Five Stages of Evolving Beta-Cell Dysfunction During Progression to Diabetes

The most common example of compensation is found with the insulin resistance due to obesity, which is accompanied by higher overall rates of insulin secretion (2) and increased acute glucose-stimulated insulin secretion (GSIS) following an intravenous glucose challenge (3). Much of the increase in insulin secretion undoubtedly results from an increase in β-cell mass, as has been found in autopsy studies in humans (4–7) and numerous rodent models (8,9). Beta-Cell mass is normally tightly maintained through a balance of β-cell birth (β-cell replication and islet neogenesis) and β-cell death through apoptosis (10). Most of the increase in β-cell mass with insulin resistance is probably due to increased β-cell number, but β-cell hypertrophy may also contribute. It is not yet clear if the higher plasma insulin levels can be entirely explained by the larger β-cell mass or whether there is also increased secretion per given unit of β-cell mass. Although compensation is usually thought of in the situation of insulin resistance, similar changes presumably occur in the early stages of autoimmune destruction. As β-cell mass falls, there must be a signal to increase mass and secretion, which presumably prolong the prediabetic period, which can last for years (11).

There is much interest in the signal leading to increased β-cell mass in this situation. The facile, but probably correct, explanation is that there is a feedback loop with insulin resistance producing increased glucose levels that stimulate β-cell secretion and growth. However, concerns are raised about how this can be compatible with seemingly normal glucose levels. One explanation is that the feedback loop is tightly regulated, like a thermostat that can maintain temperatures within a very narrow range. A molecular mechanism that could facilitate this control is activation of glucokinase in β-cells, the enzyme that controls the rate of glycolysis, thereby determining insulin secretory rates (12). Thus, very small elevations in glucose levels could lead to a change in the set point in GSIS that allows maintenance of “normal” plasma glucose levels (13).

The unique differentiation of β-cells is responsible for the extraordinary efficiency of these cells in storing and secreting insulin to provide precise regulation of metabolism. With successful compensation for insulin resistance, little change in β-cell phenotype for the secretory machinery would be expected. The unique specialization of β-cells is presumably necessary for optimal GSIS; the impressively large acute insulin responses to glucose challenge seen in obesity strongly suggest the phenotype is maintained. The complexity of this differentiation is far from being understood, but some key features are apparent. The glucose transporter GLUT2 allows rapid equilibration between extra- and intracellular glucose levels. Glucose is phosphorylated by glucokinase with a K_m of ∼8 mmol/l that allows it to function efficiently within the normal range of plasma glucose concentrations (14). Glucose metabolism is dominated by glycolysis with pyruvate directed to mitochondria for oxidation. There is little if any gluconeogenesis (15) or lactate production (16), which should be helpful for maximizing the efficiency of aerobic glycolysis. The β-cell also has specialized mitochondrial shuttles. The glycerol phosphate shuttle allows reduced nicotinamide dinucleotide (NADH) to be oxidized by mitochondria, thereby contributing to adenosine triphosphate (ATP) formation. The oxidation of NADH should also enhance glycolytic flux. In β-cells the malate/pyruvate shuttle may facilitate the generation of NADPH, which could somehow enhance secretion. The need for these shuttles probably explains why β-cells have very high levels of mitochondrial glycerol phosphate dehydrogenase (mGPDH) and pyruvate carboxylase (17,18).

To maintain this degree of specialization, the genes that are highly expressed in β-cells include those of the secretory products (insulin and islet amyloid polypeptide [IAPP]), key genes for glucose metabolism (GLUT2 and glucokinase), key enzymes for the mitochondrial shuttles (glycerol phosphate dehydrogenase [mGPDH]) and pyruvate carboxylase, and critical transcription factors (PDX-1 and Nkx-6.1). Yet, enzymes that participate in unwanted pathways such as gluconeogenesis and lactate production are expected to be suppressed. Some of these include phosphoenolpyruvate carboxykinase (PEPCK), glucose-6-phosphatase and fructose-1,6-bisphosphatase, and lactate dehydrogenase. Hexokinase would also be unneeded, so its expression could also be suppressed. Clearly these changes in gene expression are just the tip of the iceberg, and a full picture of β-cell phenotype awaits gene array studies, which should emerge in the near future.

Some experimental data support the concept that much of the β-cell phenotype is kept intact during successful compensation for insulin resistance. When glucose was infused into rats for 4 days, compensation occurred and glucose levels remained normal during the latter stages of the infusion because of increases in insulin secretion that were accompanied by increased β-cell mass (19). While responding to the increased demand, the β-cell phenotype, as determined by analysis of a selected group of genes, remained remarkably similar to the control profile (19). There must have been induction of some genes required for β-cell replication, but the genes responsible for maintaining the machinery allowing GSIS did not seem to be perturbed. This preservation of phenotype, coupled with an increase in β-cell mass, is consistent with the exuberant GSIS seen in obesity.

It is not possible to assign a precise glucose range for this stage of β-cell adaptation, but fasting levels between 5.0 and 7.3 mmol/l (89–130 mg/dl) are reasonable approximations. In stage 2, β-cells can no longer be considered to be compensating because truly normal glucose levels can no longer be maintained. However, this stage can be considered stable because, as shown in the Diabetes Prevention Program, individuals at the upper end of stage 2 with IGT progressed to diabetes at the rate of ∼11% per year while those who adhered to a diet and exercise regimen progressed at a rate of only 5% per year (20). Thus, unless some process, such as autoimmunity, is producing rapid β-cell destruction, individuals in stage 2 usually evade progression to diabetes for years.

Nonetheless, as glucose levels rise to stage 2, important changes in β-cell function and differentiation occur. The most striking and best studied is the loss of acute GSIS. The acute phase of secretion occurs ∼3–10 min after the start of the glucose infusion and represents the first phase of GSIS (21). This experimental assessment has the pathophysiological correlate of impaired early insulin release during an oral glucose challenge, which contributes to glucose intolerance (22). An important study published in 1976 (23) showed that subjects maintained normal acute GSIS as long as fasting plasma glucose levels remained <5.6 mmol/l (100 mg/dl). However, acute GSIS dropped drastically when fasting plasma glucose values were above this level and disappeared completely once fasting glucose levels increased to only 6.4 mmol/l (114 mg/dl). This pattern has now been confirmed in countless studies in humans and animals. In contrast to this loss of acute GSIS, the second phase of insulin secretion is partially preserved (24), and acute insulin responses to nonglucose secretagogues, such as isoproterenol and arginine, remain largely intact (25).

The concept of glucose toxicity is that β-cells normally function within a narrow range of plasma glucose levels and that even modestly higher glucose levels create an unnatural environment, which leads to alteration in function and most notably a loss of acute GSIS. On the molecular level, the exact mechanisms responsible for this loss of GSIS are not known; however, we postulate that a loss of specialized gene expression leads to complete disruption of the acute phase of GSIS. In contrast, the agents that stimulate insulin release through depolarization (arginine) or cyclic AMP (isoproterenol) work through more ubiquitous and less fragile mechanisms that are maintained (26). Although some have proposed that increased free fatty acid (FFA) levels are important for the β-cell dysfunction of diabetes (27), these levels correlate poorly with the loss of acute GSIS, whereas the correlation with glucose levels is very precise (26). This interpretation does not rule out important glucose-driven derangements of lipid pathways in β-cells that could justify use of the term “lipotoxicity.”

The loss of acute GSIS seen in the diabetic state is accompanied by marked changes in β-cell phenotype demonstrated by changes in gene and protein expression. These changes have been found in islets from both diabetic ZDF rats (28) and rats following partial pancreatectomy in which residual β-cells are exposed to varying degrees of hyperglycemia (29–32). The alterations may be thought of as dedifferentiation, or loss of phenotype, in that the highly expressed genes in β-cells (mentioned above) are downregulated while those that are normally suppressed are upregulated in their expression. Some of the upregulated genes include glucose-6-phosphatase, fructose-1,6-bisphosphatase, lactate dehydrogenase, and hexokinase. In addition, a stress response is found with a variety of antioxidant, apoptotic, and proapoptotic genes being activated (31). These are accompanied by a marked increase in the expression of c-myc and activation of nuclear factor (NF)-κB. These changes are consistent with a recent report that interleukin (IL)-1 expression in β-cells is activated by high glucose levels (33). Another interesting facet is the presence of β-cell hypertrophy, which may be a glucose-driven growth response stopping short of replication (29–32). Strengthening the hypothesis that these changes are secondary to chronic hyperglycemia are data showing that lowering of the elevated glucose levels with phlorizin, which inhibits renal glucose reabsorption, leads to reversal of virtually all of the changes in gene expression and hypertrophy (29).

It should be emphasized that some changes of gene expression occur in stage 2, when glucose levels might be considered “normal.” In some rats following a 90% partial pancreatectomy, fed plasma glucose levels of 6.9 mmol/l (124 mg/dl) were maintained over a 14-week period, whereas control rats maintained levels at 5.8 mmol/l (104 mg/dl) over the same time period (32). When considered in human terms, this increase of only 1.1 mmol/l (20 mg/dl) would probably not have reached the designation of IGT and therefore would be categorized as normal. However, the islets isolated from the pancreatic remnant of these rats 14 weeks after surgery had similar, albeit less marked, changes in gene expression as those of partially pancreatectomized rats with more severe hyperglycemia. Despite the minimal degree of hyperglycemia, these changes in gene expression correlate with the losses of GSIS seen in this and similar rodent models (34,35) and in humans (23), supporting the hypothesis that changes in β-cell phenotype are responsible for the disruption of GSIS. We recognize that blaming the loss of GSIS on altered phenotype may be an oversimplification. Certainly there could be important changes in signal transduction pathways that could be unrelated to changes in protein levels, but we expect that the changes in phenotype are of fundamental importance. In the future it should be possible to dissect out the many candidates that could contribute to β-cell dysfunction, such as mitochondrial shuttles, lactate dehydrogenase, glucose-6-phosphatase, uncoupling protein-2, peroxisome proliferator–activated receptor (PPAR)-α, and c-myc.

We postulate that during progression toward diabetes, stable stage 2 ends and glucose levels rise relatively rapidly from the range of 7.3 mmol/l (130 mg/dl), through an unstable transient stage 3 of decompensation, to a more stable stage 4 at ∼16–20 mmol/l (285–350 mg/dl). Thus, individuals headed to type 2 diabetes can remain in stage 2 for many years, but when β-cell mass becomes inadequate at some critical point, glucose levels rise over a relatively short period of time to stage 4, which may or may not be associated with noticeable symptoms, such as polyuria and weight loss. A similar progression occurs in type 1 diabetes, but stage 2 does not last as long because autoimmune destruction of β-cells occurs much more rapidly than whatever process is responsible for the β-cell attrition in type 2 diabetes. Thus, a child nearing adolescence might be in stage 2, whereupon flu or some other stress might push glucose levels rapidly through stage 3 to stage 4. While in stage 4, the precipitating stress may recede, allowing the child to fall back through stage 3 to stage 2 for a remission, which would only last until further β-cell destruction results in a return to stage 3 and then stage 4. Stage 4 might not be long lasting because continued β-cell loss will usually lead to the severe decompensation of stage 5, with propensity to ketoacidosis. The same pathophysiology probably occurs with pancreas and islet transplantation, in that loss of β-cells in the grafts would lead to progression through the above stages. While a successful pancreas transplant will typically bring a recipient to stage 1, an islet transplant performed according to the Edmonton approach (36) will usually place recipients in stage 2, where they will remain at risk for later decompensation (37).

When transplanting islets contained in immunoprotective devices into mice, we noted that a marginal number of islets produced either success with normal blood glucose levels or failure with glucose levels >15 mmol/l (280 mg/dl), with almost no glucose levels in between (38) (Fig. 2). A marginal number of syngeneic islets transplanted under the kidney capsule of streptozocin diabetic mice produced a similar pattern of either failure or success, with only a few glucose levels in the stage 3 range (data not shown). This concept was examined again in rats with a marginal reduction of β-cell mass following a 90% partial pancreatectomy (Fig. 3). Although a spectrum of glucose levels was seen at 4 weeks, the rats showed two distinct outcomes by 14 weeks: they were either near normal (glucose levels of 6.9 mmol/l [124 mg/dl] and 5.8 mmol/l [104 mg/dl] in fed and control rats, respectively) or unequivocally hyperglycemic (32). Somehow there were forces pushing these rodents to either stage 2 or 4, with their time in stage 3 being only transient.

Support is also provided by studies of islet transplantation in which canine islets were transplanted into the livers of young diabetic dogs (39). As the recipients matured, rapid decompensation to marked hyperglycemia was seen, presumably a result of the β-cell mass becoming inadequate to meet the insulin requirements of larger dogs. A recent study in humans also supports the concept that decompensation is not a continuum but can occur in a relatively short period of time. In studies on a population from the Mexico City Diabetes Study, it was concluded that within a 3-year timeframe the development of diabetes was often rapid rather than gradual (40). The investigators also thought the rapidity might have been underestimated because some of the subjects were receiving diabetes treatment.

At some point in stage 2, β-cells are no longer able to keep glucose levels in the prediabetic range. This failure presumably occurs because of a critical decline of β-cell mass and/or increase in insulin resistance. We postulate that there is a conspiracy of forces that push glucose levels upward, which include the insulin resistance associated with the diabetic state with its complex gluco- and lipotoxicity influencing the key insulin target tissues (i.e, liver, muscle, and fat) (41). The increases in glucose concentration probably also worsen glucotoxic effects on β-cells, leading to less efficient insulin secretion. For example, as insulin mRNA falls with increasing hyperglycemia, there is evidence that insulin biosynthesis becomes rate limiting for secretion (42). With these factors working synchronously, glucose levels rise to stage 4 but do not immediately progress to stage 5 because enough insulin production continues to prevent severe ketosis. By treating type 2 diabetes with diet, exercise, and drugs, individuals can quickly return to stage 2 and remain in that stable range for a considerable period of time as long as they continue the treatment and have no further decline in β-cell mass.

Clinicians often see patients who are asymptomatic and have unremarkable glucose levels at one point and then present with glucose levels in the range of >16 mmol/l (285 mg/dl) with the absence or presence of symptoms. Certainly individuals progressing to type 1 diabetes often rapidly progress to very high glucose levels, but clearly many people with diabetes have glucose levels within the stage 3 range. The most obvious explanation is that external forces can overwhelm the processes that make stage 3 transient (in particular, treatment with oral medications and/or insulin). Even changes in caloric intake and exercise should allow individuals to move in and out of stage 3. Another point is that the transition from stage 2 to stage 4 is not instantaneous; even in our rodent studies (Fig. 3) progression took a few weeks. Nonetheless, we postulate that stage 3 is by nature transient.

Once individuals progress through stage 3 to the unambiguous diabetes of stage 4, they typically have enough insulin secretion to remain in this stage rather than progressing to ketoacidosis. In most cases this stage lasts a lifetime for people with type 2 diabetes, while the rapid progressive autoimmune destruction of β-cells in type 1 diabetes can lead to stage 5 relatively quickly. Individuals with failing islet transplants can probably stay in stage 4 for relatively long periods of time if the immunosuppression regimens allow sufficient β-cell survival.

Morphometric studies on postmortem pancreases of people with type 2 diabetes provide convincing evidence that β-cell mass is reduced to ∼50% of that of control subjects (4–6). Data generated in rodents suggest β-cell dedifferentiation is more severe at higher glucose levels (29), which should result in less efficient insulin secretion. This lack of efficiency fits with insulin secretion studies in type 2 diabetes that have indicated that the capacity for secretion is considerably less than the 50% reduction in β-cell mass (43).

Important questions are raised by the finding that a stress response is induced by chronic hyperglycemia (31). Because the stress response is characterized by the expression of a mixture of proapoptotic, antiapoptotic, and antioxidant genes, it is difficult to know if β-cells in this situation would be more or less susceptible to the apoptosis. For destruction by autoimmunity and/or transplant rejection, the stress response could make β-cells either more vulnerable or provide protection. For type 2 diabetes, this phenotypic change may accelerate β-cell loss. Even if accelerated, the rate of β-cell apoptosis in type 2 diabetes is certainly very slow, given the assumption that the β-cell death rate must be similar to β-cell birth rate, which is probably considerably less than 1% per day (4,44). Even less is known about the rate of neogenesis, which is probably very slow as well (4). Somehow, regardless of β-cell loss in typical type 2 diabetes, apoptosis rarely progresses to near-complete loss, which leads to ketosis, even after decades of the disease. Although it is often assumed that accelerated apoptosis of β-cells is important for the pathogenesis of type 2 diabetes, limitations in β-cell replication and/or neogenesis could be just as important. Although some rodent models of diabetes exhibit increased rates of apoptosis (8,45), in rats following 90% partial pancreatectomy with glucose levels in the stage 4 range of diabetes, rates of β-cell replication and the incidence of apoptotic bodies were no different than in control rats (46).

In this last stage of diabetes, the marked loss of β-cells is so severe that people become ketotic and truly dependent on insulin for survival. Glucose levels are typically >22 mmol/l (350 mg/dl) but will vary with eating and hydration. This situation is typically found with type 1 diabetes or patients with pancreas or islet transplants when β-cells have been mostly destroyed by the immune system. It can also occur in unusual situations, such as exposure to certain toxins or very severe pancreatitis, but it rarely occurs in typical type 2 diabetes.

The development of the concept of five stages of β-cell function that characterize progression of diabetes has emerged from clinical observation, awareness of clinical investigation by many workers, and work with animal models of diabetes. This perspective provides only a broad picture of these stages and will hopefully stimulate studies that will more precisely define the mechanisms leading to diabetes.

This work was supported by grants from the National Institutes of Health/Diabetes Endocrinology Research Center (grant nos. DK-36836 and DK-35449), the Juvenile Diabetes Research Foundation, and the Diabetes Research and Wellness Foundation. We thank Dr. Rafael Nesher for thoughtful insights and suggestions.