Balancing needs and means: the dilemma of the β-cell in the modern world


Erol Cerasi, MD, PhD, Endocrinology & Metabolism Service, Hadassah University Hospital, P.O.Box 12000, 91120 Jerusalem, Israel.


The insulin resistance of type 2 diabetes mellitus (T2DM), although important for its pathophysiology, is not sufficient to establish the disease unless major deficiency of β-cell function coexists. This is demonstrated by the fact that near-physiological administration of insulin (CSII) achieved excellent blood glucose control with doses similar to those used in insulin-deficient type 1 diabetics. The normal β-cell adapts well to the demands of insulin resistance. Also in hyperglycaemic states some degree of adaptation does exist and helps limit the severity of disease. We demonstrate here that the mammalian target of rapamycin (mTOR) system might play an important role in this adaptation, because blocking mTORC1 (complex 1) by rapamycin in the nutritional diabetes model Psammomys obesus caused severe impairment of β-cell function, increased β-cell apoptosis and progression of diabetes. On the other hand, under exposure to high glucose and FFA (gluco-lipotoxicity), blocking mTORC1 in vitro reduced endoplasmic reticulum (ER) stress and β-cell death. Thus, according to the conditions of stress, mTOR may have beneficial or deleterious effects on the β-cell. β-Cell function in man can be reduced without T2DM/impaired glucose tolerance (IGT). Prospective studies have shown subjects with reduced insulin response to present, several decades later, an increased incidence of IGT/T2DM. From these and other studies we conclude that T2DM develops on the grounds of β-cells whose adaptation capacity to increased nutrient intake and/or insulin resistance is in the lower end of the normal variation. Inborn and acquired factors that limit β-cell function are diabetogenic only in a nutritional/metabolic environment that requires high functional capabilities from the β-cell.


Most discussions and reviews on type 2 diabetes mellitus (T2DM) start with the statement that T2DM is a heterogeneous disease [1]. It is not always clear what is meant by heterogeneity: the pathophysiological cause of the disease; the precise genetic modifications that led to the metabolic disorder; the fact that the clinical course of the disease, including response to treatment, may be quite variable. These are probably as valid for T2DM as for many other chronic diseases (e.g. hypertension [2]), and hardly place T2DM in a special category. Biological variability is a hallmark of life as well as of disease.

One question that has been debated intensively regarding the pathophysiology of T2DM is whether the disease is mainly because of insulin resistance or because of impaired insulin production. After decades of passionate and extreme positions [3–6], a consensus has emerged indicating that, as is often the case, both views are right and that to develop the diabetic metabolic disturbances both insulin resistance and failure of insulin production are necessary [7,8]. Nevertheless, we are still of the opinion that reduction of insulin secretion is the specific event that provokes T2DM, while insulin resistance is a more common phenomenon which, alone, may permit maintenance of normal glucose metabolism throughout life [9]. A convincing demonstration of the essential role played by the lack of normal insulin production in T2DM is the fact that moderate doses of insulin can near-normalize blood glucose concentrations in typical T2DM patients provided it is administered as physiologically as possible. As shown in table 1, both in Caucasian and Chinese moderately obese T2DM patients, excellent fasting and postprandial blood glucose control was achieved with insulin doses that are quite similar to those used in lean type 1 diabetic patients [10–12]. Thus, the insulin resistance of T2DM can be overcome by physiological insulin availability, at least in patients whose obesity is not excessive.

Table 1.  Continuous subcutaneous insulin infusion (CSII) treatment in T2DM patients
T2DM populationFasting glucose (mmol/l)2 h postprandial glucose (mmol/l)Daily CSII insulin dose (U/kg/24 h)
  1. *Calculated from (1) and (2); n = 23, mean ± s.e.m.

  2. Calculated from (3); n = 133, mean ± s.e.m.

Pre-CSII11.8 ± 1.416.9 ± 1.8 
On CSII4.7 ± 0.56.2 ± 0.60.55 ± 0.07
Pre-CSII11.3 ± 0.316.1 ± 0.5 
On CSII6.6 ± 0.17.5 ± 0.20.68 ± 0.02

β-Cell Adaptation and mTORC1

The above chapter is not to negate the existence of insulin resistance in T2DM, nor to reduce its importance in diabetes pathophysiology. Its molecular basis has been the focus of extensive research, and recently mTOR, an enzyme that integrates signals from nutrients and various growth factors and hormones to regulate protein biosynthesis and cell growth, has come to the forefront [13,14]. Indeed mTOR, via its mTORC1 complex and the downstream effector S6K1, phosphorylates IRS1 at serine residues, leads to its degradation, hence attenuation of cellular insulin signalling [15–21]. From such studies, the idea is emerging that inhibition of the mTOR/S6K1 pathway by pharmacological means could become a therapeutic choice in insulin-resistant states and T2DM.

We tested this hypothesis in Psammomys obesus (P. obesus), a gerbil with innate insulin resistance which, when kept on a high-fibre low-energy (LE) diet, maintains normoglycaemia and normal β-cell function; when fed a diet with higher caloric density (HE), the animals rapidly develop hyperglycaemia, initially with hyperinsulinaemia [22]. With prolongation of the HE diet, the hyperglycaemia worsens, hyperlipidaemia and high serum FFA appear, and β-cell mass is reduced drastically by massive death of the β-cells [23]. As shown in figure 1A, treatment of normoglycaemic animals with the relatively specific mTORC1 inhibitor, rapamycin [13], did not modify blood glucose. When diabetes was induced by HE diet, the addition of rapamycin, instead of improving, markedly impaired the diabetic state and was accompanied by collapsing plasma insulin levels (figure 1B), which was mainly because of marked β-cell apoptosis and loss of the β-cell mass (figure 1C) [24]. Moreover, we have shown that in the chronic in vivo situation of nutrition-induced insulin resistance and diabetes, rapamycin increased insulin resistance and had dramatic effects on blood glucose, serum triglycerides, FFA and ketone bodies [24]. This is in contrast to in vitro studies indicating that inhibition of TORC1/S6K1 may alleviate insulin resistance and improve β-cell survival [18,25,26]. Thus, prolonged inhibition of mTORC1 by rapamycin impedes pathways of major importance for reducing the metabolic disorders of T2DM, most markedly, β-cell survival in the face of chronic hyperglycaemia. We therefore regard the mTOR/S6K1 pathway as a critical regulator of β-cell survival in diabetes. This adaptation-promoting action, while not sufficient to correct the diabetic state, seems to be essential for limiting the degree of impairment of metabolism in T2DM. An unfortunate clinical implication of our results concerns islet transplantation, where rapamycin (sirolimus) is used as an immune-suppressor [27]; it is highly probable that the drug adversely affects the pre-existing metabolic state and harms the transplanted islets [28,29].

Figure 1.

Effect of rapamycin treatment on the evolution of diabetes in Psammomys obesus. LE, low-energy diet; HE, high-energy diet; LE-R and HE-R, diets with rapamycin treatment; Rapa, rapamycin treatment (0.2 mg/kg i.p. once-daily). Mean ± s.e.m. is shown. (A) Effect of rapamycin treatment on blood glucose concentration (in the LE-diet group differences were smaller than the symbols). (B) Serum insulin in the various treatment groups at termination of the experiment. (C) Effect on β-cell apoptosis (percentage of TUNEL-positive β-cells). (D) β-Cell mass at the end of the treatment periods. Adapted from [14].

Progression of β-Cell Failure

Once established, T2DM tends to progress relentlessly towards increasingly deranged metabolism; while the degree of insulin resistance does not change dramatically, β-cell function seems to decline linearly with time [30,31]. Figure 2 schematizes the changes in glucose-induced insulin secretion that are observed in the various phases of the progression of T2DM. First-phase insulin secretion is markedly diminished already in subjects with impaired glucose tolerance (IGT) in the presence of normal fasting blood glucose concentrations; with the appearance of even modest hyperglycaemia, first-phase response is abolished [32,33], while late, second-phase insulin response to glucose may be relatively maintained. The latter diminishes, and finally almost disappears, as the severity of T2DM progresses.

Figure 2.

Schematic illustration of the modifications in the insulin response to a 1-h hyperglycaemic clamp throughout the progression of T2DM. NGT, subjects with normal glucose tolerance; IGT, with impaired glucose tolerance.

Why does insulin secretion diminish with duration of diabetes? The dominating view is that β-cell mass is markedly reduced in T2DM [34]. While this information, based on autopsy material, is accepted by the majority of the researchers in the field, the fact that β-cell mass can not be measured in vivo makes it difficult to assess the quantitative correlation between the reduction of β-cell number and the degree of metabolic derangement. Indeed, careful measurements of β-cell mass (rather than β-cell volume as in most studies) by Rahier et al. [35] established its extraordinary variability, both in T2DM patients and their matched controls, such that in many diabetic patients β-cell mass may even be higher than in non-diabetic subjects. It is therefore reasonable to assume that β-cell function can be severely impaired also in the absence of reduced β-cell mass; thus, the factor of importance for glucose homeostasis in this context should be the ‘functional’β-cell mass, that is, the mean functional capacity of β-cells multiplied by β-cell number. Despite these reservations, it is clear that β-cell mass is in general found diminished in T2DM patients; in any case it is not augmented to compensate for insulin resistance and hyperglycaemia. Furthermore, abundant experimental data in animal models of diabetes and in vitro (reviewed in [36,37]) show that the main feature of the diabetic milieu–high concentrations of glucose and free fatty acids (FFA)–is deleterious for the β-cell, both in terms of its function and its survival.

The Two Faces of mTOR in T2DM

It is accepted that both high concentrations of glucose and FFA induce β-cell dysfunction and apoptosis, a process called gluco-lipotoxicity [37,38]. Palmitate (the FFA with most potent lipotoxic effect) induces β-cell dysfunction and apoptosis by endoplasmic reticulum (ER) stress mechanisms [39–42]. This activates the unfolded protein response (UPR) aimed at restoration of normal ER function; if it fails to do so, signalling pathways are turned on to initiate β-cell apoptosis [42–44]. The toxic effect of FFA on the β-cells is of significance mainly under the in vivo hyperglycaemic state [45,46]. One explanation for the hyperglycaemia-dependency of lipotoxicity could be activation of mTOR, with subsequent increased protein synthesis, mainly proinsulin. As secretory proteins are the main clients of the ER protein folding machinery, it may be expected that augmented proinsulin production in hyperglycaemia increases the ER stress induced by high-FFA concentrations.

We tested this hypothesis in the INS-1E β-cell line and in islets of P. obesus. A first finding of interest is that the stress induced by FFA is not because of increased load on the ER, because addition of palmitate to a high glucose-containing culture medium markedly reduces, rather than increase, the total protein synthesis as well as proinsulin biosynthesis [47]. Yet, there was ample evidence of ER stress induction by FFA, as shown in figure 3: addition of palmitate to high glucose markedly augmented the expression of CHOP, phosphorylated PERK, cJun and JNK (figure 3A), and this led to β-cell death as evidenced by the increase in cleaved caspase 3 expression and β-cell apoptosis (figure 3B). The FFA–glucose synergism at the level of ER stress is apparently expressed via the IRE-1 pathway of the UPR, because palmitate increased the levels of both total and phosphorylated IRE-1α, as well as that of its downstream effector, spliced Xbp-1 (figure 3C) [47].

Figure 3.

FFA-induction of ER stress: role of mTORC1. (A) INS-1E β-cells were incubated for 16 h in the presence of 22 mmol/l glucose without and with 0.5 mmol/l palmitate (FFA– and +). Addition of 50 nmol/l rapamycin markedly reduced the stimulatory effect of palmitate on ER stress signalling molecules (Western blots). (B) Palmitate increased cleavage of caspase 3 (Western blots) and caused increased β-cell apoptosis (Cell Death ELISAPLUS assay) in INS-1E β-cells subjected to the conditions of (A). Addition of rapamycin decreased the above FFA effects. (C) Palmitate acts through the IRE-1α pathway of the UPR. Conditions are similar to those of (A). Left, a representative Western blot for total and phosphorylated IRE-1α; right, spliced Xbp-1 levels assessed by quantitative real-time PCR. Adapted from [37].

Is mTORC1 involved in gluco-lipotoxic ER stress? Inhibition of mTORC1 with rapamycin markedly diminished the glucose-amplification of palmitate-induced ER stress and apoptosis (figure 3A–C). Thus, the mechanisms by which glucose amplifies the cytotoxic effect of palmitate on the β-cell are indeed mTORC1-dependent. Interestingly, rapamycin did not affect proinsulin and global protein synthesis in β-cells exposed to high glucose and palmitate, while decreasing IRE1α expression and phosphorylation. This may suggest that the ER stress induced by gluco-lipotoxicity is not because of ‘fatigue’ of the protein folding machinery via increased biosynthesis of proinsulin in the overworking β-cell; it is rather because of the augmented synthesis of IRE1α, with activation of JNK as a consequence [47]. It remains to be shown, nevertheless, that this mechanism is operative in vivo.

At first sight, the above results–where mTORC1 activation is shown to lead to β-cell death, and those narrated in the former chapter where mTORC1 is described as necessary for β-cell adaptation to chronic hyperglycaemia and protection from death–indeed seem contradictory. However, it has to be borne in mind that the conditions of the experiments were quite different, both duration and intensity of exposure to hyperglycaemia and high-FFA varying. We therefore propose that mTORC1 plays a dual role in the β-cell exposed to a diabetic milieu, according to environmental and perhaps genetic conditions that remain to be defined. On the one hand, by stimulating protein synthesis and cell proliferation, it boosts insulin production and thus permits a certain adaptation to the augmented requirements imposed by the hyperglycaemia, thereby limiting the severity and progression of T2DM. On the other hand, mTORC1 being instrumental for the expression of ER stress and β-cell death induced by gluco-lipotoxicity, it is a link in the progression and worsening of T2DM by reduction of the β-cell mass. It seems crucial to us that the conditions favouring the positive vs. negative effects of mTORC1 on β-cell survival and function be clarified in detail, as this will be the conditio sine qua non for designing antidiabetic treatments targeting mTORC1.

Why Do People Become Diabetic?

With the above experimental approaches we aimed to clarify the mechanisms that compromise β-cell function and survival in T2DM; all have dealt with the impact on the β-cell of the diabetic environment with its characteristic high concentrations of glucose and FFA. While providing a satisfactory explanation for the gradual deterioration of metabolism throughout the duration of the disease, they do not answer the question of how the noxious diabetic environment was established, in other words, how IGT/T2DM starts. One obvious possibility is that, with or without insulin resistance, in the presence of nutrient overload the augmented functional requirements on the β-cell elicit the same molecular events described above for the hyperglycaemic state, but at barely detectable levels. These subtle modifications in the long run might lead to the same results as in the above exaggerated conditions of abrupt in vitro exposure to high glucose and palmitate. Unfortunately, we lack the tools to experimentally create the chronic ‘prediabetic’ condition and measure the expected subtle changes in the ER and other functions of the β-cell; therefore this hypothesis is difficult to test.

A normal β-cell has an extraordinary capacity to adapt to insulin resistance, as demonstrated by the fact that normal glucose-tolerant obese persons, or women in the 3rd trimester of pregnancy, show a twofold to fourfold amplified insulin response to glucose, and autopsy material indicates approximately 50% increase in β-cell mass [34,48–50]. But does β-cell function perfectly adapt itself to insulin resistance in order to maintain unchanged glucose homeostasis? The hyperbolic ‘disposition index’ curves suggest that this is the case [51]. However, a more careful approach challenges this view [52]. Lean normal glucose-tolerant subjects were separated into highly insulin-sensitive and insensitive groups and challenged with a fixed-dose intravenous (i.v.) bolus-constant rate glucose infusion (figure 4A). As expected, the insulin response of the subjects with a lower sensitivity to insulin was approximately twofold higher than in highly insulin-sensitive persons. However, against expectation, integrated blood glucose levels achieved with the equal-dose glucose infusion were also higher (∼1.5-fold). Thus, the insulin response that was augmented as compensation for the reduced insulin sensitivity, while sufficient to maintain the glucose tolerance within normal limits, was insufficient in the face of a more robust challenge. An extension of this observation relates to the existence of healthy glucose-tolerant subjects with reduced first-phase insulin response to glucose (and other stimuli) [53]. While their glucose tolerance, by selection, was strictly within normal limits, it was significantly lower than in healthy subjects who were matched for insulin sensitivity but with a high insulin response to glucose [54]. Thus, we do seem to pay a metabolic price for even small reductions in the functional capacity of the β-cell. Indeed, when healthy lean subjects were followed for 25 years, the 2-h glucose level at the final oral glucose tolerance test (OGTT) was correlated significantly with first-phase insulin response to glucose at the initial test (corrected for insulin sensitivity [Homa-IR]) [55]. In 53 out of the 267 lean normal glucose-tolerant subjects, IGT or T2DM had developed after 24.3 ± 2.9 years; in these, first-phase insulin response to glucose, measured a quarter-century earlier, was significantly lower than in the subjects who remained glucose tolerant (figure 4B). Thus, in lean subjects an initially reduced insulin response, but not insulin resistance, predicts future impairment in glucose homeostasis [55]. This is different from the situation in an obese population, where both decreased insulin response and decreased insulin sensitivity predicts future IGT/T2DM [56].

Figure 4.

Impact of the insulin response to glucose on glucose homeostasis in lean healthy subjects. (A) Plasma insulin and glucose responses to an i.v. bolus injection of glucose (0.5 g/kg) followed by constant-rate infusion of glucose (20 mg/kg/min) for 60 min. The test was performed in normal glucose-tolerant lean subjects with higher (H) and lower (L) sensitivities to insulin. Plasma insulin and glucose values were integrated over a 2-h period. Adapted from [42]. (B) First-phase insulin response to 0.5 g/kg glucose i.v. was assessed in lean normal glucose-tolerant subjects and expressed as 5-min insulin peak corrected for the 5-min blood glucose value and the insulin sensitivity (measured as HOMA-IR). OGTT was performed 25 years later and correlated to the initial insulin response. Adapted from [45].

In our early studies we postulated that glucose-tolerant subjects with low insulin response are prediabetic (in the sense of being at risk to develop IGT/T2DM, not as presently used to delineate subjects with IGT or IFG, we regard them as early-T2DM) [47]; the above follow-up study may justify this assumption. Although we have known for a considerable time that the insulin response is partly genetically controlled [57,58], its molecular basis was unknown. Recently, from genome-wide association studies polymorphisms of several genes controlling various aspects of the biology of β-cells have emerged as T2DM risk factors, with TCF7L2 being the most prominent one [59]. Most interestingly, it has been shown that in normal glucose-tolerant subjects the insulin response to glucose decreases as the number of risk-associated alleles increases [60]. These recent findings may explain why we were unable to delineate a distinct prediabetic group in the normal population, because the insulin response to glucose showed a continuum from lowest to highest values [61].

Four decades after our initial thoughts about T2DM development [61], we would like to present a modified view on how we see prediabetes. A ‘prediabetic’β-cell is a perfectly normal β-cell, but its capacity to adapt to increased demand of insulin production and/or mass expansion is in the lower-end of the normal biological variability. Thus, the ‘prediabeticity’ of the β-cell is entirely context-dependent: until a given degree of nutritional overload and/or insulin resistance is reached, this β-cell functions normally and maintains normal glucose homeostasis. The ‘given degree of load’ may be a function of the genetic background of the β-cell. To take an example, the T variant of TCF7L2 gives a 40% increased risk of developing T2DM [59]. However, this is true for the world of today, with its excessive nutrition and physical inactivity; we postulate that had the genome-wide association studies been performed in the immediate post-World War 2 Europe, TCF7L2 would not have emerged as a T2DM risk factor, because the T-homozygous β-cell would have been perfectly capable of handling the limited nutritional load of the epoch. These considerations lead us also to propose an alternative to the ‘thrifty gene hypothesis' [62], which sees an evolutionary advantage to obesity/insulin resistance: the ‘lazy gene hypothesis'–gene variants that limit the functional capabilities of the β-cell had no evolutionary disadvantage as these β-cells, over the whole period of evolution, were never subjected for extended periods to the excessive demands of a nutritional overload, except for the past half-century.

In conclusion, the slowly increasing understanding of the interplay that exists between a multitude of genetic variations that control how the β-cell copes with the ‘stress of life’, and the modern life conditions that seem to generate β-cell stress, is opening a vast area of research, many heretofore unknown cellular mechanisms being clarified. One area that begs further development relates to the dual role of mTORC1 in type 2 diabetes: how to enhance the β-cell adaptation-promoting effects of mTORC1 while preventing its ER stress-accentuating action? We hope that from such research optimal means for the treatment and prevention of diabetes will emerge in a not-too-distant future.


The research described above is the result of decades-long collaborations with too many scientists to be named here. We wish to acknowledge our gratitude to all of them, but more specifically to our more recent collaborators, Merav Fraenkel, Mali Ketzinel-Gilad, Yafa Ariav, Maayan Shaked and Etti Bachar. We are also grateful to the multiple granting agencies that supported our work over the years.

Conflicts of Interest

The authors have declared no conflicts of interest.