Effects of Diazoxide on Gene Expression in Rat Pancreatic Islets Are Largely Linked to Elevated Glucose and Potentially Serve to Enhance β-Cell Sensitivity

Diazoxide (Hyperstat) was from Schering-Plough (Labo, Heist-op-den-Berg, Belgium). Hank's balanced salt solution and RPMI 1640 were purchased from SVA (National Veterinary Institute of Sweden, Uppsala, Sweden).

Male Sprague-Dawley rats were purchased from Scanbur (Sollentuna, Sweden). The ethical guidelines of the Karolinska Institute for the care and use of laboratory animals were followed. The rats were maintained in a 12-h (0600–1800 h) light/dark cycle with free access to water and standard diet. They weighed 250–350 g at the time they were used for experiments. Islets of Langerhans were isolated by collagenase (Roche Diagnostics) digestion in Hank's balanced salt solution basically as described previously (2) followed by sedimentation. Islets were then selected under a stereomicroscope and transferred to Petri dishes (Sterilin, Teddington, U.K.) containing RPMI-1640, 2 mmol/l glutamine, 10% (vol/vol) FCS, 100 units/ml benzylpenicillin, 0.1 mg/ml streptomycin, and 5.5, 11, or 27 mmol/l glucose, each with or without the copresence of 325 μmol/l diazoxide. Islets were then cultured free-floating for 24 h at 37°C in an atmosphere of 5% CO₂ in air. Islets visibly free from non-islet tissue were then transferred to dishes containing 5 ml Krebs-Ringer bicarbonate medium, 10 mmol/l HEPES, 0.2% BSA, and 3.3 mmol/l glucose and preincubated for 30 min at 37°C.

Islets were collected for RNA extraction after the postculture preincubation described above, and total RNA was extracted using an RNeasy Micro kit (Qiagen). Nine, three, and four separate experiments, respectively, were performed for the 27, 11, and 5.5 mmol/l glucose with or without diazoxide conditions. Because it was difficult to obtain enough material for microarray analysis from a single experiment, we pooled islets from three experiments (each with pooled islets from two to four rats) to obtain material for each analysis of the 27 mmol/l glucose with or without diazoxide condition. In this way we obtained material for three replicate analyses. For the 5.5 mmol/l glucose with or without diazoxide condition, islets were pooled from two experiments (each with pooled islets from four rats) to obtain material. For the 11 mmol/l glucose with or without diazoxide condition, islets from four rats were pooled for each of the three experiments. In this way we obtained material for two replicate analyses. The pooled material for each analysis contained 11–22 μg total RNA. The purity and nondegraded state of the RNA was assured using a Bioanalyzer from Agilent. These total RNA samples were used to synthesize labeled cRNAs and were hybridized to a rat expression array (230A; Affymetrix, Santa Clara, CA) at the Novum Affymetrix Core Facility, Karolinska Institutet. Significance testing indicated that the likelihood of false-positive results was <10%.

An equal number of islets for each experimental condition was collected and washed twice with ice-cold PBS. Extracts corresponding to 50 islets (15 μg protein as confirmed by Bradford protein assay) were denatured in 60 μl loading buffer at 80°C for 10 min. Samples were analyzed on 10% SDS-PAGE run for 1 h at 150 V and were then transferred to nitrocellulose for 1 h at 250 mA. Membrane was blocked for 2 h at room temperature with 5% (wt/vol) fat-free milk and 0.1% Tween 20 in Tris-buffered saline, pH 7.6, and then incubated overnight at 4°C with primary antibodies at the following dilutions: 1:500 for Aldolase B (Santa Cruz Biotechnology, Santa Cruz, CA) and 1:2,000 for pancreatic and duodenal homeobox 1 (PDX1; kindly provided by Dr. Helena Edlund, Umeå University, Umeå, Sweden). After the addition of anti-IgG (anti-goat or -rabbit IgG) second antibody at a dilution of 1:5,000, membranes were incubated for 1 h at room temperature. Immunoreactive bands were visualized using an enhanced chemiluminescence kit (Amersham Biosciences), exposed to X-ray film (Hyperfilm; Amersham Biosciences), and documented with a flatbed scanner (Scanjet 5300; Hewlett-Packard) and quantitation software (1D; Kodak).

After culture and preincubation as described above, islets were incubated in groups of three for 60 min at 37°C in 300 μl Krebs-Ringer bicarbonate containing 3.3 or 16.7 mmol/l glucose, each with or without specific additives. The insulin accumulated was measured as previously described (3). Islet insulin contents were measured in batch-incubated islets after sonication for 10–15 s, followed by extraction of insulin overnight at 4°C in 200 μl acid-ethanol (70%, vol/vol).

GeneChip operating software (GCOS version 1.4; Affymetrix) was used for absolute and comparison analysis. Scaling was set to “all probe set” with a target signal of 100, and normalization was set to “user defined” with a normalization value of 1. For estimation of regulated genes, pairwise comparisons of test versus control were performed, resulting in a quantitative signal log ratio and a qualitative “change call.” The signal log ratio is the logarithmic (base = 2) ratio of intensities from test and control samples. Change call is based on the “change P value,” where GCOS settings used were P < 0.002 for “increase call” and P > 0.998 for “decrease call.”

For transcripts to be considered increased by diazoxide or glucose, an increase call and signal log ratio ≥0.5 were required, and the corresponding requirement for decrease were a decrease call and signal log ratio ≤−0.5. Furthermore, for a diazoxide effect to be present, significant effects were required in all three replicate comparisons (i.e., nine of nine) for 27 and 11 mmol/l glucose with or without diazoxide and in both replicate analysis (i.e., four of four) for 5.5 mmol/l glucose with or without diazoxide. In cases where less stringent criteria were used, this is explicitly stated in the tables and/or text. For registering a glucose effect, four to six of the six comparisons were required to show a glucose effect per se >0.5. Table 1 summarizes the comparisons performed.

For basic functional classification, records from the Gene Ontology project (www.geneontology.org) were used together with annotation information from Affymetrix (www.affymetrix.com). For the genes thereby annotated to the category of metabolism, we further explored 1) whether these genes were annotated in other categories and 2) whether information was available in the Beta Cell Gene Expression Bank (http://test.t1dbase.org). Such additional information was incorporated into tables and/or text.

Islets were cultured for 24 h in low, moderately elevated, or high (5.5, 11, and 27 mmol/l, respectively) glucose in the presence or absence of diazoxide. Insulin secretion was measured postculture and in the absence of diazoxide; also, islet insulin contents and insulin in culture media were measured (Fig. 1). As expected, diazoxide almost completely blocked insulin release into the culture media (Fig. 1A). Culture at 11 and 27 mmol/l glucose significantly reduced the insulin response to postculture stimulation with 16.7 mmol/l glucose (Fig. 1B). Culture with diazoxide upheld islet insulin contents, which, in the absence of diazoxide, were decreased by 48 and 74%, respectively, after culture at 11 and 27 mmol/l glucose (Fig. 1C). When insulin responses to 27 mmol/l glucose were calculated as percent of insulin content, there was no decrease caused by elevated glucose during culture. However, exposure to diazoxide during culture with 27 mmol/l glucose markedly increased the percentage of insulin released during postculture stimulation with glucose (Fig. 1D).

Of 15,923 genes present on the array, the expression of 8,179 genes was detected after culture at 27 mmol/l glucose, 7,539 genes after culture at 11 mmol/l, and 8,233 genes after culture at 5.5 mmol/l glucose. The corresponding figures obtained after inclusion of diazoxide during culture at 27, 11, or 5.5 mmol/l glucose were 8,360, 7,773 and 8,537 genes, respectively. Culture with 27 mmol/l glucose (27 vs. 5.5 mmol/l glucose) upregulated 303 genes (signal log ratio ≥0.5) in six of six comparisons and downregulated 320 genes (≤ −0.5). As to the expression of insulin genes (INS1 and INS2), both were upregulated (signal log ratio +0.9) in four of six comparisons at 11 versus 5.5 mmol/l glucose, whereas they were downregulated in 27 versus 11 mmol/l glucose in six of nine comparisons (signal log ratio −1.3 for INS1, −1.2 for INS2). The 20 known genes that were most highly influenced by 27 mmol/l glucose per se are shown in supplemental appendix 1, which can be found in an online appendix (available at http://dx.doi.org/10.2337/db06-0322).

Few genes were affected by diazoxide after coculture at 5.5. mmol/l glucose. Of diazoxide's effects, 92% (up- and downregulated) were observed only after coculture with 11 and 27 mmol/l glucose (Fig. 2A).

Increasing the glucose concentration during culture markedly increased the number of genes that were downregulated by diazoxide. At 5.5 mmol/l glucose, 13 genes were upregulated by glucose and 19 downregulated. At 11 mmol/l glucose, 97 genes were upregulated and 21 downregulated. At 27 mmol/l glucose, 101 genes were upregulated and 160 downregulated. Thus, of the total number of genes affected, not only the number but also the percentage of downregulated genes was markedly changed (Fig. 2B and C).

Of 101 genes upregulated by diazoxide (signal log ratio ≥0.5) only in the presence of 27 mmol/l glucose, only 8% were also upregulated by glucose (Table 2). A total of 31 (31%) of the genes upregulated by diazoxide were downregulated by 27 mmol/l glucose. Of the genes upregulated by diazoxide, 62 (61%) were not affected by 27 mmol/l glucose (Table 2). Similar findings were obtained for diazoxide effects at 11 mmol/l glucose. Thus, only 3% of genes affected by diazoxide were regulated in parallel by 11 mmol/l glucose (Table 2). None of the few genes affected by diazoxide after coculture at 5.5 mmol/l glucose were upregulated in parallel with effects of 27 mmol/l glucose per se (Table 2).

A majority of genes downregulated by diazoxide were affected in an opposite fashion by 27 mmol/l glucose or by 11 mmol/l glucose (Table 2). Of the genes downregulated by diazoxide at 5.5 mmol/l glucose, only two were downregulated in parallel with effects of 27 mmol/l glucose per se (Table 2).

As shown in Tables 3–5, diazoxide-induced alterations in gene expression were mostly moderate with only a few genes showing more than a threefold (i.e., signal log ratio 1.59) upregulation in expression level. At 27 and 11 mmol/l glucose, a larger amount of genes were downregulated more than threefold by diazoxide (Tables 3 and 4).

The annotated genes affected by diazoxide were classified according to the GeneSpring software program (Table 6). This classification shows similar percentages of genes affected by diazoxide in several of the categories. There were no obvious differences in the proportion of genes (i.e., the sum of up- and downregulated genes) affected by 11 and 27 mmol/l glucose. Interestingly, the highest percentage of genes affected by diazoxide (in percent of the number of genes represented on the array) was in the subcategory of monosaccharide metabolism (14.4%).

Because the enhancing effect of diazoxide on postculture glucose-induced insulin secretion was much more marked after coculture at 27 than at 11 mmol/l glucose (Fig. 1), it seemed of special interest to select for a diazoxide effect on genes affected only at 27 and not at 11 mmol/l glucose or genes that were affected in an opposite fashion by these two glucose concentrations. By stringent criteria, a limited number of genes, most of them downregulated by diazoxide, could be identified (Table 7). Genes potentially influencing growth, such as IGF binding protein and cyclin D, belonged to these downregulated genes. A large percentage (5 of 25 genes) belonged to the category of metabolism, which is further discussed below.

The total number of genes affected by diazoxide at 27 mmol/l in the category of metabolism are shown in supplementary appendixes 2 and 3. Of these genes, we focused first on genes for enzymes with specific effects on hexose, fatty acids, or amino acid metabolism (Table 8). Second, we focused on genes with generalized effects on ATP production and processing.

The most marked upregulation by diazoxide in the entire array was seen for aldolase B, which was upregulated 5.3-fold (signal log ratio 2.4) by 27 mmol/l glucose but 42-fold (5.4) by diazoxide (in the presence of 27 mmol/l glucose). This upregulation by diazoxide was only seen with 27 mmol/l and not with 11 mmol/l glucose (Table 8). Apart from aldolase B, only the gene coding for malic enzyme was upregulated, and by less stringent criteria (signal log ratio +0.6, eight of nine comparisons) (results not shown in table). Downregulation by diazoxide of genes affecting glycolysis included enzymes of the pentose phosphate shunt and production of lactate.

The increase by diazoxide in expression of the aldolase B gene was extraordinary and has to our knowledge not been previously reported. This prompted verification of the effects of diazoxide on aldolase B expression at the protein level. Using Western blotting, we found that culture with 27 mmol/l glucose increased aldolase B protein levels by 2.2-fold and that addition of diazoxide increased levels of this protein by another 4.3-fold (Fig. 3A). Also, there was no diazoxide effect visible when coculture was performed together with 11 mmol/l glucose (mean of two experiments, results not shown). Thus, diazoxide appears to regulate aldolase B protein levels in parallel with its effects on aldolase B mRNA levels.

Regarding diazoxide-induced changes in genes related to fatty acid and amino acid metabolism, most of these did not parallel effects induced by glucose per se (Table 8). The expression changes in genes involved in fatty acid metabolism followed a clear pattern, insofar as they would serve to inhibit fatty acid β-oxidation and promote fatty acid synthesis. Regarding the diazoxide-induced changes in genes related to amino acid metabolism, diazoxide inhibited expression of the GAD gene, a gene of possible importance in the etiology of type 1 diabetes (4). This effect was observed at all glucose concentrations. A marked effect on argininosuccinate synthetase in the presence of 27 and 11 mmol/l glucose was also noted.

PDX1 (additionally annotated as transcription factor) was moderately upregulated by diazoxide at all glucose concentrations. Downregulation of PDX1 has been coupled to desensitization of β-cell function, in particular glucose-induced insulin secretion (5). It was therefore of special interest to verify gene effects on the protein level. Western blotting (Fig. 3B) confirmed the effects of diazoxide on the protein level. Also, NKX6.1 (Table 8), a gene of importance for β-cell growth and function (5,6), was upregulated by diazoxide at 27 mmol/l glucose.

Other genes affecting mitochondrial functioning and hence ATP production included fumarate hydratase 1 and uncoupling protein-2 (UCP-2). In the case of UCP-2, this gene was downregulated by diazoxide during coexposure both with 5.5 and 27 mmol/l glucose.

The most striking finding of this study was the intimate relationship between glucose and diazoxide effects on islet expression. Thus, >90% of the effects of diazoxide that were detected by microarray were seen only after coculture with an elevated concentration (11 and 27 mmol/l) of glucose. To our knowledge, such a strong coupling with glucose has not been described for the effects of other agents on gene expression.

The glucose dependency of diazoxide's effects on mRNA levels parallels to some extent the effects of diazoxide on insulin secretion previously described (1) and confirmed and extended here. Thus, diazoxide augments postculture glucose-induced insulin secretion only when the culture is performed in the presence of an elevated concentration of glucose. Furthermore, the enhancing effect of diazoxide on glucose-induced insulin secretion was much stronger at 27 than at 11 mmol/l glucose, and this corresponds to an increase in diazoxide-regulated genes, in particular downregulation.

How to define desensitization is a debated issue. One may argue that our insulin data do not demonstrate desensitization of glucose-induced insulin secretion because secretion was not reduced if one expresses results as percent of islet insulin contents. However, diazoxide unquestionably enhanced postculture-induced insulin secretion after coculture with 27 mmol/l glucose. Regardless of semantics, it seems important to investigate mechanisms behind this large and apparently beneficial effect.

We have previously attributed the beneficial effects of diazoxide on insulin secretion to β-cell rest, i.e., alleviation of negative effects of profound and prolonged stimulation. The possibility of overstimulation was suggested by the partial depletion of insulin stores (7 and current study), reflecting a disparity between demand and capacity. Severe overstimulation could increase endoplasmic reticulum stress, as shown in other experimental systems (8). A reduction of protein biosynthesis is an early sign of endoplasmic reticulum stress (8). Our data do suggest some reduction of preproinsulin mRNA when the glucose concentration was increased from 11 to 27 mmol/l. However, we did not find upregulation of the stress-sensitive genes CHOP, P8, GRP94, EDEM1, ATF4, RAMP4, or caspase 12 by 27 mmol/l glucose nor any effects on these genes by diazoxide (results not shown). Two exceptions were FKBP11 and P58, which have been linked to endoplasmic reticulum stress. FKBP11 was upregulated by glucose (signal log ratio 1.6, six of six comparisons) and downregulated by diazoxide (signal log ratio −0.9, nine of nine comparisons). P58 was also upregulated by glucose (signal log ratio 0.9, five of six comparisons). Taken together, our data do not support a major role for endoplasmic reticulum stress behind the functional effects of overstimulation that have previously been outlined (1).

In a recent microarray study (9), MIN6 cells were analyzed for short-term (45 min) effects of elevated glucose in comparison with KCl, which stimulates secretion bypassing ATP-dependent K⁺ channels. Similarity was found for these early effects on mRNA for glucose and KCl; furthermore, genes upregulated by glucose were inhibited by diazoxide in separate PCR experiments. These findings provide proof of concept that stimulated insulin secretion per se can lead to gene induction, at least in a short-term perspective, and that blocking secretion with diazoxide can nullify such induction. The genes deduced to be affected by the secretory process in the study by Ohsugi et al. (9) were not affected in this study (results not shown), with the exception of EGR1 and SFRS5, which were upregulated by diazoxide at 27 mmol/l glucose (signal log ratio 0.8 and 0.6, respectively, in nine of nine comparisons).

It remains possible that the observed coupling of diazoxide's effects to the permissive effect of elevated glucose could, in part, be secondary to one or several aspects of the insulin secretory process. For example, because insulin secretion requires energy, it is plausible that glucose metabolism is adjusted for lesser energy demands when secretion is blocked by diazoxide. Such putative adjustment could then secondarily alter expression of genes related to metabolism.

Could diazoxide exert effects through its effects on ambient insulin levels? The possibility of a feedback loop between secreted insulin and β-cell function has been proposed and debated for many years. Because of divergent experimental results (10,11), the issue is not settled. Under experimental conditions similar to the current ones, we did not observe any regulation of islets cultured with diazoxide in the presence of exogenous insulin with respect to the subsequent enhancement of glucose-induced insulin secretion (7). Furthermore, in the current experiments, there was no major difference in the concentration of insulin in culture media after culture with 11 versus 27 mmol/l glucose, even though the influence by diazoxide on gene expression was vastly different. Hence, the reducing effects of diazoxide on insulin levels in culture media would probably not factor into the current results.

Previous observations indicate that diazoxide exerts primary effects on metabolism in different tissues, including β-cells. Diazoxide enhances ischemic preconditioning in the heart (12) and neurons (13,14) through a mitochondrial interaction that may be coupled to reduction of mitochondrial reactive oxygen species production (14). In β-cells diazoxide decreases mitochondrial membrane potential in acute experiments (15,16) and may inhibit succinate oxidation (17). We did not observe any effect on gene expression for succinate dehydrogenase, represented in the array by four genes encoding for separate subunits. On the other hand, we observed an inhibitory effect on the enzyme fumarate hydratase 1, whose action is sequential to that of succinate dehydrogenase and could produce a similar inhibitory effect on mitochondrial metabolism. Inhibition of mitochondrial metabolism could serve to reduce reactive oxygen species produced by elevated glucose in β-cells (18) and could thus be a potentially beneficial effect of diazoxide in β-cells.

Thus, we envisage that the permissiveness of elevated glucose for effects of diazoxide is caused by both an inhibitory effect of diazoxide (during its presence) on glucose-induced insulin secretion and by a modulating influence of the mitochondrial state of respiration. The latter statement we base, first, on evidence, albeit not conclusive, that diazoxide interacts directly with mitochondrial metabolism and, second, on general knowledge of the importance of the mitochondrial state of respiration for a modulatory influence of drugs and other agents. In line with the latter notion, we find that culture with diazoxide exerts opposite effects on mitochondrial membrane potential after coculture at high versus low glucose (unpublished observations).

Upregulation of gene expression for aldolase B constituted the most marked effect of diazoxide in this microarray, and this effect was confirmed on the protein level. The effect was not seen at low glucose nor at 11 mmol/l glucose. Mathematical modeling of β-cell glycolysis indicates that upregulation of aldolase B by glucose is crucial for normal metabolic oscillations in β-cells (19). Such oscillations may in turn drive oscillations of glucose-induced insulin secretion (20). A stimulatory effect by diazoxide on aldolase B activity could then serve to uphold and increase metabolic oscillations, which, in turn, regulate oscillations of insulin secretion. Indeed, data from human (21) as well as from rat pancreatic islets (our unpublished observations) show that diazoxide can preserve the amplitude of insulin oscillations otherwise decreased during prolonged exposure to elevated glucose.

Also, the effects on fatty acid metabolism that we observed could potentially be beneficial. It is well established that long-term excess of fatty acid metabolism can negatively affect glucose-induced insulin secretion (22). The effects of diazoxide on fatty acid metabolism could be thought to improve glucose-induced insulin secretion. Thus, diazoxide inhibited the gene expression of several enzymes that participate in β-oxidation of fatty acids (including, by less stringent criteria, 2,4 dienoyl CoA reductase 1, signal log ratio −0.6 in seven of nine comparisons). These effects correspond to the finding that diazoxide decreases oxidation of fatty acids in islets (23). Reciprocally, genes promoting lipid synthesis, such as acetyl-CoA carboxylase, were upregulated in the current array. Mechanistically, it may be relevant that the expression of solute carrier family 25, member 10 was also upregulated by diazoxide, albeit by less stringent criteria (signal log ratio +1.1, seven of nine comparisons). This gene was recently shown in adipose tissue to transport citrite out of mitochondria in exchange for malate, the citrate then participating in lipid biosynthesis (24).

Desensitization of β-cells by hyperglycemia in vivo involves downregulation of NKX6.1 and PDX1 and upregulation of lactate dehydrogenase (5) genes. Here, we demonstrate that diazoxide exerts the opposite effects. Expression of both NKX6 and PDX1 may be necessary not only for normal β-cell growth and maturation, but also for normal insulin secretion (25,6). Interestingly, mitochondrial transcription factor A (TFAM) was, by less stringent criteria, also upregulated by diazoxide together with 27 mmol/l glucose (signal log ratio +0.63, eight of nine comparisons). This factor translocates to the mitochondria, where it controls mitochondrial gene transcription (26) and potentially regulates the effect of PDX1 on mitochondria-coded genes (25). In the case of lactate dehydrogenase, enzyme activity and lactate production is low in glucose-sensitive β-cells (27). This ensures that a rise in ambient glucose leads to increased production of ATP and possibly other metabolic signals for secretion through a tight coupling between glycolysis and oxidative phosphorylation. Increased production of lactate will then attenuate a glucose-induced signal for insulin secretion, and halting this process by diazoxide should produce the opposite effect.

The diazoxide effect on cAMP response element modulator (CREM) could constitute another beneficial effect. The CREM gene was recently shown to be upregulated both by glucotoxic and lipotoxic conditions in vitro (28); furthermore, forced overexpression of the gene was inhibitory on glucose-induced insulin secretion. We confirm that glucose upregulates the CREM gene and additionally demonstrate that diazoxide exerts the opposite effect.

Diazoxide also inhibited gene expression of UCP-2, and this effect was not glucose dependent. Such effects were recently demonstrated by PCR for gene expression (23) and also on the protein level (our unpublished observations). Because fatty acid metabolism induces UCP-2 (29), it is conceivable that the UCP-2 and fatty acid effects shown here are interlinked. As to functional effects, previous studies have implicated UCP-2 as a negative factor for glucose-induced insulin secretion (30). However, this notion has been debated (31), and it is not clear to what extent the current effects by diazoxide on UCP-2 are implicated in insulin secretion.

The inhibitory effect by diazoxide on gene expression of GAD is also interesting from a functional and pathophysiological point of view. The GAD enzyme is a key enzyme in the pathway leading to GABA synthesis; furthermore, antibodies against GAD are at least a marker and possibly a causative factor in the etiology of autoimmune diabetes (4). In agreement with our findings, diazoxide has previously been shown to decrease the efflux of GAD from rat pancreatic islets (32). Furthermore, diazoxide was reported to decrease the production of GAD protein (33). However, an effect by diazoxide on GAD gene expression has not previously been documented. Another enzyme of interest in the perspective of autoimmunity is argininosuccinate synthetase, the downregulation of which could be important for NO production (34).

It should be noted that regulation by diazoxide did not extend to genes expressed specifically in exocrine pancreas. This gives evidence for islet-specific effects of the drug in our islet preparations. In this context it should be mentioned that cholecystokinin (regulated here by diazoxide) is expressed not only in exocrine but also in endocrine pancreas (35).

A comment is warranted on what should be considered a “basal” glucose concentration for rat islets maintained in culture. Long-term culture of rat β-cells at 5.5 mmol/l glucose has been reported to induce β-cell malfunctioning and apoptosis (36,37). However, in the current study and in previous ones, we did not record any negative effects. This is exemplified by the large insulin response to an acute glucose challenge in islets cultured at 5.5 mmol/l glucose (Fig. 1B). A short culture time (24 h) and—in contrast to the aforementioned studies—use of serum-supplemented media could account for preserved functionality after culture at 5.5 mmol/l glucose. However, culture at 11 mmol/l glucose has been found in some studies to provide optimal conditions for rodent β-cell function and survival, and only further elevation of glucose levels was found to induce negative effects (38).

For these reasons we also included microarrays for the 11 mmol/l glucose condition. The results from these arrays show that a number of genes were affected by diazoxide that were not affected at 5.5 mmol/l glucose. However, an increase of glucose concentration from 11 to 27 mmol/l further increased the number of genes affected by diazoxide. In particular, the number of genes that are downregulated by diazoxide was dramatically increased. The latter genes would be primary suspects for being involved in attenuation of glucose-induced insulin secretion by prolonged hyperglycemia. As already mentioned, 20% of these genes are involved in metabolism, some with clear linkage to regulation of glucose-induced insulin secretion. Of further note is the downregulation of IGF-binding proteins. Of these, IGF-binding protein-3 has been implicated in growth and as a signaling molecule in β-cells (39), and cyclin D has been known to stimulate β-cell growth (40). An intriguing possibility, which is testable in future studies, is a trade-off between β-cell growth and insulin secretion during prolonged hyperglycemia.

In summary, we demonstrate that the effects of diazoxide on gene expression are largely linked to a high-glucose environment (important exceptions being genes such as GAD and UCP-2), that effects are exerted on glucose and lipid metabolism, and that these effects potentially uphold normal functioning and sensitivity to glucose in β-cells. Such effects are consistent with already-documented beneficial effects by previous diazoxide exposure on insulin secretion. Our findings support the rationale of using diazoxide and other ATP-sensitive K⁺ openers in trials to improve insulin secretion in diabetes.

This work was supported by the Swedish Research Council (grant 72P-15180-01A), the Swedish Society of Medicine, Funds of Karolinska Institutet, the Norwegian Medical Research Council (grant 136139/310), and the Magnus Bergvall, the Fredrik and Ingrid Thuring, the Ragnhild and Einar Lundström, the Åke Wiberg, and the Loo and Hans Osterman foundations.