Effects of Glucocorticoid Treatment on β- and α-Cell Mass in Japanese Adults With and Without Diabetes

Type 1 diabetes (T1DM) and type 2 diabetes (T2DM) are both characterized by a deficit of β-cell mass (BCM) (1,2). Preservation or recovery of BCM is therefore an important therapeutic strategy for both T1DM and T2DM. However, the regenerative capacity of BCM in humans remains largely unknown.

In rodents, β-cells have been shown to be able to adaptively increase in response to an increased insulin demand such as obesity or pregnancy (3–5). However, β-cell proliferation in humans has been reported to rapidly decrease within 5 years after birth, and only minimal β-cell proliferation is observed in adults (6–8). Estimation of β-cell life span by lipofuscin accumulation or radiocarbon dating has also suggested minimal β-cell turnover in adult humans (9,10). Therefore, clarification of endogenous regenerative capacity in adult humans is critical for interpretation of the results of rodent studies and their application to humans.

It has been reported that BCM is increased by 20 to 50% in obese adult humans without diabetes (8,11), to a smaller degree than in rodents, which usually show a two- to threefold increase (4,5), consistent with lower β-cell turnover in adult humans. Recently, we have also reported that in the Japanese population, no significant increase in BCM was observed in obese adults without diabetes (12). These findings further underscore the limited capacity of BCM expansion in adult humans.

Glucocorticoids (GCs), such as prednisolone and dexamethasone, also generally called “steroids,” are potent anti-inflammatory agents that are commonly used to treat a broad range of inflammatory and autoimmune conditions (13). GCs are known to increase insulin resistance by facilitating hepatic glucose production and reducing peripheral glucose disposal (14–16). As a result, the use of GCs is associated with a risk of development or worsening of glucose intolerance, which is well known as one of the major adverse effects of GC therapy (17). In rodents, GC-induced insulin resistance was shown to increase insulin biosynthesis and secretion and promote BCM expansion through β-cell proliferation (18–20). GC administration in humans has also been shown to augment insulin secretion (21–23); however, the effect of GC therapy on BCM in humans is unknown.

Therefore, in this study, to gain more insight into β-cell regenerative capacity and the pathophysiological role of BCM in GC-induced glucose intolerance in humans, we sought to address the following questions:

• 1) Does BCM adaptively increase to compensate GC-induced insulin resistance?

• 2) Does BCM decrease in individuals with GC-induced diabetes (i.e., steroid diabetes), as well as T1DM and T2DM?

• 3) Is there any relationship between glucose intolerance and BCM in individuals with GC treatment?

The Keio University School of Medicine Review Board approved this study.

Specimens of pancreas obtained at autopsy were obtained with the permission of the bereaved families. Potential case subjects were first identified by retrospective analysis of the Keio University autopsy database. To be included, case subjects were required to have 1) been aged 40 to 79 years, 2) a full autopsy within 24 h of death, 3) medical information before death, 4) no history of pancreatitis, pancreatic tumor, or pancreatic surgery, and 5) stored pancreatic tissue that was of adequate size and quality. Case subjects were excluded if pancreatic tissue had undergone autolysis.

We reviewed ∼1,000 autopsy case subjects between 2000 and 2013, and found 49 case subjects that had received GC therapy with long-term (e.g., prednisolone 30 mg/day for 2 years) or short-term/intermittent (e.g., 3 days of pulse methylprednisolone) use before death. We only included case subjects that had received GC administration at least within 3 months before death. Twenty-three case subjects had received long-term GC therapy and 26 case subjects short-term/intermittent GC therapy (Supplementary Table 1). The case subjects were classified into three groups: 1) those without diabetes (GC-NDM), 2) those with GC-induced diabetes (GC-DM), and 3) those diagnosed with T2DM before GC therapy (DM2+GC). We also obtained samples from 51 age- and sex-matched case subjects with or without T2DM who had not received GC therapy before death as control groups (DM2 and NDM, respectively). All were Japanese. Mean time from death to autopsy was 7.5 ± 5.5 h. Samples for 95 specimens were from the body or tail of the pancreas, and 5 specimens were from the head of the pancreas. In addition, we were able to obtain glycosylated hemoglobin (HbA_1c) level within 1 year before death in 83 case subjects (NDM, 18; GC-NDM, 23; GC-DM, 9; DM2, 20; DM2+GC, 13). The HbA_1c values were expressed as the National Glycohemoglobin Standardization Program (NGSP) value (24).

The pancreas was fixed in formaldehyde at autopsy, embedded in paraffin for subsequent analysis, and 5-μm sections were stained for light microscopy as follows: 1) with hematoxylin-eosin, 2) for insulin (peroxidase staining) with hematoxylin, 3) for glucagon with hematoxylin, 4) for insulin and Ki67 for assessment of β-cell replication, and 5) for insulin and single-stranded DNA (ssDNA) for assessment of β-cell apoptosis, as previously described (12,25). For immunohistochemical staining, guinea pig polyclonal antibodies against porcine insulin and rabbit polyclonal antibodies against human glucagon were used (DAKO Japan, Kyoto, Japan). Murine monoclonal antibodies against human Ki67 (DAKO Japan) and rabbit polyclonal antibodies against ssDNA (IBL, Takasaki, Japan) were used for the detection of proliferating cells and apoptotic cells, respectively.

To quantify fractional β-cell area (%BCA), the entire pancreatic section was imaged at original magnification ×200 (×20 objective) using a Mirax Scan and Mirax Viewer (Carl Zeiss MicroImaging GmbH, Goettingen, Germany). The ratio of BCA to total pancreas area was digitally measured using Image Pro Plus software (Media Cybernetics, Silver Springs, MD), as previously reported (12,25). Interlobular connective tissue, large blood vessels, and adipocytes were excluded from the total pancreas area; thus, the total pancreas area consisted to the greatest extent of pancreatic acinar tissue and pancreatic islets. Likewise, the ratio of α-cell area to total pancreas area (%ACA) was also digitally measured. All measurements were conducted by a single investigator (S.S.), and the intraobserver coefficient of variance (computed in five case subjects studied on five occasions) was 7%. All measurements were conducted twice, and the mean of the two measurements was used. At the time of measurement, the investigator was blinded to GC use and the glucose metabolism status of each specimen.

To conduct further morphometric analysis, islet size and density, scattered β-cells and insulin-positive duct cells were quantified in randomly selected areas of the pancreas that contained more than 100 islets in each case (112 ± 16 islets per case, 10,619 islets total) using a Mirax Viewer. Scattered β-cells were defined as a cluster of three or fewer β-cells in acinar tissue, and the density of scattered β-cells was determined as the number of scattered β-cells/pancreas area (/mm²). Likewise, the density of islets (/mm²) and islet size (μm²) were also determined in the same area. Insulin-positive duct cells were also counted and expressed as the number of insulin-positive duct cells/pancreas area (/mm²). In addition, β-cell replication and apoptosis were quantified in total pancreas sections, and the frequencies of β-cell replication and apoptosis were expressed as the percentage of islets. Assuming a difference in %BCA among the groups, these values were further adjusted for %BCA, as previously described (2). A total of 49,450 islets (610 ± 28 islets per case) were assessed for these analyses.

Data are presented as mean ± SD in the text and tables. Data with a non-normal distribution are expressed as median (interquartile range). The Mann-Whitney U test was used to analyze the difference between the groups, and the Spearman correlation coefficient was used to assess the correlation between two parameters. For multivariate regression analysis, nonnormally distributed data were log-transformed. Statistical analyses were performed using SPSS 22 software (IBM, Armonk, NY). A P value of <0.05 was considered to be significant for all analyses.

The characteristics of the case subjects are summarized in Table 1, with causes of death in Supplementary Table 1. Mean BMI in each group was 19–22 kg/m², and the mean BMI in GC-DM was slightly but significantly lower than that in DM2. Mean HbA_1c in GC-DM, DM2, and DM2+GC was 7.0–7.5%, which was significantly higher than that in NDM and GC-NDM.

There was no significant difference in %BCA between NDM and GC-NDM (1.66 ± 1.05% vs. 1.21 ± 0.59%, P = 0.13; Fig. 1A). The %BCA in DM2 was significantly lower than that in NDM and GC-NDM (0.92 ± 0.63%, P = 0.01 and P = 0.03 vs. NDM and GC-NDM, respectively). However, the difference in %BCA in GC-DM was not significant compared with NDM or GC-NDM (1.34 ± 0.53%, P = 0.55 and P = 0.55, vs. NDM and GC-NDM, respectively), and the %BCA in GC-DM was significantly higher than that in DM2 (P = 0.03). The %BCA in DM2+GC was not significantly different from that in DM2 (0.73 ± 0.43%, P = 0.35) but was significantly lower than that in NDM, GC-NDM, and GC-DM (P = 0.01, P = 0.02, and P = 0.01, respectively).

No significant correlation was found between %BCA and total days or total dose of GC in GC-NDM, GC-DM, or DM2+GC group (Fig. 2A–F), although the correlation between %BCA and total days of GC use was almost significant in GC-NDM (R = 0.40, P = 0.06; Fig. 2A). When these case subjects were classified according to short-term/intermittent or long-term use of GCs, these results were largely unchanged (Fig. 2).

In line with our prior study (12), there was no significant correlation between %BCA and BMI in either total or each group of subjects, and the relationships between %BCA and total days or total dose of GC use were unchanged after adjustment for BMI (data not shown).

Similarly to %BCA, there was no significant difference in %ACA between NDM and GC-NDM (1.07 ± 0.68% vs. 0.88 ± 0.52%, P = 0.36; Fig. 1B). The %ACA in DM2 was significantly decreased compared with NDM (0.64 ± 0.43%, P = 0.02). There was no significant difference in %ACA in GC-DM compared with NDM or GC-NDM (0.82 ± 0.54%, P = 0.24 and P = 0.50 vs. NDM and GC-NDM, respectively). %ACA in DM2+GC was also significantly lower than that in NDM or GC-NDM (0.45 ± 0.29%, P = 0.01 and P = 0.01, respectively).

A significant positive correlation was found between %ACA and total dose of GC in GC-NDM (R = 0.52, P = 0.03; Fig. 3) and between %ACA and total days of GC therapy in total subjects (i.e., GC-NDM, GC-DM, and DM2+GC combined; R = 0.34, P = 0.02), although there was no significant correlation between %ACA and total days or total dose of GC treatment in the GC-DM and DM2+GC groups. These correlations were unchanged after adjustment for BMI (data not shown).

The α-cell–to–β-cell ratio (%ACA-to-%BCA ratio) did not differ significantly among the groups (Fig. 1C), and no significant correlation was found between the α-cell–to–β-cell ratio and total days or dose of GC treatment (Fig. 4), which, again, was unchanged after adjustment for BMI. There was a significant positive correlation between %BCA and %ACA in case subjects with and without GC therapy (R = 0.70 and R = 0.59, both P = 0.0001; Fig. 5).

Mean islet size and islet density were not significantly different between NDM and GC-NDM (6,631 ± 2,985 vs. 5,970 ± 2,656 μm² [P = 0.48] and 6.11 ± 3.15 vs. 5.53 ± 2.06/mm² [P = 0.70], respectively) or GC-DM (5,211 ± 890 μm² and 4.99 ± 2.44 /mm², P = 0.66 and P = 0.29, respectively; Fig. 6A and B). Although the difference in islet density was not significant, mean islet size was significantly decreased in DM2 and DM2+GC compared with NDM, GC-NDM, and GC-DM (all P < 0.05), consistent with the reduced %BCA and %ACA in DM2 and DM2+GC.

On one hand, there was no significant difference in the density of scattered β-cells among the groups (Fig. 6C). On the other hand, the density of insulin-positive duct cells was significantly higher in GC-NDM and DM2 compared with NDM (0.01 ± 0.03 vs. 0.12 ± 0.16 /mm² and 0.05 ± 0.06 /mm², P = 0.001 and P = 0.002, respectively, Fig. 6D). No significant correlation was found between the density of insulin-positive duct cells and total days or total dose of GC use (Supplementary Fig. 1).

There was no significant difference in β-cell replication (i.e., Ki67-positive β-cells) among the groups (Fig. 6E). It is of note that β-cell replication was not significantly correlated with BMI in either group (data not shown). Only a few β-cells showing apoptosis (i.e., ssDNA-positive β-cells) were found in DM2, but they were not found in the other groups (Fig. 6F).

There was a significant negative correlation between %BCA and HbA_1c in total case subjects (R = −0.34, P = 0.001; Fig. 7E). When the case subjects were stratified by the presence or absence of GC therapy, a significant negative correlation was found between %BCA and HbA_1c in case subjects without GC therapy (i.e., NDM and DM2 groups combined, R = −0.45, P = 0.004; Fig. 7A). However, the correlation between %BCA and HbA_1c was attenuated in case subjects who had received GC therapy (i.e., GC-NDM, GC-DM, and DM2+GC groups combined, R = −0.27, P = 0.08; Fig. 7C).

Intriguingly, a significant negative correlation was also observed between %ACA and HbA_1c in total case subjects (R = −0.36, P = 0.001; Fig. 7F). The negative correlation between %ACA and HbA_1c remained significant in case subjects both with (R = −0.44, P = 0.006) and without GC therapy (R = −0.36, P = 0.02; Fig. 7B and D). These results were unchanged even after excluding one NDM subject with higher %BCA (4.9%) and %ACA (2.9%).

In this study, by examining autopsy pancreas, we report that:

• 1) There was no significant increase in BCM in individuals without diabetes who received GC therapy.

• 2) There was no significant reduction in BCM in individuals with GC-induced diabetes, whereas BCM in individuals with T2DM with or without GC therapy was significantly decreased compared with that in individuals without diabetes.

• 3) In individuals who received GC therapy, the correlation between BCM and degree of hyperglycemia (i.e., HbA_1c) was attenuated compared with that in those without GC therapy.

• 4) There was no significant increase in ACM in individuals who received GC therapy.

GCs, such as prednisolone and dexamethasone, are well known to have diabetogenic effects by facilitating hepatic glucose production and reducing peripheral glucose disposal (14–16). In healthy humans, it has been reported that insulin secretion is two- to fourfold increased to compensate increased insulin resistance after short-term GC administration (21–23). In rodents, GC administration has been shown to promote an increase in BCM and in insulin secretion through an increase in β-cell replication and β-cell neogenesis (18–20). On the basis of these previous studies, we hypothesized that BCM might also be increased in humans who received GC therapy.

In this study, however, we found no significant difference in %BCA in Japanese individuals without diabetes who had received GC therapy prior to death compared with age-, sex- and BMI-matched individuals without diabetes who had not received GC therapy prior to death. There was also no significant correlation between %BCA and total days or total dose of GC administration, although a higher amount of GC increases the risk of development of diabetes (26). These results were unchanged after adjustment for BMI. We also observed no significant difference in islet size or islet density between individuals without diabetes who received or did not receive GC administration. Thus, our findings suggest that in adult humans, in contrast to rodents, there is little adaptive increase in BCM in response to GC-induced insulin resistance.

Notably, we found no significant decrease in %BCA in individuals with GC-induced diabetes, also called “steroid diabetes,” compared with individuals without diabetes. β-Cell dysfunction is a hallmark of both T1DM and T2DM (27). It has been reported that BCM is decreased by >90% and by 30% to 65% in patients with T1DM (1) and T2DM (2,11,28–31), respectively, suggesting that deficits of both β-cell function and BCM, collectively called “β-cell functional mass,” are a core pathogenetic feature of diabetes. Indeed, in this study we confirmed that %BCA was decreased by ∼45% in individuals with T2DM compared with age- and BMI-matched control subjects without diabetes, consistent with other Japanese studies (30,31). Furthermore, we found a significant negative correlation between %BCA and HbA_1c level in individuals with and without T2DM, in line with the previous observation (30), whereas this correlation was attenuated in individuals who received GC therapy. These findings suggest that, unlike T1DM and T2DM, the development of GC-induced diabetes is mainly associated with insulin resistance and/or β-cell dysfunction, but not necessarily a deficit of BCM. This may explain why patients with GC-induced diabetes often achieve complete remission after withdrawal of GC administration (21), whereas patients with T2DM rarely achieve its remission (27).

Although it has been reported that insulin secretion is increased after a single dose or short-term GC administration in healthy humans, β-cell function assessed by disposition index has been reported to be unchanged or even impaired in susceptible populations such as first-degree relatives of patients with T2DM and obese individuals (21–23,32), suggesting that GC administration not only induces insulin resistance but also impairs β-cell function. In vitro studies in rodent islets have shown that GCs acutely impair the insulin secretory pathway by reducing the uptake and oxidation of glucose, augmenting outward potassium currents, and interfering with the parasympathetic nervous system (17). It has also been reported that GCs augment endoplasmic reticulum stress and induce β-cell apoptosis (33). However, we did not observe an increase in β-cell apoptosis in individuals who received GC therapy.

We did not observe any significant change in mean islet size and islet density in individuals who received GC therapy, consistent with no significant change in %BCA in those individuals. We also observed no difference in β-cell replication and apoptosis between individuals with and without GC therapy, suggesting little change in β-cell turnover after GC therapy, although it is possible that we were not able to detect a significant difference among the groups because of the limited number of Ki67- and ssDNA-positive β-cells under the autopsy condition (34,35). Scattered β-cells and insulin-positive duct cells are considered to be a surrogate marker of β-cell neogenesis (2,36). A recent study reported that β-cell neogenesis was increased in patients with impaired glucose tolerance (IGT) and new-onset T2DM (31), suggesting that β-cell neogenesis is compensatorily increased during the development of diabetes. In this study, we observed a significant increase in insulin-positive duct cells in individuals without diabetes who received GC therapy as well as in individuals with T2DM compared with control subjects without diabetes, suggesting that β-cell neogenesis may compensatorily increase under GC-induced insulin resistance. However, we stress that this compensatory increase in insulin-positive duct cells did not result in an increase in %BCA or islet density, indicating that the compensatory increase in β-cell neogenesis is not sufficient for BCM expansion in humans.

Lastly, we found no significant increase in %ACA in individuals who received GC therapy, whereas GC administration has been reported to increase plasma glucagon level (23,37). However, we note that there was a significant positive correlation between %ACA and the total dose of GC in GC-NDM, suggesting the possibility that a high dose of GC may increase ACM. Thus, further study is warranted to clarify this possibility in subjects who received higher dose and/or longer GC therapy than that in the current study. Examining whether ACM is increased in individuals with endogenous GC overproduction such as Cushing syndrome will also be of interest.

We also note that we did not find a significant increase in %ACA in individuals with T2DM. Hyperglucagonemia and a paradoxical increase in postprandial glucagon is a common pathological feature of T2DM (38). However, whether ACM is increased in patients with T2DM remains controversial (30,39). We rather observed a significant decrease in %ACA in patients with T2DM, inconsistent with another Japanese study (30). This inconsistency may be derived from the patients’ characteristics and methodological differences, as discussed previously (40). Because the proportion of α-cells increases with islet size, this inconsistency may also be derived from the difference in islet size distribution between individuals. Moreover, because we did not examine α-cell turnover in this study, the mechanism by which ACM decreased remains unclear. Nonetheless, the significant decrease in islet size with similar islet density in patients with T2DM observed in this study also indicates that the total number of islet endocrine cells is indeed reduced in patients with T2DM. Although recent rodent studies have suggested β-cell–to–α-cell transdifferentiation as a cause of β-cell loss in patients with T2DM (41), our and previous studies have shown no evidence of α-cell expansion in patients with T2DM (39). The significant negative correlation between %ACA and HbA_1c observed in this study also implies that ACM expansion is not a main contributor to hyperglycemia in humans with T2DM.

As with other autopsy studies, our study was not free of limitations. First, the underlying diseases, such as inflammatory and autoimmune diseases, that necessitated GC therapy might have affected BCM in individuals who received GC therapy, although matching underlying diseases between individuals with and without GC therapy is extremely challenging. In this study, we tried our best to compare the subjects with control subjects matched for age, sex, and BMI. Although the presence of a chronic systemic inflammatory status may decrease insulin sensitivity (42), this effect would tend to increase the difference between subjects with and without GC therapy. However, we are not able to exclude the possibility that other factors, such as family history of T2DM, duration of diabetes, and concomitant medications, as well as decreased body weight related to the cause of death, might also have affected our findings.

Second, in this study we assessed BCM and ACM by measuring %BCA and %ACA. Thus, any difference in pancreas volume between the groups might have affected our findings. However, the groups were mostly matched for age and BMI to minimize these effects on pancreas volume (43). It has been reported that the proportions of α- and β-cells are constant throughout the pancreas, except in the pancreatic polypeptide cell–rich ventral portion of the pancreatic head (28,29). In this study, only five case subjects were sampled from the pancreatic head, and the results were unchanged, even after excluding these case subjects (data not shown). We measured %BCA and %ACA in a single section of the pancreas, which might have resulted in greater interindividual variation; however, a wide range of BCM and ACM has been reported even in the population without diabetes (8,11,30,31,44,45).

Third, the use of different kinds of GCs and different regimens of GC therapy might have affected our findings, although we took account of the total days and total dose of GC use. We also confirmed that the results did not change when the subjects were classified according to long-term or short-term/intermittent administration.

Fourth, although the diagnosis and classification of diabetes were based on medical records before death, undiagnosed or misclassified case subjects might have affected our findings. However, we also confirmed the results based on HbA_1c values. We note that HbA_1c values in individuals without diabetes were relatively high in this cohort. A recent study showed that the HbA_1c level was 0.2–0.5% greater in Asians compared with Caucasians with the same plasma glucose levels (44). Other factors, such as anemia, also might have affected HbA_1c values (45). Finally, because the subjects of this study were Japanese, who are leaner and more insulin sensitive than other ethnicities, such as Caucasians, Hispanics, and Africans (46), and that there is an ethnic difference in β-cell change in response to obesity has been suggested (12,47), our findings may not be generalizable to other ethnicities. Because of these limitations, as well as the relatively small sample size of each group, our results should be confirmed in further studies with different populations.

In conclusion, GC therapy affected neither BCM nor ACM in adult humans with and without diabetes. These results suggest that GC-induced diabetes is mainly caused by insulin resistance and/or β-cell dysfunction, but not necessarily a deficit of BCM, and also underpin the minimal capacity of β-cell expansion in adult humans.

Acknowledgments. The authors thank Yuko Madokoro, Department of Pathology, Keio University School of Medicine, for technical assistance, and Dr. Wendy Gray, self-employed, for editing the manuscript.

Funding. This study was supported by funding from the Nateglinide Memorial Toyoshima Research and Education Fund, the Daiwa Securities Health Foundation, the Japan Diabetes Foundation, and Keio Gijuku Academic Development Funds (Y.S.).

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

Author Contributions. S.S. and Y.S. researched data and wrote the manuscript. J.I., K.K., R.M., and H.I. contributed to discussion and reviewed and edited the manuscript. T.Y. researched data, contributed to discussion, and reviewed and edited the manuscript. Y.S. 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.