluorescence, we evaluated pancreatic tissue for metaplastic changes, fibrosis, angiopathy, inflammation, and adiposity and correlated these findings with donor clinical characteristics.

Pancreatic tissue or images of stained tissue sections and deidentified clinical information from 74 ND donors, 20 donors with T1D, and 25 donors with T2D were obtained through partnerships with the International Institute for the Advancement of Medicine, the National Disease Research Interchange, local organ procurement organizations, and nPOD (n = 16) and efforts related to HPAP (n = 35) and VPB (n = 68). All organs were processed within 20 h from intraoperative cross clamp as part of the above programs, as described previously (14–19). The Vanderbilt University Institutional Review Board does not classify deidentified human pancreatic specimens as human subject research. Experimental schematic and donor cohort characteristics are summarized in Fig. 1, with individual donor information listed in Supplementary Table 1. Characteristics of age- and sex-matched cohorts are summarized in Supplementary Tables 2 to 4.

A detailed protocol for assessing the histologic features described in this manuscript, including additional images and examples, is available on protocols.io (20). Sections (5 μm) from formalin-fixed paraffin-embedded tissue blocks were stained with H-E or Masson trichrome blue (cat. no. 9176A; Newcomer Supply). Stained slides were imaged at ×20 magnification using an Aperio ScanScope CS (Leica Biosystems). Sections were assessed and scored for exocrine histologic features by two independent blinded pathologists (A.E. and A.W.). Scores of both pathologists were compared, and discrepancies were discussed and resolved to obtain a final consensus score. One section each from the head, body, and tail of the pancreas was analyzed per donor.

H-E slides were assessed for the presence of acinar-to-ductal metaplasia (ADM), pancreatic intraepithelial neoplasia (PanIN), acinar atrophy, medium blood vessel angiopathy, and adiposity. Per donor, the number of ADM-positive lobules or PanIN lesions was quantified for a total count per square centimeter of tissue analyzed (Supplementary Fig. 1). Acinar atrophy was graded according to a 3-point scale (0–3) representing normal, mild, moderate, or severe atrophy (Supplementary Fig. 2). Scores from each region were added together to obtain a cumulative score (maximum of 9) (Supplementary Table 5). To evaluate angiopathy, medium-sized intralobular blood vessels, defined as those with lumens distinguishable at ×20 magnification, were graded on a 4-point scale (0–4) according to degree of vessel wall thickening (21) (Supplementary Fig. 3). Large interlobular arteries were excluded from analysis because of their sporadic occurrence in sampled tissue sections and increased variability in ratios of lumen to vessel wall. Each section was assigned a score based on the most advanced stage of angiopathy observed throughout the entire tissue section, and scores from each region were added together to obtain a cumulative score (maximum of 12) (Supplementary Table 5). Pancreatic adiposity, defined as the fraction of tissue area composed of adipocytes, was determined in intralobular and interlobular compartments using semiquantitative and quantitative methods. For semiquantitative assessment of adiposity, intralobular and interlobular adipocytes in each section were graded on a 3-point scale (0–3), representing no, mild, moderate, or severe adiposity (22). The total pancreatic adiposity score was calculated as a sum of the inter- and intralobular adiposity scores in all pancreatic regions (maximum score of 18) (Supplementary Table 5). To confirm the semiquantitative adipocyte scoring results, we developed an automated method for adipocyte quantification (Supplementary Fig. 4) that provides a higher-resolution measurement of tissue adiposity, although it introduces some error as a result of the possible characterization of small ducts and blood vessels as adipocytes. Scanned images were analyzed using HALO software (version 3.1; Indica Labs). Manual annotations were drawn to define the total cross-sectional pancreatic area, intralobular area (excluding nonfat areas, such as large blood vessels and ducts), and interlobular area (excluding extrapancreatic tissue, large vessels, and ducts). Unstained tissue within the intra- and interlobular annotations, representing adipocytes because of manual exclusion of large vessels and ducts, was automatically detected and quantified relative to the total tissue area. Adipocyte size (measured by manually outlining adipocytes in representative tissue regions) and total adipocyte number (estimated by dividing total adipocyte area by mean adipocyte size) were evaluated for contribution to total pancreatic adiposity.

Masson trichrome blue staining was used for quantitation of fibrosis in all donors for whom tissue was available for special staining, specifically tissues that were from VPB (n = 68) (Supplementary Table 1). The percentage of tissue area staining blue was automatically quantified using the HALO software random forest classification algorithm, as previously described (23). To quantitatively characterize the localization of pancreatic fibrosis, duct vessel area was manually outlined and quantified in H-E sections.

For assessment of tissue inflammation and capillary morphology, we used the PhenoCycler Open (previously CODEX) platform (Akoya Biosciences), as described previously (18) and detailed on protocols.io (24). In brief, paraformaldehyde-fixed pancreas sections from a subset of donors with short-duration T2D (n = 10) and ND donors of similar age (n = 6) were mounted onto 22 × 22 mm glass coverslips (Electron Microscopy Sciences) coated in 0.1% poly-L-lysine (Sigma-Aldrich) and stained with the CODEX Staining Kit. Tissue sections were incubated with oligonucleotide-conjugated antibodies specific for the indicated antigens (Supplementary Table 6). Sections were counterstained with nuclear stain diluted in 1:1,000 PBS, and automated image acquisition and fluidics exchange were performed using the Akoya CODEX instrument and CODEX Instrument Manager (version 1.29) driver software (Akoya Biosciences) integrated with a BZ-X810 epifluorescent microscope (Keyence), according to manufacturer instructions. Hybridization/stripping of the fluorophore-conjugated complementary oligonucleotides was performed using dimethyl sulfoxide (Sigma-Aldrich). For tissue visualization, the imaging area was set by center point and tile number using BZ-X810 viewing software (Keyence). All images were acquired using a CFI Plan Apo I ×20/0.75 objective (Nikon) with 30% tile overlap and five z-planes (1.5 μm/z). Image alignment, stitching, background subtraction, and deconvolution were performed using the CODEX processor (version 1.7.0.6; Akoya Biosciences). TIFF images were imported into HALO software (Indica Labs) for analysis.

The exocrine area of the entire tissue section was annotated, excluding islets based on chromogranin-A staining. Cell segmentation based on DAPI nuclear counterstain and cell-type annotations were performed using the HALO HighPlex FL (version 3.2.1) module with consistent cytonuclear parameters. Because of marker intensity variability among samples, thresholds were manually set for each marker and donor. Cells were counted positive for a given marker if minimum intensity was reached in 50% of the cytoplasm area or nuclear area. Endothelial cell area (CD31) was measured by a random forest classification algorithm (HALO Tissue Classifier module).

Each T1D or T2D donor was matched to a single sex-concordant ND control with the smallest difference in age. Donors for whom tissue was available for trichrome staining (those from VPB) were preferentially paired with other VPB donors, as long as they were within 8 years of age. Wilcoxon matched-pairs signed rank tests were used to calculate P values for all intergroup comparisons. For calculations on CODEX images, which were not age matched, Welch t tests with correction for multiple comparisons using the Holm-Šídák method were used. Graphs of continuous variables (ADM, PanIN, trichrome, and CODEX cell percentages) show the mean ± SD. Graphs of ordinal variables (atrophy, angiopathy, and adiposity scores) are shown as box and whisker plots, with median, interquartile range, and range shown. Simple linear regression analysis was used to generate regression lines, 95% CIs, P values, and r² values for Fig. 2. Multivariable regression was used for correlation of histologic and clinical parameters, with parameter estimate and P value shown for each parameter (Fig. 8A and Supplementary Figs. 6–9). Statistical analyses were performed and graphs made using GraphPad Prism or R software (version 4.2.1; R Foundation for Statistical Computing, Vienna, Austria).

Tissue section images and donor information are available at https://www.pancreatlas.org (HPAP and VPB), https://hpap.pmacs.upenn.edu (HPAP), and https://npod.org/for-investigators/data-portal/ (nPOD). All data generated or analyzed during this study are included in the published article (and its online supplementary files) or available from the corresponding author upon reasonable request. Organ procurement organizations partnering with nPOD to provide research resources are listed at https://www.jdrfnpod.org/for-partners/npod-partners/.

Similar to prior reports (11), ND pancreas weight increased through the third decade of life and plateaued thereafter (Supplementary Fig. 5). To establish the effects of age and BMI on key histologic parameters, we evaluated sections from the pancreatic head, body, and tail of 74 ND donors (Figs. 1 and 2). As shown in Fig. 2A–C, there was a broad range of histologic features in tissues from ND donors with no known pancreatic disease. ADM, thought to be a dedifferentiation of acinar cells in response to injury that lies very early on the premalignant spectrum, correlated weakly with age and BMI (P ≤ 0.1) (Fig. 2D). PanIN, a neoplastic lesion originating from ductal cells also thought to have some premalignant potential, correlated weakly with age but not BMI (Fig. 2E). In contrast, acinar atrophy did not correlate with either age or BMI (Fig. 2F) in our cohort. Tissues available for trichrome staining showed no correlation of pancreatic fibrosis with age or BMI (Fig. 2G). The angiopathy score increased with age but not BMI (Fig. 2H), and the pancreatic adipocyte score increased with both age and BMI (Fig. 2I). Adipocyte size and number both contributed to differences in adiposity (Supplementary Fig. 6). Multivariable analysis confirmed the findings of the simple linear regression and showed a significant correlation of male sex with higher total adiposity score (Supplementary Table 7).

To determine how the acinar and ductal cells change in diabetes, we evaluated ADM, PanIN, and acinar atrophy in donors with T1D and T2D and age- and sex-matched ND donors. Examples of acini and ducts with typical morphology (Fig. 3A), ADM (defined by dilatation of acini and formation of duct-like tubular structures lined with flat epithelial cells [25]) (Fig. 3B), PanIN (defined by ductal structures lined by flat or papillary epithelium with various content of mucus and cytologic atypia) (Fig. 3B), and acinar atrophy (defined by parenchymal loss with replacement by loose peri- and interlobular connective tissue [26]) (Fig. 3C) are shown. ADM was more frequent in T2D, but not in T1D, tissues compared with matched ND tissues (Fig. 3D). No difference in number of PanIN lesions, all low grade (1A or 1B) (27), was seen in T1D or T2D tissues (Fig. 3E). Acinar atrophy was more common in T1D relative to ND tissues, but not in T2D compared with ND tissues (Fig. 3F). Therefore, the two types of diabetes are associated with differing patterns of change to the acinar and ductal cells of the pancreas.

We previously demonstrated increased fibrosis in longstanding T1D that correlates with smaller pancreas size (23). To better understand if pancreatic fibrosis differs between T1D and T2D, we quantified fibrosis in T2D and further characterized the patterns of fibrosis in T1D and T2D. Using automated quantification of trichrome-positive tissue, we found that the T2D pancreata had a 58% larger trichrome-positive area relative to matched ND tissues (29.41 ± 6.3 vs. 18.6 ± 9.8%; P = 0.03) (Fig. 4A and B), similar in severity to the increase in T1D previously reported (32.4 ± 13.8 vs. 17.8 ± 4.3%; P = 0.01) (23). However, morphologic changes in the pancreas resulting from fibrosis differed substantially between T1D and T2D (Fig. 4C). Mildly fibrotic T2D pancreata frequently had focal intralobular fibrosis, whereas heavily fibrotic T2D pancreata commonly had dense intralobular fibrosis spanning multiple pancreatic lobes. In contrast, the intralobular parenchyma was relatively normal in the T1D cohort, with most fibrosis in periductal, perivascular, and interlobular spaces. Because periductal/perivascular fibrosis in the Masson trichrome stain correlated well with hyaline tissue in the H-E stain (Fig. 4D), we used H-E staining to quantify total duct and vessel areas visible at low magnification (duct–vessel). Duct–vessel area was greater in T1D compared with ND tissues, whereas there was no difference between T2D and ND tissues (Fig. 4E), providing quantitative support to the observation that the fibrosis in T1D is predominantly periductal and perivascular. Thus, although fibrosis is common in both diabetes types, the patterns of fibrosis are markedly different between T1D and T2D.

Because micro- and macrovascular changes are seen in other organs in longer-duration diabetes, we sought to determine to what extent pancreatic blood vessels are affected in T1D and T2D. Grading each tissue section using a standardized semiquantitative system based on vessel wall thickening (21) (Fig. 5A and Supplementary Fig. 3), we found that medium-sized blood vessels were significantly affected by angiopathy in both T1D and T2D compared with matched ND pancreata (Fig. 5B). In contrast, immunofluorescent staining of the capillary endothelial cells (Fig. 5C) revealed no difference in endothelial cell area (Fig. 5D) or endothelial cell activation, as characterized by HLA-DR positivity (Fig. 5E).

Because varying degrees of exocrine inflammation have been reported in T2D (28,29), we applied CODEX multiplexed imaging to evaluate the distribution and characteristics of immune cells in the exocrine tissue in T2D (Fig. 6A). We observed no difference in total immune cells (CD45⁺) or macrophages (IBA1⁺) in the T2D pancreata (Fig. 6B). CD3⁺ T cells, a majority of which were CD8⁺, were increased in T2D, although the ratio of CD4⁺, CD8⁺, and CD4⁻CD8⁻ T cells was unchanged compared with ND ( Fig. 6C). Macrophage polarization was also unchanged in T2D, with no differences in the percentage of M1-like macrophages (defined as CD206⁻ CD163⁻ HLA-DR⁺), M2-like macrophages (CD206⁺ or CD163⁺ and HLA-DR⁻), M1–M2 mixed macrophages (CD206⁺ or CD163⁺ and HLA-DR⁺), or unpolarized macrophages (CD206⁻ CD163⁻ HLA-DR⁻). Therefore, although the T2D exocrine pancreas is characterized by signs of chronic inflammation (fibrosis), there was only minimal evidence of classical cellular inflammation in the early-stage T2D cohort studied.

Whether pancreatic steatosis occurs in T2D has been an area of debate, with some imaging studies showing an increase in fat (30) and other imaging and histologic studies showing no change (11). We investigated pancreatic adipocyte distribution in ND donors and donors with T1D and T2D (Fig. 7A) using a published semiquantitative scoring method (22). Compared with matched ND donors, the T1D pancreata had fewer intralobular adipocytes but no changes in interlobular or total adipocytes (Fig. 7B). We found no difference in adiposity in T2D (Fig. 7C). We confirmed these findings by applying an automated adipocyte quantitation method (Supplementary Fig. 4). Our results support prior histologic studies suggesting that T2D is not associated with increased pancreatic fat (11), whereas the T1D pancreata had fewer intralobular adipocytes than ND pancreata. Because adipose tissue fibrosis is associated with inflammation and metabolic dysfunction (31), we asked whether pancreatic adipocyte fibrosis frequency is increased in T2D as a marker of altered metabolic phenotype. Adipocyte fibrosis was unchanged in T2D compared with ND tissues (Supplementary Fig. 7).

To investigate clinical parameters that may contribute to histologic findings, we analyzed observations from all donors using a multivariable regression model that incorporated age, BMI, sex, and diabetes status (Fig. 8A). This analysis corroborated our findings that T1D was associated with greater rates of acinar atrophy and angiopathy, whereas T2D was associated with greater rates of ADM and angiopathy. Angiopathy correlated with age among all donors, whereas pancreatic adiposity varied with age, BMI, and sex but not disease status. We also used multivariable regression to evaluate the effects of disease duration and HbA_1c at death on the various histologic parameters in T1D (Supplementary Table 8) and T2D (Supplementary Table 9). Among T1D donors, we saw no correlation of disease duration or HbA_1c with any histologic parameters used in our study. Among T2D donors, ADM and atrophy correlated positively with both disease duration and HbA_1c, whereas PanIN correlated with disease duration only.

In a large cohort of well-preserved tissues from organ donors with redacted medical information, we systematically analyzed morphologic changes in the exocrine pancreas across age, BMI, and diabetes status (results summarized in Fig. 8B). In donors with no diabetes, we report that angiopathy increased with age, and pancreatic adiposity increased with both age and BMI. ADM and PanIN were present in the pancreata of ND donors and correlated weakly with age and BMI. Male donors had greater rates of pancreatic adiposity than female donors. Compared with age- and sex-matched ND donors, the pancreata of donors with T1D had greater rates of acinar atrophy and angiopathy, whereas the T2D pancreata had greater rates of ADM and angiopathy, with no change in pancreatic adiposity. The T1D pancreata had greater incidences of periductal and perivascular fibrosis, whereas the T2D pancreata had more frequent intralobular fibrosis but only minimal evidence of classical cellular inflammation, as quantified using CODEX multiplex imaging for the first time in the exocrine pancreas. These observations in tissues from donors with and without diabetes have important implications for future mechanistic studies of intercompartment exocrine–endocrine communication, chronic pancreatic injury, and risk of pancreatic cancer in diabetes.

To facilitate the accurate description of the exocrine pancreas in diabetes, we first sought to define the histologic characteristics of pancreata from donors with no known pancreatic disease. Studying tissues from a cohort of donors across a broad range of ages and BMIs, we observed a surprisingly high frequency of pathologic findings in the exocrine pancreas of ND donors, many of which correlated with age, sex, and/or BMI (Fig. 2). Cohort size may have limited our ability to detect additional correlations. Observation of the ND pancreata provided a critical baseline for describing changes in diabetes and insight into possible pathologic mechanisms in diabetes. For example, the exocrine angiopathy that occurred in both T1D and T2D may represent an acceleration of processes that normally occur in an ND pancreas with age, whereas atrophy, which occurred in T1D but did not correlate with age in ND pancreata, likely occurs as a direct result of processes that are independent of normal aging. Our findings in the ND pancreata also emphasize the importance of appropriate donor matching for studies of the pancreas in diabetes to avoid attributing age-related changes to the disease process.

An age-matched study design allowed us to identify differences and similarities between the T1D and T2D pancreata, which provide a basis for further insights into endocrine–exocrine intercommunication. Consistent with multiple prior studies (23,32), pancreas weight was markedly lower in T1D than ND donors (Supplementary Table 3). Although imaging studies have shown a mild reduction in pancreas volume in T2D (11,33), studies of donor pancreas weight in T2D have been limited. In our cohort, we found no difference in pancreas weight between T2D and ND tissues (Supplementary Table 3). We speculate that discrepancies between our findings and imaging studies may be due to minor variability in the trimming of extrapancreatic fat and connective tissue that masks mild differences in pancreas size. Our findings of greater rates of acinar atrophy in T1D, but not in T2D, and the correlation of acinar atrophy with HbA_1c among T2D donors (Supplementary Table 9) suggest a possible role of islet–acinar communication in the maintenance of normal acinar size and structure. Presumably, the acinar tissue of all T1D donors is exposed to markedly lower insulin concentrations than that of ND donors as a result of the peripheral injection of insulin (34). Taken together with reports of reduced pancreas size in recent-onset T1D (35,36) and in a family with monogenic insulin-deficient diabetes without autoimmunity (37), we hypothesize that an islet-derived product, probably insulin but possibly other factors, such as IGF-I (38), has significant trophic effects on acinar size and structure. Loss of this trophism, as in T1D, may then result in acinar tissue atrophy and smaller total pancreas size. First-degree relatives of individuals with T1D have also been reported to have a smaller pancreas than unrelated control individuals (36). Whether this represents early subclinical impairment of islet–acinar trophism or an underlying familial predisposition to T1D associated with reduced total islet mass requires further study. In contrast to T1D, acinar tissue of T2D donors likely experiences a much broader range of insulin exposure, depending on the degree of endogenous insulin deficiency. Donors with higher HbA_1c levels likely have more severe insulin deficiency (39), which may explain the correlation of HbA_1c and acinar atrophy in T2D donors, although alternative hypotheses, including inflammatory or metabolic insults accounting for acinar atrophy in T2D, are also possible. At the same time, the possibility of bidirectional blood flow between the endocrine and exocrine compartments of the pancreas has been suggested (40). Although blood flow patterns in the human pancreas require further investigation, this raises the possibility that exocrine perturbations may influence islet function through vascular communication. Additional studies are required to understand how angiopathic changes influence islet blood flow and function in diabetes.

The presence of pancreatic fibrosis in both T1D and T2D suggests that chronic injury affects the exocrine pancreas in both conditions. We found that the fibrosis in T1D was predominantly perivascular and periductal (Fig. 4), emphasizing the importance of considering spatial context when quantifying histologic features. A recent study demonstrated that injection of bacteria into the pancreatic duct can induce insulitis in rats, supporting the hypothesis that ductal inflammation may instigate or contribute to the initial autoimmune process in T1D (41). Our data support the possibility of ductal and perivascular inflammation in T1D, although we could not determine whether such inflammation began before or after onset of autoimmunity or hyperglycemia. In contrast, the T2D pancreata had intralobular fibrosis, apparently replacing acinar tissue with dense collagenous connective tissue and ADM, although analysis of a subset of early-stage T2D donors revealed minimal cellular inflammation (Fig. 6). Larger and more in-depth studies are needed to better characterize the identity and phenotype of pancreatic stromal cells in T2D and determine the role they play in metabolic and histologic disturbances.

More tissue adiposity can be another marker of chronic injury, and visceral adiposity has been linked to T2D and insulin resistance. Imaging studies in living individuals have demonstrated increased fat content in the T2D pancreas (42–44), but no such difference was seen in autopsy studies (11,13). Consistent with these histologic studies, we did not see a quantitative difference in pancreatic adiposity in T2D, whether or not accounting for differences in BMI (Figs. 7 and 8,A). Intracellular fat accumulation in the exocrine pancreas, which our study was unable to detect, could explain discrepancies between histologic and imaging studies of pancreatic fat in T2D. Intracellular lipid droplets likely affect human islet function (45,46) and have been observed in pancreatic acinar cells (47), but the role of intracellular fat in the exocrine pancreas and how it changes in diabetes are topics that require further study.

Our finding of a weak correlation between pancreatic adiposity and BMI in ND donors (Fig. 2H) is consistent with similarly equivocal findings in imaging studies (47,48). The lack of difference in adiposity between our ND and T2D cohorts, despite an average BMI difference of 7 kg/m², is likely a result of the relatively small effect of BMI on tissue adiposity. Whether relative insulin deficiency in T2D also contributes, as is likely the explanation for lower intralobular adiposity in T1D (Fig. 7B), is not known. In addition to adipocyte quantity, our assessment of adipocyte characteristics (size and surrounding fibrosis) showed no differences in T2D compared with ND tissues (Supplementary Figs. 5 and 6). Further characterization of pancreatic adipocyte phenotype may reveal differences that our methods did not detect.

The strengths of our study include standardized analysis, large cohort size, and access to the redacted hospital records, allowing analysis of the correlation between donor characteristics and pancreas histology. To probe the poorly understood association of T2D with the risk of pancreatic ductal adenocarcinoma, we analyzed ADM and PanIN, two early premalignant lesions, in relation to diabetes duration and glycemic control. Hyperinsulinemia, hyperglycemia, and inflammation have all been proposed as possible mechanistic links between the two conditions. In mouse models of pancreatic ductal adenocarcinoma, hyperglycemia promoted cancer progression (49). In humans, a prospective cohort study of 1.3 million Korean patients identified hyperglycemia as an independent risk factor for several cancers, including pancreatic cancer (50), but randomized clinical trials have not demonstrated a reduction in cancer mortality with intensive glycemic control (51). Most cases of ADM or low-grade PanIN do not progress to pancreatic ductal adenocarcinoma, but the increased presence of these regenerative and metaplastic changes in the pancreata of donors with elevated HbA_1c levels and longer disease duration suggests that individuals with T2D and poor glycemic control face ongoing insults to the exocrine pancreas that may predispose to malignancy. Our data suggest that hyperglycemia alone is insufficient to induce these changes, because donors with longstanding T1D and elevated HbA_1c levels had no increase in ADM. Obesity, independent of T2D, is also a major risk factor for pancreatic adenocarcinoma (52). ADM and PanIN correlated weakly with BMI among ND donors (Fig. 2D and E) but not among T1D or T2D donors (Fig. 8A). Additional studies are required to understand the mechanisms responsible for pancreatic metaplasia in diabetes and obesity.

In summary, we report marked age- and BMI-associated findings in the exocrine pancreas that vary in ND, T1D, and T2D donors. Acinar metaplasia, intralobular fibrosis, and angiopathy are more prevalent in T2D tissues, whereas rates of acinar atrophy, periductal and perivascular fibrosis, and angiopathy are increased in T1D, demonstrating important differences in exocrine histopathology between these two diseases. Additional mechanistic studies are required to better understand how the observed histologic changes in T1D and T2D relate to changes in exocrine function, altered pancreas size, and pancreatic cancer risk.

Acknowledgments. The authors thank the organ donors and their families for their invaluable donations. They also thank investigators from HPAP and nPOD, the International Institute for the Advancement of Medicine, the National Disease Research Exchange, and local organ procurement organizations for their partnership in obtaining these human pancreatic tissues or sections for research.

Funding. This work was supported by the Human Islet Research Network (Research Resource Identifier [RRID] SCR_014393); HPAP (RRID: SCR_016202); grants DK106755, DK123716, DK123743, DK120456, DK104211, DK108120, DK112232, DK112217, DK123594, DK133691, DK117147, T32DK007061, and DK20593 from the National Institute of Diabetes and Digestive and Kidney Diseases and EY032442 from the National Eye Institute (to the Vanderbilt Diabetes Research and Training Center); and the Leona M. and Harry B. Helmsley Charitable Trust, JDRF, the Doris Duke Charitable Foundation, and the Center for Integrated Healthcare, U.S. Department of Veterans Affairs (grant BX000666). This research was performed with the support of nPOD (RRID: SCR_014641), a collaborative T1D research project supported by JDRF (nPOD: 5-SRA-2018-557-Q-R), and the Leona M. and 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 views of nPOD.

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

Author Contributions. J.J.W. and A.E. compiled and analyzed the data and wrote the manuscript. A.E. and A.W., as blinded clinical pathologists, reviewed and scored tissue sections. R.B., R.J., A.M.B., R.A., and D.C.S. processed pancreatic tissues and facilitated access to tissue sections. S.P. and H.K. performed biostatistical analysis. D.C.S. collected and analyzed CODEX data and assisted in figure design and writing. M.B. and A.C.P. provided research funding and access to tissues, advised on experimental design, and reviewed and edited the manuscript. A.C.P. 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. This study was presented in part at the American Diabetes Association Scientific Sessions, New Orleans, LA, 3–7 June 2022, and the Human Islet Research Network Annual Investigator Meeting, Washington, DC, 14–16 September 2022.