Involvement of Oxidative Stress–Induced DNA Damage, Endoplasmic Reticulum Stress, and Autophagy Deficits in the Decline of β-Cell Mass in Japanese Type 2 Diabetic Patients

There is globally a drastic increase in type 2 diabetic patients. The prevalence of type 2 diabetes is more conspicuous in Asian countries and will be more so in the coming 30 years compared with Western countries (1). Type 2 diabetes is characterized by hyperglycemia linked with insulin resistance in peripheral tissues and low insulin secretion of β-cells. Such features have led to a conception that there are two types of type 2 diabetes: obese and lean. The former predominates Western diabetic patients and the latter Asians (2,3). Distinction between obese and lean type of diabetes may indicate different pathogenetic processes.

β-Cell deficit is a main pathology of the islet in both obese and lean type 2 diabetes (4–6). However, the question whether it is primary cause for diabetes or secondary to the long-term metabolic abnormalities remains unanswered. In addition, the reason why the β-cell decline takes place is unknown. Increased oxidative stress, endoplasmic reticulum (ER)–induced stress, or deficit of autophagy is suspected to be the cause of premature death of β-cells in type 2 diabetes (7–11). In animal models, oxidative stress has well been confirmed to cause premature death in β-cells (7,8). On the other hand, ER stress and autophagy deficit are also found to initiate β-cell death (9–11). The information is still incomplete as to what extent or how the single factor of these variables is responsible for the β-cell death. In humans, it is not feasible to study the dynamic process of the β-cell death in spite of plethora of experimental evidence that environmental insults cause β-cell damage leading to premature death (12,13). Most of the previous studies using autopsy samples had limitations that the results were confounded by factors other than diabetes such as circulatory collapse, malnutrition, or parenteral alimentation, inevitably present in the end stage of the disease. Nevertheless, with increasing materials, collection of the data may provide important information related to the pathologic features in the genesis of this disease. In this study, we have therefore explored the relationship between β-cell pathology and oxidative stress, ER stress, and autophagy deficit in Japanese type 2 diabetic patients to elucidate its pathogenesis.

Pancreases from 47 Japanese type 2 diabetic patients and 30 nondiabetic control subjects were studied (Table 1). The subjects were limited to fresh autopsy cases conducted within 5 h of death. Cases with diabetes had a history of hyperglycemia that fulfilled the criteria of diabetes proposed by the Japan Diabetes Society (14). Diagnosis of type 2 diabetes in a diabetic group was further confirmed by the clinical record of the patients. Cases were excluded if 1) potential secondary causes of type 1 or type 2 diabetes were present, 2) patients had been exposed to chronic glucocorticoid treatment, or 3) pancreatic tissue had undergone autolysis or showed evidence of pancreatitis. Cases with a history of gastrointestinal surgery (gastrectomy) or under pregnancy or long-term treatment with antischizophrenic drugs were also excluded from the evaluation. Nondiabetic control subjects fulfilled the criteria of “normal type” without previous and family history of diabetes. The pancreas weight was measured after fine dissection of the organ from duodenum or other surrounding tissues. Clinical records were examined in each case as to the last evaluation of body weight, height for BMI, and HbA_1c. Duration of diabetes was defined as the time from diagnosis of diabetes. The data of insulin secretory capacity like serum C-peptide and the results of GAD antibody were also sought. Diabetes treatment and major accompanying disorders as well as other specific agents if any for the treatment were also carefully checked.

Formalin-fixed pancreases were each dissected into two portions of the head, body, and tail for embedding in paraffin. From our previous studies, we found that the body portions represented the islet morphometry because the head contained pancreatic polypeptide–rich islets and the tail was often confounded by marked fat infiltration. The use of paraffin blocks and study design was approved by the ethics committee of the Hirosaki University School of Medicine, and the study conforms to the provision of the Declaration of Helsinki.

Immunohistochemical staining of islet markers were processed on the paraffin blocks of the body of the pancreas, from which several consecutive cross 4-μm-thick sections were obtained. The sections were immunostained by streptavidin-biotin-peroxidase system (Nichirei Co., Tokyo, Japan) first incubated overnight at 4 °C with antibody to chromogranin-A (DakoCytomation, Glostrup, Denmark) for the identification of the islet: anti-insulin antibody (Santa Cruz Biotechnology, Santa Cruz, CA) for β-cell and antiglucagon (Dako) for α-cell. The sections were lightly counterstained with hematoxylin to count the number of nuclei of endocrine cells.

To explore the cell death, the degree of oxidative stress–related cell damage, or cell damage by ER stress, double immunostaining of anti-insulin with terminal deoxynucleotidyl TUNEL of an ApopTag (Millipore, Bellerica, MA), immunoreactive γH2AX (Ser139; Clone JBW301; Millipore) (15), and CCAAT/enhancer 1 binding protein-β (C/EBP-β; EPR3358; Epitomics, Burlingame, CA) was performed, respectively (10,16). To evaluate the implication of autophagy deficit, the expression of immunoreactive P62 (F-12; Santa Cruz) (17) were examined. To evaluate proliferation rate of β- and α-cells, double immunostaining of Ki67 (MIB1; Dako) and insulin or glucagon was also performed. Choice of antibodies was determined by high specificity and reproducibility of staining results in the literature and our previous studies (5,18). For double immunostaining, sections were first incubated with TUNEL agents, or antibodies against Ki67, γH2AX, and C/EBP-β, and then antibody to insulin or glucagon. Immunostaining of P62 was applied to single sandwich reactions. Following incubation with the last antibody conjugated with peroxidase or alkaline phosphatase, the reaction products of the enzyme were visualized with diaminobenzidine or with a fuchsin staining kit (Nichirei).

For determination of the fractional area of the islet (Vi), β-cell (Vβ), and α-cell (Vα) relative to total parenchymal pancreatic area (including endocrine and exocrine pancreas and stroma) defined as volume densities, the pancreatic sections immunostained with chromogranin-A or double immunostained for insulin and glucagon were applied for the measurement by point counting method using ImageJ system (version 1.56, National Institutes of Health, Bethesda, MD), as previously described (4,18). For the measurement of Vi, we captured the image of the whole pancreatic sections stained with chromogranin-A, comprising more than 150 serial frames as JPEG file at ×10 magnification (×5 objective). The area of the islet and pancreas parenchyma on each section was obtained by counting the number of crossing points that hit the islet and pancreas on the overlaid grid (625 points) on the sections. The number of points hit on the islet was divided by the total number of points hit on the pancreatic parenchymal area, thus yielding the value of Vi. We counted ∼4,500–5,000 islets in each subject.

Vβ and Vα were obtained from the measurement on the double immunohistochemistry sections of insulin and glucagon at high magnification (×40). More than 150 islets were captured digitally as JPEG files. The images were overlaid on a grid consisting of 2,500 points. Vβ and Vα were determined by dividing the total number of points hit on insulin- or glucagon-positive cells by the total number of points hit on the pancreatic parenchyma, respectively. β-Cell or α-cell occupancy (percentage) within islet was obtained in each subject by division of Vβ or Vα by Vi. The mass of the islet (Mi), β-cell (Mβ), and α-cell (Mα) was determined by multiplying Vi, Vβ, and Vα by pancreas weight, respectively.

The population of the islet (density of islet) per unit area was obtained by counting of the number of chromogranin-A positive area composed of more than three cells. Average islet size and β-cell size were obtained by division of the total islet area or total β-cell area by the number of islet or the number of β-cells per unit area, respectively.

To determine β- and α-cell growth, double-positive cells for insulin or glucagon and Ki67 (among over 2,000 β-cells and 1,000 α-cells) in all the islets were examined, and the number of Ki67-positive cells expressed as percentage of total β- or α-cells was evaluated. For determination of apoptotic cells, we used slides double-stained for insulin and TUNEL. Since apoptotic cells were rarely found, we did not quantify the number, because of unfeasibility to statistical analysis. Intensities of immunoreactions of γH2AX, C/EBP-β, and P62 were evaluated by comparison with positive control specimens of lymph nodes showing many apoptotic cells (19) for oxidative stress–related DNA damage (γH2AX), pancreas with neutrophilic infiltrations (20) for ER stress, and alcoholic liver with Mallory bodies (21) for autophagy deficits. The reaction of the cells was defined as positive when the nuclei or cytoplasms were clearly stained brown compared with negative background of the exocrine acinar cells. The percentage of positive cells was calculated by evaluating more than 1,000 β-cells in each subject, and comparison between diabetic and nondiabetic subjects were made. These analyses on the immunostained sections were conducted on at least 50 islets in sizes larger than ∼4,000 μm² in a random manner.

Data are presented as mean ± SE. Statistical comparisons on the mean values between control and diabetic groups were carried out using the nonparametric Mann–Whitney U test, with a P value of <0.05 taken as significant (StatView, version 5.0.1, Mountain View, CA). A simple regression was carried out for the correlation analysis. The Wilcoxon rank sum test was performed to compare endocrine cell replication between groups due to skewed distributions of the observations. To evaluate the implication of cell damage markers in the β-cell deficit, we adopted hierarchical “cluster analysis,” which enables grouping (clustering) of subjects with similar (but not identical) values of staining results of γH2AX, C/EBP-β, and P62 and the relationships with β-cell deficit based on the Ward’s method (22) using a software (JMP, version 10.0.2, SAS institute Inc., Cary, NC).

Clinical summary of 47 type 2 diabetic patients and 30 nondiabetic age-matched (control) subjects is summarized in Table 1. Mean age was comparable in two groups. BMI was 22.4 ± 0.7 in the nondiabetic group and 22.7 ± 0.6 in the diabetic group (P > 0.1). Duration of diabetes in the diabetic group was 10.8 ± 1.5 years. HbA_1c value in diabetic patients nearest to their death was 7.6 ± 0.3% (60 ± 3.1 mmol/mol), while Japanese healthy control values were reported to be 5.49 ± 0.55% (37 ± 5.6 mmol/mol) (23). There was no difference in pancreas weight between nondiabetic and diabetic groups. Cardiovascular events were predominant causes of death in diabetic group. Detailed information on the age, BMI, pancreas weight, diabetes duration, and HbA_1c values is described in Supplementary Table 1. The causes of death are separately described in Supplementary Table 2. Although there were only limited patients who had the data of GAD antibody in diabetic patients, all the examined cases were negative.

As shown in Table 1, Vi was not different between diabetic and nondiabetic groups with insignificant decrease (13%) in the mean value in the diabetic group. Vβ in the diabetic group (1.02 ± 0.08%) was 31% less than that in nondiabetic subjects (1.47 ± 0.17%; P < 0.01). Similarly, Mβ in the diabetic group (1.27 ± 0.11 g) was significantly (32%) smaller than that in nondiabetic group (1.86 ± 0.24 g; P < 0.05). In contrast, there was significant increase in Vα (31%) in the diabetic group (0.67 ± 0.05%) compared with that in nondiabetic subjects (0.51 ± 0.06%; P < 0.05). The diabetic group also showed an increased Mα (37%; 0.84 ± 0.07 g) compared with the nondiabetic group (0.61 ± 0.07 g; P < 0.05). Vβ was in parallel with Vα in either the nondiabetic or the diabetic group, in which the increase in Vα was marked in subjects with reduced Vβ (Supplementary Fig. 1). β-Cell occupancy inversely correlated with α-cell occupancy in either the nondiabetic or the diabetic group, in which the increased α-cell occupancy was marked. Subdivision of diabetic patients into hypertensive, atherosclerotic, or cardioischemic groups did not show any specific tendency for β-cell loss or α-cell changes. There were no significant differences in the mean cellular sizes of β- and α-cell or in the population (cell number) of β- and α-cell per unit pancreatic area between diabetic and nondiabetic groups.

Population of small neogenetic islets composed of less than four endocrine cells was significantly greater in the diabetic group than that in the nondiabetic group (P < 0.001).

Double immunostaining of Ki67 and insulin disclosed very low positive reactions in β-cells, yielding only 0.41 ± 0.25% in the diabetic group and 0.32 ± 0.12% in nondiabetic subjects. There was no significant difference in the mean Ki67 positivity of β-cell between the diabetic and nondiabetic groups. Double immunostaining of Ki67 and glucagon revealed that the frequency of Ki67-positive α-cells was 0.43 ± 0.13% in the diabetic group and 0.44 ± 0.16% in the nondiabetic group. There was no difference in their frequencies between these two groups. There were no obvious islet cells positive for TUNEL in either the nondiabetic or the diabetic group.

Although there was a trend toward a decrease in Vi and Vβ with increasing duration of diabetes, there was no significant correlation between these two parameters (Fig. 1A). Age did not significantly influence on the values of Vi, Vβ, and Vα (data not shown). In contrast, there was a significant inverse correlation of Vi and Vβ, but not Vα, with the levels of HbA_1c in diabetic patients, indicating that the higher the HbA_1c, the greater the loss of β-cells (Fig. 1B). BMI did not show any correlation with Vi, Vβ, and Vα (Fig. 1C). There was a trend of correlation (P = 0.08) though not statistically significant between duration of diabetes and Mβ, but not Mi or Μα (Supplementary Fig. 2). Mi and Mβ, but not Mα, correlated with HbA_1c levels (P < 0.05 for both). BMI did not show any impact on Mi, Mβ, and Mα in either the nondiabetic or the diabetic group. There was no apparent correlation between serum C-peptide levels and Vβ or Mβ in the diabetic group, although the data were available in only limited cases.

The β-cell damage was evaluated by immunohistochemistry of γH2AX for oxidative stress–related DNA damage, C/EBP-β for ER stress, and P62 for autophagy deficit (Fig. 2A). Positive controls showed strong reactions of γH2AX in apoptotic cells in the lymph node, C/EBP-β in neutrophils in the lesion of acute pancreatitis, and P62 in Mallory bodies in hepatocytes of alcoholic liver cirrhosis. In the islets of pancreases obtained from diabetic patients, immunostaining of γH2AX and C/EBP-β often revealed clearly positive reactions in the nuclei of damaged β-cells in contrast to weak or negative background nuclei of acinar cells or other stromal cells. Such reactions were infrequent in the islets from nondiabetic subjects. In contrast, P62 reactions were often detected in the cytoplasm of endocrine cells in the islets from diabetic cases, while nondiabetic subjects rarely showed positive P62 reactions.

Population of positive cells for γH2AX was 0.47 ± 0.29% on average in nondiabetic controls, while it was 3.91 ± 1.27% in the diabetic group, significantly greater in the latter (P < 0.02) (Fig. 2B). There was an inverse correlation between γH2AX and Vβ (r = 0.41; P < 0.05) or Mβ in the diabetic group (r = 0.49; P < 0.05), while it was not the case in nondiabetic controls (Fig. 2C). Mean values of C/EBP-β-positive cells were 0.93 ± 0.59% in the nondiabetic group and 7.71 ± 2.69% in the diabetic group, and the difference was significant (P < 0.05). Some diabetic cases showed high frequencies of P62-positive cells compared with nondiabetic ones, and there was a significant difference in P62 positivity between the diabetic group (3.75 ± 1.53%) and the nondiabetic group (0.72 ± 0.60%; P < 0.05). There was no significant correlation between C/EBP-β and Vβ or Mβ or between P62 and Vβ or Mβ.

To see the implication of γH2AX, C/EBP-β, and P62 expressions in the β-cell decline, hierarchical cluster analysis on the signature of these factors was conducted and sought to see the relationship to β-cell decline. There was an expression profile yielding three distinct groups (groups 1, 2, and 3) (Fig. 3A). Group 1 consisted of subjects who showed mostly high scores for all three factors or γH2AX and P62, and group 2 showed low or modest scores for all three factors. When the mean value of Vβ was compared among the groups, there was a significant reduction of Vβ in group 1 compared with nondiabetic subjects (P < 0.01) and group 2 (P < 0.05), in which Vβ was not significantly reduced compared with nondiabetic control (P > 0.10) (Fig. 3B). Vβ in group 3, which showed high scores of C/EBP-β but low or modest for other two factors, was significantly smaller than that in the nondiabetic group (P < 0.05). The difference in Vβ between group 2 and group 3, however, was not significant.

Diabetic patients treated with oral diabetic drug and insulin therapy exhibited significant decreases in both Vβ and Mβ compared with a group in diet treatment (Supplementary Fig. 3A). In addition, there was a trend toward an increase in the expression levels of γH2AX in a group in either oral diabetic drug or insulin therapy compared with a diet group (Supplementary Fig. 3B).

In this study, we confirmed our previous data that β-cells are selectively depleted in Japanese type 2 diabetic patients (4). There was no obvious difference in the population of proliferating cells identified by Ki67 between diabetic and nondiabetic groups, and we could not detect appreciable levels of apoptotic β-cells in diabetic subjects. Exploration for variables that may affect the morphometric data disclosed strong association of the β-cell loss with HbA_1c, but duration of diabetes, BMI, or age did not show significant impact on the data. Concurrently, β-cell loss was associated with accumulation of multiple indices of cellular damage from oxidative stress, ER stress, and autophagy deficits. Despite numerous data that indicate various pathogenetic mechanisms in type 2 diabetes animal models (7–11), our findings for the first time indicate the combined implication of oxidative stress, ER stress, and autophagy deficit in the pathogenesis of type 2 diabetes in humans.

Significant correlation of β-cell decline with HbA_1c in this study is of particular importance in view of the implication of chronic hyperglycemia in the progressive decline of β-cells. It is not clear at present, however, whether high concentrations of HbA_1c are the results of β-cell decline or long-standing hyperglycemia per se elicits enhanced β-cell death. The findings of accumulated γH2AX in remaining β-cells possibly indicates that cellular injury caused by hyperglycemia may play a role in the β-cells loss since γH2AX reflects nuclear DNA damage as well as organelle damage exposed to oxidative stress (15,19). Our previous results of increased 8-hydroxydeoxyguanosine expression that we found in type 2 diabetic islets are in accord with this contention (4,24,25).

The modest reduction of β-cell mass in this study (∼30%) compared with those in Americans (5) (∼50–60%) may be ascribed to the lack of influences of increased BMI on Vi and Vβ in Japanese nondiabetic subjects (26,27). Mean BMI was 22 in Japanese lean diabetic subjects while it was 35 in Americans (5). In contrast to the marked increase in β-cells in obese European or American nondiabetic subjects (28), compensatory hyperplasia in obese Japanese was not apparent (26,27). Such a modest reduction of β-cells may not itself account for the poor insulin secretion encountered in diabetes, underscoring the importance of β-cell dysfunction in type 2 diabetes rather than β-cell loss (2,6,29). It has been shown that oxidative stress elicits reduced pancreatic duodenal homeobox-1 or insulin gene expression, resulting in impaired insulin secretion in animal models and human type 2 diabetes (7,30). Consistent with this contention, our results indicate that, even in the modest decline of β-cell mass, the extent of cell damage by oxidative stress may determine the capacity of insulin secretion.

In this study, there was an increase in α-cell mass in diabetic patients, but this issue is still controversial in previous studies (31). Regulation of α-cell mass is much less explored in diabetic islets, and the question remains whether changes we found are adaptive processes to β-cell disappearance or specific processes may operate to promote α-cell proliferation (32). One may consider that loss of paracrine factors released from depleted β-cells may account for the increase in α-cells. This contention may be supported by the inverse correlation between α- and β-cell occupancy as shown in this study (Supplementary Fig. 1). It is interesting to note that incretin treatment promoted proliferation of α-cells in human type 2 diabetic patients (33). Unfortunately, we could not obtain any meaningful association of α-cell increase with clinical parameters. Future studies are needed to clarify the process of α-cell changes.

Validity of the markers for ER stress and autophagy deficits in diabetic islets was confirmed by the reactions to nuclei of neutrophils and Mallory bodies on the same slides, and positive β-cells were much fewer than expected, thus indicating low sensitivity but high specificity of these markers for cell injury. Despite abundance of the data to support the importance of ER stress and autophagy deficits in the β-cell pathology in animal models of type 2 diabetes (9–11), such significance was not sufficiently confirmed in human diabetic patients. For example, ER stress was shown to be typical in Wolfram syndrome (34), but C/EBP homologous protein expressions were not necessarily enhanced in type 2 human diabetic islets as shown in this study (35). On the other hand, animal models for autophagy deficits recapitulate β-cell reduction with impaired glucose tolerance (36), but only a fraction of patients with type 2 diabetes showed high scores for P62 in this study. These results may indicate possible involvement of ER stress and autophagy deficits in β-cell decline in type 2 diabetes. We should be cautious, however, of interpretation of the results because of our limited survey on a variety of markers for the cell damage related to ER stress and autophagy deficits.

Ki67-positive cells of islet endocrine cells were extremely low in adult pancreatic islets in either diabetic or nondiabetic subjects, and there was no difference in these two groups. Since apoptosis cells were rarely found in all the subjects investigated in this study, the decline of β-cells may be relevant to slow and gradual disappearance of senescent β-cells undergoing cell death, not to low replicating activity of β-cells. In fact, it is shown that life expectancy of β-cells is extremely long (nearly 20 years) (37), and the aged cells are sensitive to cell death (38). Accumulation of γH2AX-positive cells in the aged pancreas underscores this contention (26) since γH2AX was shown to be indicative for preapoptotic state (18,19). The lack of apparent evidence of apoptotic cells may not indicate the real absence of apoptotic cells, which are cleared by macrophages too rapidly to be appreciated on the sections.

Obviously, our study has a number of limitations for interpretation. We are lacking ample clinical data related to the diabetes condition except for most recent values of HbA_1c, duration of diabetes, and BMI. Much more information is required such as glucose control, serum concentrations of insulin or lipids, and detailed past history or family history of diabetes. Such information would extremely be valuable for the connection of the structural data to the function of the islet. One may wonder if the background disorders such as the presence of malignancy or inflammatory diseases may have affected the data. We believe, however, that comparison of the data with or without such cases did not significantly change the fundamental results shown in this study. Very recently, a single genetic factor, such as TCF7L2 or KCNQ1, was suggested to correspond with irregular islet contour (39,40). Although we did not specifically look into the shape of islet, further investigations may be warranted to confirm the implication of altered islet architecture in the pathogenesis of diabetes. There may also need to confirm the alterations of cell damage markers by detailed molecular analyses on the isolated islets. Currently, our methodology is not complete enough to analyze the molecules, due to difficulty to obtain ample nucleic acids or proteins from fixed human autopsy samples. Development of new techniques or accumulation of fresh samples may be expected to solve this issue in future.

Acknowledgments. Technical assistance from Saori Ogasawara, Misato Sakamoto, and Hiroko Mori (Department of Pathology and Molecular Medicine, Hirosaki University Graduate School of Medicine) is highly appreciated. The authors also appreciate valuable suggestions for and help with the staining of γH2AX from Ken Kurose and Naoya Kumagai (Department of Diagnostic Pathology, Hirosaki University Hospital).

Funding. This study was supported by a grant in aid of the Japanese Ministry of Education, Culture, Sports, Science and Technology; the Japanese Ministry of Health, Labour and Welfare; and the Takeda Science Foundation.

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

Author Contributions. H.M. and S.Y. designed the study, conducted the study, interpreted and discussed the results, and wrote the manuscript. K.Ta., W.I., K.Ts., S.O., and T.Y. conducted the study and interpreted and discussed the results. S.Y. 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 at the 73rd Scientific Sessions of the American Diabetes Association, Chicago, IL, 21–25 June 2013.