A differentiated β-cell results not only from cell-specific gene expression, but also from cell-selective repression of certain housekeeping genes. Indeed, to prevent insulin toxicity, β-cells should handle insulin stores carefully, preventing exocytosis under conditions when circulating insulin is unwanted. Some ubiquitously expressed proteins would significantly jeopardize this safeguard, when allowed to function in β-cells. This is illustrated by two studied examples. First, low-Km hexokinases are disallowed as their high affinity for glucose would, when expressed, significantly lower the threshold for glucose-induced β-cell function and cause hypoglycaemia, as happens in patients with β-cell tumours. Thus the β-cell phenotype means not only expression of glucokinase but also absence of low-Km hexokinases. Secondly, the absence of MCTs (monocarboxylic acid transporters) in β-cells explains the pyruvate paradox (pyruvate being an excellent substrate for mitochondrial ATP production, yet not stimulating insulin release when added to β-cells). The relevance of this disallowance is underlined in patients with exercise-induced inappropriate insulin release: these have gain-of-function MCT1 promoter mutations and loss of the pyruvate paradox. By genome-wide ex vivo mRNA expression studies using mouse islets and an extensive panel of other tissues, we have started to identify in a systematic manner other specifically disallowed genes. For each of those, the future challenge is to explore the physiological/pathological relevance and study conditions under which the phenotypically disallowed state in the β-cell is breached.
Specifically repressed metabolic housekeeping genes in the β-cell
As often stated in the world of islet biologists, the β-cell is a specialized cell type, unique in its physiological responsibilities, remarkably distinct from other cells in terms of metabolic control and very relevant if we want to understand diabetes. The implicit assumption of a mechanism that would explain this specialized phenotype is that β-cells use in a cell-specific manner a set of genes that are collectively responsible for the required functions. Examples of such genes encode secreted peptides such as insulin and IAPP (islet amyloid polypeptide), transporters such as the secretory granule zinc transporter ZnT8  and the tissue-specific transcription factors Pdx1 (pancreatic and duodenal homeobox 1), Nkx6.1 (NK6 homeobox 1) and Pax6 (paired box 6). In the present paper, we add another phenomenon to this classical viewpoint on the β-cell phenotype, namely the necessity to exclude expression of a small number of genes that are ubiquitously expressed elsewhere and mediate housekeeping functions in other tissues. Thus the complementation between specifically expressed and specifically disallowed genes creates a new perspective on genetic programming of differentiated cellular phenotypes, and, as will be illustrated by the case of exercise-induced hyperinsulinism, is a newly discovered mechanism of human disease.
Hexokinase isoforms and the triggering setpoint of the glucose sensor
The most studied examples of disallowed genes in β-cells are the low-Km hexokinase isoenzymes. Hexokinase activity is essential for glycolysis as it catalyses the first step, the phosphorylation of glucose into glucose 6-phosphate . Tissue-specific expression of different hexokinase isoforms, which differ in allosteric regulation, subcellular localization and metabolic flux, contributes to the appropriate glycolytic flux in different tissues and cell types . Indeed, major differences exist between these isoforms at the level of enzyme kinetics and regulatory properties, so that one can distinguish low-Km hexokinases (HK I, II and III; Km for glucose of 7–150 μM) from high-Km hexokinase (HK IV, also called glucokinase; Km of ∼8 mM) [2,3]. Moreover, in contrast with low-Km hexokinases, glucokinase has co-operative kinetics for glucose as a substrate. The difference in affinity for glucose is very relevant for metabolic physiology, as cells expressing HK I, II or III, have saturated enzymatic flux when glucose concentrations exceed 1 mM, whereas cells expressing glucokinase only can maintain a flux of glucose phosphorylation that is proportional to glucose concentrations in the physiological range (5–10 mM) . This crucial difference, in addition to allosteric properties, and the fact that glucose transport is not rate-limiting for glycolysis in the β-cell , makes expression of glucokinase essential for the ‘glucose sensor’ in β-cells, i.e. the molecular mechanism that detects how much glucose is present in the circulation in order to allow regulation of how much insulin is released . The relevance of this important aspect of β-cell function has been supported at different levels. First, inactivating mutations in the glucokinase gene with partial loss of enzyme activity in the β-cell (caused by decrease in Vmax or increase in Km) is the basis of MODY (maturity-onset diabetes of the young) Type 2 diabetes . Secondly, mutations that increase the affinity of glucose for glucokinase, causing high enzymatic flux at basal glucose levels, cause inappropriate basal hyperinsulinaemia and hypoglycaemia . Finally, pharmacological activators of glucokinase facilitate glucose-induced insulin release .
In fact, one could argue that the differentiated ‘glucose-sensing phenotype’ of β-cells is not only explained by specific expression of glucokinase, but also relies on profound repression of the three other hexokinase genes (Figure 1A). Several arguments support this idea. In the first place, normal primary β-cells do repress in an almost perfect manner the expression of low-Km hexokinases (Figure 1). In whole islets of Langerhans, where a substantial low-Km component can be measured in addition to high-Km glucokinase, this perfection is not fully appreciated (Figure 1). However, (most of) this expression is a false-positive result explained by high levels of low-Km hexokinases (HK I–III) in non-β-cells. Indeed, detailed analysis of the cellular composition of FACS-purified β-cell preparations showed that most low-Km hexokinase activity can be attributed to contaminating non-β-cells, especially acinar cells (Figure 1) . A second argument is forced violation of low-Km hexokinase repression, which leads to abnormal β-cells that are no longer capable of sensing glucose. For instance, transient overexpression of mRNA encoding low-Km HK I in MIN6 cells  and in primary β-cells  resulted in an increased basal glycolytic flux, elevated basal insulin release and poor responsiveness to elevated glucose. When we accept these two arguments, it means that the glucose-sensing phenotype of the differentiated β-cell, conditioned by a sufficiently low hexokinase/glucokinase activity ratio, is possible via profound low-Km hexokinase gene repression. This means that conditions that favour such repression are important to achieve the differentiated β-cell phenotype and it may be of interest to pay attention to this concept in developmental studies of the endocrine pancreas. In previous work, the ratio was closely monitored in efforts to make glucose-responsive clonal β-cells from insulin-producing tumours [11,12] or to explain the passage number-dependent progressive loss of glucose-responsiveness of βTC7-cells . It could also mean that imperfect repression of low-Km hexokinase expression may occur in metabolically stressed β-cells even before failure occurs. Several examples of a shift in phenotype towards abnormally high hexokinase over glucokinase have been observed. One example is 85–95% pancreatectomy in rats in which the remaining β-cells are forced to cope with metabolic demands; under these stressed conditions, some HK I is expressed, while the normally allowed glucokinase is down-regulated [14,15]. Other examples are the ZF (Zucker fatty) and ZDF (Zucker diabetic fatty) rats. These strains are hyperinsulinaemic and have increased islet hexokinase activity compared with lean Wistar rats  and ZDF lean controls . Similar observations were made in DBA/2 mice . In these rodent models, hyperglycaemia was one of the plausible factors responsible for changes in β-cell gene expression . Dietary conditions during pancreatic development may also influence this ratio as suggested by Jimenez-Chillaron et al.  in a model of maternal malnutrition during pregnancy. Offspring from mothers that underwent 50% caloric restriction during days 12.5–18 of pregnancy developed postnatal glucose intolerance. The underlying mechanism was attributed to poorly differentiated β-cells that exhibited loss of glucokinase activity, strong up-regulation of low-Km hexokinase activity and loss of glucose-induced insulin release . Although the exact molecular mechanisms that are responsible for this altered path of β-cell differentiation is unknown, the excitement of this discovery is that environmental factors both during fetal and adult life can influence the delicate balance between allowed and disallowed genes to shape the glucose sensor of β-cells.
Disallowed low-Km hexokinase mRNA expression in β-cells
Monocarboxylic acid (lactate/pyruvate) transporters and exercise-induced insulin release
Another example of genes of which the expression is disallowed in the β-cells is the family of MCTs (monocarboxylic acid transporters) and LDH (lactate dehydrogenase) isoforms. Members of the MCT family, of which eight have now been identified, are ubiquitously expressed and mediate pyruvate/lactate transport across the plasma membrane . Pancreatic β-cells are special in that most glycolytic glucose carbon that enters metabolism via the glucose-sensing step is transported as pyruvate into the mitochondria [21,22]. Glycolytic pyruvate is an excellent substrate for β-cell mitochondria, for oxidative decarboxylation to support ATP synthesis and for carboxylation to sustain cataplerosis [21,22]. However, in contrast with the intracellular fate of glycolytic pyruvate, extracellular pyruvate added to whole isolated rat and mouse islets is well metabolized, but unable to stimulate insulin secretion . This so-called ‘pyruvate paradox’ could be explained by the presence of non-β-cells in the islets, which might be responsible for uptake and subsequent metabolism of extracellular pyruvate [24,25], an interpretation analogous to the low-Km hexokinase activity measured in islets . Absence of pyruvate uptake in β-cells would explain why this metabolite cannot stimulate insulin release when added to intact β-cells. This idea is supported by using hydrophobic membrane-permeable methyl pyruvate which does not need solute transporters and which is an excellent insulin secretagogue . That membrane transport in β-cells is the reason for the poor pyruvate stimulation of insulin is supported further by low levels of lactate transport activity in FACS-purified β-cells  and low expression of MCT1 and the other MCT isoforms [27,28] (Figure 2). Final evidence for the explanation of the paradox was given by overexpression of MCT1 in isolated rat islets which was sufficient to induce an insulin secretory response to pyruvate . For lactate, however, which is also a poor insulin secretagogue [29,30], an additional metabolic hurdle is very low LDH expression . Therefore double overexpression of both MCT1 and LDH was required to obtain robust lactate-induced insulin secretion from islets . Single overexpression of LDH in MIN6 cells attenuated glucose-stimulated insulin release , which might be explained by low, but detectable, expression of MCT1 and lactate efflux in MIN6 cells. In fact, as illustrated by detectable levels of HK III (Figure 1A) and MCT1 (Figure 2), even the most differentiated β-cell lines could reflect the metabolic needs for growth and replication which involve anaerobic and anabolic modes of glycolysis . In fact, for several aspects discussed above, there is a parallelism between suppression of low-Km hexokinase and LDH/MCT1 isoforms in β-cells as it explains the unusual metabolic organization of β-cells [8,22] to serve physiological control of insulin release.
Disallowed MCT1 mRNA expression in mouse islets
What genetic programming could be responsible for the suppression of LDH and MCT in normal β-cells? For LDH, it may be normal blood glucose, as suggested by an elegant series of recent islet transplantation studies . When a sufficient amount of islets are transplanted in a diabetic rat, normal glucose is restored; however, when insufficient islets are transplanted, hyperglycaemia remains. Laybutt et al.  used this model to examine the islet phenotype after chronic exposure to normoglycaemic compared with diabetic conditions in vivo. Compared with freshly isolated islets and islets in a non-diabetic environment, the transplanted islets that had been exposed to hyperglycaemia for 2.5 weeks were characterized by loss of insulin, Pdx1 and Nkx6.1 expression and strong up-regulation of Ldh mRNA. The precise transcriptional or mRNA-(de)stabilizing event that is responsible for this shift in gene expression remains to be discovered.
In all tissues other than in the pancreatic islet, the expression of MCT1 and LDH contributes to flexibility of metabolism, for instance by switching from aerobic to anaerobic glycolysis or by using extracellular lactate as a substrate for ATP synthesis. This knowledge leads to a number of questions. Why would β-cells avoid expressing a housekeeping gene such as MCT1 and what could be the benefit of a profound disallowance of MCT1 in β-cells? Might disallowance be related to the fact that β-cells have a glucose sensor which might not work in the presence of MCT1? Without having complete answers for these questions, patients with EIHI (exercise-induced hyperinsulinism)  have given us the first glimpse of a rationale. The hallmark of the EIHI phenotype is inappropriate insulin release followed by hypoglycaemia after a short period of anaerobic exercise. As in non-affected individuals, physical exercise by the patients causes a rapid rise of blood lactate and pyruvate [33,34]; however, EIHI patients respond to exogenous pyruvate with increased insulin secretion, and hence have lost the pyruvate paradox. In line with what was discussed above, a likely disease mechanism is allowance of pyruvate transport into β-cells. To identify the gene responsible for this genetic disorder, the coding regions of all pyruvate transporters were sequenced, but no polymorphism was linked to the disease . In a genome-wide linkage analysis, however, several mutations within the promoter region of MCT1 were found to be associated with the EIHI phenotype . This implies that regulation of the MCT1 gene, rather than changes in the encoded protein, is responsible for the phenotype. This idea was supported by MCT1 promoter studies in vitro . Of course, the final proof of the disease mechanism would be to directly study MCT1 expression and pyruvate influx into β-cells (uncontaminated by exocrine or non-β-cells) from patients with EIHI, which is virtually impossible. Generation of mice ‘knocked-in’ for the variant human SLC6A1 (solute carrier 6A1) genes may provide a more accessible alternative. Nevertheless, the combination of molecular genetic and metabolic data in patients, together with the expression profile of MCT1 in islets compared with other cell types, indicates strongly that the disease is based on inappropriate pyruvate uptake in β-cells (Figure 3). Because of the metabolic programming of these cells, distal metabolic events that occur after a rise in blood glucose are mimicked, and insulin is released even when blood glucose is low. As a rise in plasma pyruvate occurs in all animals in situations that are essential for survival (e.g. fight or flight situations), this β-cell-selective disallowance is predicted to be strongly conserved in evolution. Figure 3 also suggests another explanation of why both MCT1 and LDH should be disallowed in β-cells: as neighbouring islet and non-islet cells utilize glucose in part via anaerobic metabolism, lactate efflux is a normal (housekeeping) consequence of basal metabolism. When this lactate would have access to metabolic flux in β-cells, a local noise on measuring blood glucose would be the result, and the glucose sensor would be distracted from its function.
Physiological rationale for disallowed low-Km HK I, MCT1 and LDH expression in β-cells
A systematic search for other disallowed housekeepers
Guided by these two examples, we have developed a methodology that allows identification of other disallowed genes on a genome-wide basis. The strategy is based upon ex vivo mRNA isolation from mouse islets in addition to a great number of other tissues (approx. 25 at the time of writing), followed by mRNA expression microarray analysis. Thus it is possible to filter on a statistical basis for genes that are ubiquitously expressed except in the pancreatic islet. A major challenge will be for other specifically disallowed genes to identify the physiological or pathological rationale, e.g. by studying transgenic mice in which the phenotypically disallowed state in the β-cell is violated, or by searching for other genetic disorders equivalent to EIHI. A second challenge will be to find a molecular mechanism for the process of expression disallowance. As the encoding mRNA is very low, the most likely mechanism is very inefficient mRNA transcription, accelerated mRNA breakdown or a combination of both. The former can be explained by β-cell-specific transcriptional repressors and/or chromatin remodelling factors that prevent transcription in a cell-specific manner. We have noted that, by taking one of the other tissues in the reference panel as baseline and repeating the statistical filtering procedure that we used for islets, other specifically disallowed genes can be found, for instance in liver and adipose tissue. Therefore, although we like to cherish the thought that the β-cell is a very special cell type, the contribution of disallowed expression of housekeeping genes to the accomplishment of a specialized phenotype may have a broader significance.
Pancreatic β-Cell: Birth, Life and Death: A joint Biochemical Society, Juvenile Diabetes Research Foundation (JDRF) and EU ‘SaveBeta’ Consortium Focused Meeting held at King's College London School of Medicine, London, U.K., 3–4 December 2007. Organized and Edited by Stephanie Amiel (King's College London, U.K.), David Dunger (University of Cambridge, U.K.), Decio Eizirik (Université Libre de Bruxelles, Belgium), Peter Jones (King's College London, U.K.), Jo Lilleystone (Juvenile Diabetes Research Foundation, U.K.), Guy Rutter (Imperial College London, U.K.), James Shaw (Newcastle University Medical School, U.K.) and David Tosh (Bath, U.K.).
Our research is supported by the Research Foundation Flanders (FWO-Vlaanderen; grant G.0529.05 and postdoctoral fellowship to K.L.) by the Juvenile Diabetes Foundation (grant 1-2006-182) and by the Katholieke Universiteit Leuven (grant GOA 2004/11).