In healthy individuals, insulin resistance is associated with physiological conditions such as pregnancy or body weight gain and triggers an increase in beta cell number and insulin secretion capacity to preserve normoglycaemia. Failure of this beta cell compensation capacity is a fundamental cause of diabetic hyperglycaemia. Incomplete understanding of the molecular mechanisms controlling the plasticity of adult beta cells mechanisms and how these cells fail during the pathogenesis of diabetes strongly limits the ability to develop new beta cell-specific therapies. Here, current knowledge of the signalling pathways controlling beta cell plasticity is reviewed, and possible directions for future research are discussed.
Insufficient insulin secretion by pancreatic beta cells to compensate for developing insulin resistance of liver, muscles and adipose tissue is considered to be the cause of overt type 2 diabetes . In insulin-resistant conditions, such as during pregnancy or in response to increased body weight, there is an increase in both beta cell number and glucose competence (i.e. the amount of insulin the cells can secrete in response to a given rise in extracellular glucose concentration). This beta cell plasticity ensures that insulin secretion can precisely match the metabolic requirements of the organism under changing environmental conditions and maintains normoglycaemia throughout life (Fig. 1). The roles of increased cell number and glucose competence have been investigated in animals and humans. Histomorphometric analysis of pancreatic autopsy samples revealed a higher beta cell mass in the pancreas from obese, insulin-resistant individuals, compared with samples from normal-weight individuals, but the beta cell mass in the pancreas from individuals with type 2 diabetes was reduced in association with increased signs of apoptosis . These findings suggest that reduction in beta cell mass may underlie the decreased insulin secretion capacity. However, further analysis showed that the reduction in beta cell mass was proportional to the time from onset of diabetes and that hyperglycaemia probably developed when beta cell mass was still within normal levels . This implies that reduced glucose competence of a normal number of beta cells may lead to the onset of diabetes. Analysis of the data presented in this last publication also shows that there is no relation between beta cell mass and insulin secretion capacity, with a much higher beta cell mass in the pancreas from some diabetic subjects than from many normal individuals. Thus, there is no direct correlation between beta cell mass and glycaemic control, suggesting that glucose competence of individual beta cells is a major factor in determining pancreatic endocrine function.
Similar conclusions have been drawn from the results of animal studies. In particular, in a study of genetically obese and diabetic mice (ob/ob or db/db mice), it was shown that beta cell mass and plasma insulin levels were markedly increased during the progression of obesity and insulin resistance. However, after a few months of diabetic hyperglycaemia, a reduction in beta cell mass and hypoinsulinaemia developed [4, 5]. In a model of nutrition-induced metabolic stress, mice fed a high-fat diet that rapidly developed insulin resistance leading to compensatory insulin secretion capacity [6, 7]. This response is highly influenced by the genetic background of the mice studied [8-11]. In humans, genome-wide association studies have been used to identify single-nucleotide variants in or in close proximity to more than 50 genes that influence susceptibility to type 2 diabetes. It is noteworthy that most of these genes are expressed in beta cells and participate either in the glucose signalling pathway that controls insulin secretion or in various transcriptional mechanisms that control beta cell differentiation and proliferation [12-15]. Thus, to understand the pathogenesis of type 2 diabetes, the critical signalling or metabolic pathways that control beta cell proliferation and glucose competence must be identified and how the activity of these pathways is influenced by individual genetic variability must be determined. Here, the evidence for plasticity of adult beta cell mass and function is reviewed, together with the respective signalling pathways involved. In addition, potential future lines of research are explored.
Adult beta cell plasticity
Over the last decade, a major focus in beta cell research has been the investigation of beta cell generation from precursor cells or from embryonic stem cells. This effort led to an understanding of the transcriptional control of pancreatic endocrine development during embryogenesis [16-18]. It also led to the establishment of the optimal conditions to differentiate mouse and human embryonic stem cells into endodermal progenitors, which could further be differentiated into endocrine cells, including insulin-secreting cells [19-21]. More recently, through genetic techniques allowing immortalization of human embryonic beta cells, protocols have been established to generate human beta cell lines capable of unlimited replication whilst preserving glucose-stimulated insulin secretion and other features of adult beta cells, such as sensitivity to the action of gluco-incretin hormones . These cell lines are vital for the study of human beta cell biology; however, whether they can also be used for the treatment of diabetes, for instance using cell transplantation techniques, is still unclear.
In comparison with the study of beta cell development and differentiation, progress in understanding the molecular control of adult beta cell plasticity and in defining the methods to manipulate it has been more modest. In theory, controlling beta cell number can be achieved by increasing beta cell proliferation, inducing beta cell differentiation from precursors or protecting mature beta cells against apoptosis; this last process is caused by the combination of inflammatory cytokines released locally by inflammatory or immune cells or secreted by adipose tissue or muscle [23, 24] and high plasma levels of glucose and free fatty acids, that is, glucolipotoxicity .
Replication following destruction of adult beta cells
In normal physiological conditions, adult beta cell replication, although very difficult to precisely assess (especially in humans), is considered to be very low. Findings from studies in mice and rats suggest that the rate of beta cell renewal is ~3% per day, that is, complete replacement every month , and replication appears to be the major pathway to beta cell neoformation . The rate of replication is high in the early postnatal period and declines rapidly in adult animals. In humans, it has been suggested that once the full complement of beta cells has been generated in young adults, almost no replication occurs later in life .
Beta cell expansion can be induced experimentally in adult animals in several ways: (i) in response to partial pancreatectomy , (ii) in response to destruction of beta cells by diphtheria toxin treatment in transgenic mice expressing the diphtheria toxin receptor in their beta cells , or (iii) following induction of an inflammatory response caused by wrapping the pancreas with cellophane [31, 32] or by pancreatic duct ligation . The mechanism of beta cell neoformation varies depending on the experimental protocol. Following pancreatic duct ligation, new beta cells are formed from precursor cells recruited from an unknown source to the pancreatic duct. When beta cells are destroyed by diphtheria toxin, their regeneration mostly results from the transdifferentiation of alpha cells . Transdifferentiation was also observed in transgenic mice expressing the transcription factor Pax4 in alpha cells. This lead to a massive increase in beta cell mass, which could be sustained over time because the disappearance of alpha cells resulted in the recruitment to duct and islets of new precursors able to differentiate into Pax-4-expressing alpha cells . Beta cell neoformation can also proceed from the dedifferentiation of exocrine cells into ductal-like cells, which can then redifferentiate into mature beta cells [35-37]. It has also been proposed that beta cells may originate in the pancreatic ducts, in which precursor cells have been located. However, the importance of this pathway for beta cell regeneration in adult mice is still debated [38-40].
Thus, there is ample evidence that new beta cells can be generated in adult animals, in response to various experimental conditions and using different mechanisms (Fig. 2). This indicates that total beta cell mass is constantly monitored and that signals are produced to induce new beta cell formation. The nature of the signals and whether they differ under the various regeneration conditions discussed above remain unknown. It is an important challenge of current research to identify genes expressed in these conditions, either by beta cells themselves or possibly also by alpha or duct cells, as well as genes that trigger beta cell neoformation.
Beta cell replication in insulin-resistant conditions
The mechanisms leading to a compensatory increase in beta cell number in insulin resistance are also not known. In the setting of obesity and insulin resistance, hyperglycaemic episodes may occur during the phase of beta cell compensation. As glucose is one of the most potent stimulators of beta cell proliferation [41, 42], it may induce compensatory beta cell growth. Glucose induces beta cell proliferation by a mechanism that requires its metabolism and closure of KATP channels [43, 44] leading to membrane depolarization and insulin granule exocytosis. This leads to the secretion not only of insulin but also of other peptides such as insulin-like growth factor (IGF)-2, which could act as autocrine regulators of the insulin and IGF-1 receptors. As beta cell expansion in genetic models of insulin resistance requires expression of the insulin receptor substrate-2 (IRS-2) in beta cells, this supports the hypothesis that regulation of beta cell mass can involve activation of the insulin or IGF-2 receptors [45, 46]. The IGF-1 receptor/IRS-2/Akt pathway has also been linked to increased beta cell glucose competence , indicating that proliferation and glucose competence may, in some situations at least, be regulated simultaneously.
It has also been proposed that secreted factors, for example released by insulin-resistant muscle, can increase beta cell proliferation . Also, parabiosis experiments carried out between control mice and mice with liver-specific knockout of the insulin receptor, which have massive beta cell compensatory expansion, induce beta cell proliferation in the control animals; this suggests that factors released by the insulin-resistant liver can stimulate beta cell proliferation . Neuronal signals may also be involved. This has been demonstrated in a model of liver insulin resistance induced by activation of the extracellular signal regulated kinase activation of the extracellular signal regulated kinases (Erk1, Erk2),(Erk1/2) kinase pathway specifically in this organ . This led to a remarkable increase in beta cell proliferation, which appears to be entirely mediated by a neuronal pathway linking the liver to the endocrine pancreas.
Inflammation of the endocrine pancreas, with infiltration of macrophages and other inflammatory cells in the islets, is a hallmark of type 2 diabetes in humans and mice . This is associated with production of cytokines, which not only involves glucose-induced interleukin (IL)-1ß production by beta cells and autocrine activation of the Fas pathway but also secretion by activated inflammatory cells [50-53]. At low levels of IL-1ß expression and Fas activation by beta cells, this signal may induce beta cell proliferation, especially when the intracellular signalling molecule Flip is expressed [53, 54]; this pathway may link initial, low-grade inflammation to adaptation of beta cell mass.
There is thus very strong evidence for the involvement of metabolic, endocrine and nervous signals in the adaptation of beta cell mass to insulin resistance in liver, muscle and fat. However, the identity of these signals, how they are generated and by which tissue(s), is still far from being understood.
Beta cell replication during pregnancy
Pregnancy is an insulin-resistant state that develops to ensure sufficient provision of glucose to the foetus. However, to preserve normoglycaemia, the beta cells of the mother undergo multiple functional changes, including increased glucose-stimulated insulin secretion, increased glucose uptake, phosphorylation and oxidation capacity  and a large increase in beta cell mass. In mice, a peak of proliferation is observed at day 14 of gestation and the maximum increase in beta cell mass, reaching ~150% of the prepregnancy mass, is observed by day 19 of gestation. Following delivery, a phase of rapid apoptosis ensues to normalize the beta cell mass [56, 57]. An increase in beta cell mass during pregnancy in humans has also been reported . In rodents, beta cell proliferation as well as functional changes leading to increased glucose-stimulated insulin secretion appears to be mostly under the control of prolactin and the placental lactogen acting through activation of the prolactin receptor (PRL-R)/Jak/STAT signalling pathway [55, 59]. This activates the transcription factor FoxM1, which induces the expression of several cell cycle regulators  but also suppresses the expression of the multiple endocrine neoplasia 1 gene (menin1), which leads to reduced expression of the cell cycle inhibitors p18 and p27 [61-63] (Fig. 3). It is interesting that activation of the PRL-R induces substantial expression of the enzyme tryptophan hydroxylase, leading to serotonin production and autocrine activation of the serotonin receptor 5HTR2B[64, 65]. Further, the role of the cell surface oestradiol receptor GPR30 in inducing beta cell proliferation has recently been demonstrated; its mechanism of signalling involves the silencing of the microRNA mir338-3p, a negative regulator of the IGF-1 receptor signalling pathway .
In humans, the normal beta cell mass expansion during pregnancy may be blunted in gestational diabetes mellitus. The cause of this impaired expansion response is not known but is certainly associated with gene variants leading to an improper proliferation response to the pregnancy hormones. As gestational diabetes is associated with increased risk of developing type 2 diabetes later in life [67, 68], this suggests that the adaptive mechanisms that lead to beta cell proliferation in response to the transient insulin resistance of pregnancy also play a role in the life-long adaptation of beta cell mass and function. Identification of the genes conferring susceptibility to gestational diabetes mellitus would be of great interest. Currently available evidence suggests the participation of the already identified type 2 diabetes genes CDKAL1 and MTNR1B .
Gluco-incretins and regulation of beta cell mass and function
The gluco-incretin hormones glucagon-like peptide-1 (GLP-1) and gastric inhibitory polypeptide (GIP) have direct impact on the function of the pancreatic beta cells by binding to specific receptors located on their cell surface [69, 70]. Binding triggers intracellular signalling mechanisms initiated by the production of cAMP and activation of protein kinase A and Epac2, a cAMP-binding protein . The immediate effect of these events is the potentiation of glucose-induced insulin secretion [72, 73]; this is an important control mechanism as it is estimated that gluco-incretin action on beta cells is responsible for ~50% of insulin secreted in the absorptive phase . This acute effect of GLP-1, but not GIP, is preserved in patients with type 2 diabetes, although supraphysiological concentrations of GLP-1 are needed to trigger insulin secretion and normalize glucose levels in the blood . Nevertheless, various GLP-1 receptor agonists, as well as inhibitors of the enzyme dipeptidylpeptidase-4 (which rapidly inactivates endogenous GLP-1), have most recently been introduced for the treatment of type 2 diabetes .
Besides this acute insulinotropic effect, both GLP-1 and GIP also have trophic effects leading to increased beta cell proliferation [77-79], protection against cytokine- and glucolipotoxicity-induced apoptosis [80, 81] and increased glucose competence [82, 83]. In rodents, these effects combine to effectively increase beta cell mass and even protect beta cells against autoimmune destruction in the NOD mouse model of type 1 diabetes [84, 85].
As shown in Fig. 4, several intracellular signalling pathways are activated downstream of the initial cAMP production, which combine to control the trophic actions of GLP-1. First, the classical cAMP/protein kinase A/cAMP response element-binding protein (CREBP) pathway controls the expression of many genes. Secondly, activation of the Erk1/2 pathway requires simultaneous Ca2+ uptake (induced by high extracellular glucose) and release of Ca2+ from the endoplasmic reticulum. Thirdly, the IRS-2/Pi3K/Akt pathway plays a major role in protecting beta cells against apoptosis, inducing proliferation and increasing glucose competence [86-88]. Fourthly, Cornu et al. [83, 89] showed that the trophic actions of GLP-1 were dependent on increased expression of the IGF-1 receptor and its autocrine activation by IGF-2 produced by the beta cells. This autocrine loop controls beta cell plasticity and transmits the GLP-1 signal. Finally, a role of the Wnt signalling pathway in GLP-1 action has also been suggested. This is activated by stabilization of ß-catenin secondary to activation of the Akt, Erk1/2 and PKA pathways. This leads to expression of transcription factor 7-like 2 (TCF7L2) , a diabetes susceptibility gene , which controls glucose-stimulated insulin secretion, in part by regulating GLP-1 receptor expression [91-93].
Whether GLP-1 and GIP have similar trophic effects on human beta cells is unclear. Good evidence supports a role for GLP-1 in protecting against cytokine- and glucolipotoxicity-induced apoptosis . However, attempts to induce human beta cell proliferation in vitro using GLP-1 have so far been disappointing [94, 95], and there is an urgent need to determine whether mouse and human beta cells respond similarly to the action of gluco-incretins.
Long-term treatment with GLP-1 receptor agonists is very effective in controlling glycaemia in patients with type 2 diabetes, but diabetes quickly resumes after treatment cessation. This indicates that there is no long-term improvement of beta cell function , although it was very recently reported that beta cell mass was strongly increased in the pancreas of patients with type 2 diabetes treated with GLP-1 agonists or dipeptidyl-peptidase IV inhibitors . These observations need to be confirmed, a task that is, however, particularly difficult in the absence of proper imaging techniques for in vivo assessment of beta cell mass and function.
One important observation is that the increase in mouse islet proliferation induced by GLP-1 or other growth factors is usually very modest, with 1% to ~5% of the beta cell population showing signs of progression through the cell cycle. In GLP-1-treated cells, this low level of proliferation has been linked to the induction by GLP-1 of multiple mechanisms that limit its own signalling pathway. Indeed, GLP-1-induced proliferation requires activation of the PKA/CREBP, PI3K and Erk1/2 signalling pathways. However, immediately after GLP-1 binding to its receptor, multiple suppressors of these signalling pathways are induced, including RGS2 (an inhibitor of Gsα activation and cAMP production), ICER and CREM (inhibitors of CREBP) and DUSP14 (a dual-specificity phosphatase that inactivates Erk1/2 signalling) . Knockdown of these negative regulators of signalling increases GLP-1-induced beta cell proliferation.
Thus, beta cells have evolved mechanisms to limit their proliferative response to growth factors, probably because over secretion of insulin can be lethal. Therefore increasing beta cell mass may not only need to target the pathways that induce proliferation, but also those that prevent over-responsiveness to stimuli.
Proliferation, glucose competence and nutrient-regulated enzymes
Beta cells are highly sensitive to the levels of circulating nutrients, which control the acute insulin secretion response but also the long-term adaptation of beta cell mass. In recent years, several nutrient-sensing enzymes have been identified that are activated by changing levels of nutrients or of specific metabolites.
The serine/threonine protein kinase PAS kinase is a sensor of elevated glucose concentrations that has evolved from a large family of prokaryotic kinases containing the conserved Per-Arnt-Sim (PAS) sensor domain . In beta cells, activation of this kinase induces translocation of the transcription factor Pdx-1 in the nucleus and increases insulin gene transcription and glucose-stimulated insulin secretion [100, 101]. Thus, PAS kinase is a regulator of glucose competence, and its expression is reduced in islets from type 2 diabetic individuals . It is also expressed in alpha cells and studies with gene knockout mice suggest that the role of PAS kinase in these cells is to limit glucagon gene expression and secretion .
Mammalian target of rapamycin (mTOR)
mTOR is a serine/threonine kinase that is found in two forms, mTORC1 and mTORC2, with different substrate specificities . mTORC1 has a role in the control of beta cell size and proliferation, in response to branched-chain amino acids, and possibly also glucose, or growth factors that cause the induction of protein kinase Cζ [104-108]. Constitutive activation of mTORC1 in beta cells by genetic inactivation of the upstream TSC1/2 regulatory genes induces increased beta cell mass and hypoglycaemia, and proliferation is associated with regulation of the cell cycle regulators cyclin D2, cyclin D3 and Cdk4. Because mTORC1 is inhibited by rapamycin, an immunosuppressive drug used in organ transplantations, treatment with this drug negatively influences beta cell mass and function [109, 110].
Mammalian sirtuins comprise a family of NAD+-dependent protein deacetylases including Sirt1, which has been extensively investigated for its role in the control of cellular metabolism and ageing [111, 112]. Sirt1 is activated by fasting, when the intracellular NAD+/NADH ratio increases or by the polyphenol compound resveratrol. Sirt1 thus regulates the activity of enzymes, transcription factors, histones and structural proteins by inducing their deacetylation. In beta cells, activation of Sirt1 leads to a coordinated increased expression of Glut2, glucokinase, Pdx-1, (pancreatic and duodenal homeobox 1), HNF1α, (HNF1a : hepatic transcription factor 1) and Tfam (: transcription factor A, mitochondrial UCP2 : uncoupling protein 2), and suppression of UCP2 expression, resulting in increased ATP production and glucose-stimulated insulin secretion [113-115].
An important action of Sirt1 is to deacetylate the tumour suppressor gene LKB1, an upstream regulator of AMP kinase. When deacetylated, LKB1 phosphorylates and activates AMP kinase and several AMP kinase-related kinases . Genetic inactivation of LKB1 induces a massive increase in beta cell mass and loss of cellular polarity [117-119], effects that are most probably due to inactivation of several kinases as genetic inactivation of AMP kinase does not induce beta cell proliferation.
Of interest, a mutation in Sirt1 was recently identified in a family with a history of type 1 diabetes. Cellular studies of the Sirt1 mutant showed that its expression in beta cells caused increased nitric oxide and cytokine production, suggesting a possible role in beta cell destruction in these patients .
This is an evolutionarily conserved kinase that acts as a sensor of low-nutrient conditions and is particularly activated during hypoglycaemia [121, 122]. It is a trimeric protein composed of one of two α subunits (α1 or α2), one of two β subunits (β1 or β2) and one of three γ subunits (γ1, γ2 or γ3). Activation of AMP kinase depends on an increase in the intracellular AMP/ATP ratio, but full activity requires further phosphorylation of the α subunit on threonine 172 by the upstream kinase LKB1, itself regulated by deacetylation by Sirt1, or by CamKK1, a protein kinase activated by Ca2+. AMP kinase is also activated by the antidiabetic drug metformin, and therefore, it is important to understand its role in beta cell function. Unfortunately, this role is currently debated with several studies demonstrating that activation of AMP kinase increases glucose-stimulated insulin secretion, whereas others show the opposite, as comprehensively reviewed recently . One difficulty in studying the physiological role of AMP kinase in beta cell biology is that this enzyme is activated when glucose levels fall well below the normoglycaemic level. It is thus difficult to understand how it can acutely regulate glucose-stimulated insulin secretion. It may rather be an important sensor of hypoglycaemia or of nutrient deprivation that affects long-term adaptation of beta cells to these challenging conditions. Although an important sensor of energy status, its precise role in beta cell biology remains to be understood.
Summary and future challenges
Beta cells can display a marked plasticity under physiological conditions, with modulation of both number and glucose competence. Type 2 diabetes results when this plasticity fails to compensate for the developing insulin resistance, possibly initiated by a defect in glucose competence followed by a decrease in beta cell number. There is now extensive knowledge of the pathways controlling beta cell proliferation, yet insufficient to develop rational ways to increase beta cell mass. In particular, the diversity of mechanisms that limit beta cell proliferation remains poorly understood. It is striking that all the stimuli that have been reported to increase beta cell proliferation have similar modest effect, suggesting that the mechanisms limiting proliferation are very potent. More investigations of these mechanisms are required to enable manipulation of beta cell mass.
It is also important to note that most of our present knowledge is derived from the study of rodent beta cells and it is not clear that human beta cells will behave in exactly the same way. It is thus critical to study human beta cells from normal individuals and patients with diabetes. Recently generated human beta cell lines can also provide increased understanding of human beta cell biology.
Reliable imaging techniques, which would allow in vivo visualization of beta cells and assessment of their secretion capacity, are still lacking to study the pathogenesis of type 2 diabetes and the response to therapeutic treatments. Intensive research activities are ongoing to develop multiple modes of beta cell imaging, and some lines of investigations are already producing interesting results as discussed in recent excellent reviews [124-126].
Finally, when the signalling pathways controlling beta cell proliferation and glucose competence are fully elucidated, two challenges will remain to understand (i) how individual genetic variability impacts on beta cell cell function and susceptibility to deregulation by various metabolic stresses and ageing, and (ii) how these pathways can be targeted by pharmacological intervention, nutrition or exercise. Based on the great advances made in recent years and the importance of current challenges to improve health, there is clearly a need for strong commitments from the research community and funding bodies to better support adult beta cell research to design rational and long-term ways to preserve the insulin secretion capacity of the endocrine pancreas.
Conflict of interest statement
No conflicts of interest to declare.
Work in the author's laboratory has been supported by grants from the Swiss National Science Foundation (3100A0-113525), the National Center of Competence in Research ‘Frontiers in Genetics’ and the Innovative Medicine Initiative Joint Undertaking under grant agreement no. 155005 (IMIDIA), resources of which are composed of financial contribution from the European Union's Seventh Framework Programme (FP7/2007–2013) and EFPIA companies in kind contribution and European Union's Seventh Framework Programme Integrated Project BetaBat.