Critical Role for Cataplerosis via Citrate in Glucose-Regulated Insulin Release

Purified rat β-cells were obtained from male Wistar rat collagenase-isolated pancreatic islets using flow cytometry and were cultured in serum-free medium, as previously described (5). β-cells were precultured overnight in Ham’s F10 medium (Gibco-BRL, Grand Island, NY), supplemented with 2 mmol/l glutamine, 10 mmol/l glucose (Merck, Darmstadt, Germany), 1% charcoal treated type V BSA (Boehringer Mannheim, Germany), 0.075 g/l penicillin (Sigma, St. Louis, MO), and 0.1 g/l streptomycin (Sigma). After this period, the cells were cultured for 24 h in the same medium containing either 3 or 10 mmol/l glucose. Viability of the cells, assessed by neutral red uptake, was the same for the two cell cultures (87 ± 1% G3 cells vs. 87 ± 1% G10 cells).

Total RNA (10 μg) was isolated using the Trizol RNA isolation method (Gibco-BRL) and used as starting material for the cDNA preparation. The first- and second-strand cDNA synthesis was performed using the SuperScript Choice System (Invitrogen, Carlsbad, CA) according to the manufacturer’s instructions, except we used an oligo-dT primer containing a T7 RNA polymerase promoter site. Labeled cRNA was prepared using the MEGAscript in vitro transcription kit (Ambion, Austin, TX). Biotin-labeled CTP and UTP (Enzo, Farningdale, NY) were used in the reaction together with unlabeled NTPs. After the in vitro transcription reaction, the unincorporated nucleotides were removed using RNeasy columns (Qiagen, Hilden, Germany). Array hybridization and scanning was performed as described in Cordozo et al. (6). cRNA was fragmented at 94°C for 35 min in a fragmentation buffer containing 40 mmol/l Tris-acetate, pH 8.1, 100 mmol/l KOAc, and 30 mmol/l MgOAc. Before hybridization, the fragmented cRNA in a 6×SSPE-T hybridization buffer (1 mol/l NaCl, 10 mmol/l Tris, pH 7.6, and 0.005% Triton) was heated to 95°C for 5 min and subsequently to 40°C for 5 min before loading onto the Affymetrix probe array cartridge (Affymetrix, Santa Clara, CA). The probe array was then incubated for 16 h at 40°C at constant rotation (60 rpm). The washing and staining procedure was performed in the Affymetrix Fluidics Station. The probe array was exposed to 10 washes in 6×SSPE-T (at 25°C) and 4 washes in 0.5×SSPE-T (at 50°C). The biotinylated cRNA was stained with a streptavidin-phycoerythrin conjugate (10 mg/ml, Molecular Probes, Eugene, OR) in 6×SSPE-T for 30 min (at 25°C). The labeled RNA was exposed to 10 washes in 6×SSPE-T (at 25°C). The probe arrays were scanned at 560 nm using a confocal laser-scanning microscope with an argon ion laser as excitation source (Hewlett Packard GeneArray Scanner G2500A). The readings from the quantitative scanning were analyzed by the Affymetrix Gene Expression Analysis Software.

To assess which genes are responsible for the glucose-responsive phenotype of rat β-cells, mRNA profiles in G3 and G10 cells were compared via Affymetrix rat Genome U34A oligonucleotide arrays, as described previously (6). Genes were considered to be glucose responsive if they fulfilled the following criteria: 1) average difference (AvDiff as defined by Affymetrix software on basis of difference between perfect match/mismatch probes) is ≥150 units in the condition of increased gene expression; 2) mean fold change in expression level (three replicate experiments) of ≥2; and 3) ≥1.5-fold change in expression level between G3 and G10 cells in each of the three replicate experiments. Genes were assigned to functional clusters based on searches on PubMed and the Kyoto Encyclopedia of Genes and Genomes (www.KEGG.com). Expressed sequence tag sequences were further identified by BLAST searches onto databases.

For some selected transcripts, microarray data were confirmed or extended by semiquantitative PCR analysis. This was performed for GLUT2 (forward primer GGTGTGATCAATGCACCTC, reverse primer GTATCTGGGGCTTTCTGGAC, 25 cycles), ACL (forward primer CCATGTCAGCCAAGGCAATT, reverse primer CAGTTCGTTGAGACCTCCAA, 30 cycles), SREBP1c (forward primer GGAGCCATGGATTGCACATT, reverse primer AGGAAGGCTTCCAGAGAGGA, 25 cycles), PDX1(forward primer CTCGCTGGGAACGCTGGAACA, reverse primer CTTTGGTGGATTTCATCCACGG, 25 cycles), PPAR-α (forward primer: CCCTCTCTCCAGCTTCCAGCCC, reverse primer CCACAAGCGTCTTCTCAGCCATG, 25 cycles), and PPAR-β (forward primer TGACGGCCAGCGAGGGATG, reverse primer GAGGCTGCTGAAGTTGGGGATG, 25 cycles). Polymerization reactions were performed in a Perkin-Elmer 9600 thermocycler in a 50-μl reaction volume containing 2 μl cDNA, 50 ng RNA equivalent, 20 pmol of appropriate oligonucleotide primers, GeneAmp PCR buffer, and 5 units of AmpliTaq Gold polymerase (Perkin Elmer). The thermal cycle profile used was as follows: a 10-min denaturing step at 94°C, a number of cycles indicated between brackets (cycling temperature at 58°C), and a final extension step (at 72°C). To correct for experimental variations between samples, the gene of interest was amplified with α-tubulin (forward primer CTCGCATCCACTTCCCTC, reverse primer ATGCCCTCACCCACGTAC, 20 cycles) in each sample. Samples were loaded onto 1.5% agarose gels and photographed with the Kodak 120 System. Bands were analyzed with the 1D Image Analysis Software program (Eastman Kodak). The amount of each specific product was then expressed relative to the internal control, giving a ratio specific product/control gene for each sample. A negative control (RT reaction performed in absence of RT enzyme) was performed to exclude genomic DNA contamination of cDNA.

To examine stearoyl CoA desaturase gene expression at the protein level, extracts corresponding to 1.4 × 10⁵ pancreatic β-cells were loaded. For immunoblots we used a rabbit polyclonal anti-SCD1 antibody (1:2,000 dilution; gift from Dr. J. Ozols, Farmington, CT; this antiserum also recognizes SCD2 and SCDX) and HRP-labeled donkey anti-rabbit Ig (1:2,000 dilution; Amersham, Buckinghamshire, England) as secondary antibody. As positive controls, rat liver (50 μg) or SCD1-induced rat liver microsomes (kindly given by Dr. J. Ozols) were loaded in parallel. Protein loading was checked by immunoblotting stripped blots with anti-,α-tubulin (1:2,000; Sigma). Protein abundance was visualized using the ECL kit (Amersham) and Kodak Xomat films; intensities of SCD1 bands were quantified using Kodak software (1D Image Analysis Software) and normalized for α-tubulin.

Insulin release from rat β-cells was measured during either 2-h static incubations (5) or dynamic perifusions (8). Samples were assayed for immunoreactive insulin with guinea pig anti-insulin serum. Static incubations were performed with 3 × 10⁴ cells per tube in 0.5-ml volumes of Earle’s Hepes medium (pH 7.35). The conditions were indicated below the figure [different glucose concentrations, 10 mmol/l glucose + 10 nmol/l GLP-1, presence or absence of the ACL-inhibitors radicicol, also known as monorden, a C14-macrolide derivative originally isolated from Monosporium bonorden (9; kind gift from Dr. Yoshida, Tokyo, Japan) or (−)-hydroxycitrate (Hoffman La Roche, Nutley, NJ; 10)]. The perifusion experiments were carried out in a multiple microchamber module (Endotronics, Coon Rapids, MN). Approximately 2.5 × 10⁵ β-cells were loaded per column and preperifused for 20 min in Ham’s F10 medium, 3 mmol/glucose, supplemented with 0.5% BSA, 2 mmol/l glutamine, and 2 mmol/l CaCl₂ and were equilibrated with 95%O₂/5% CO₂. Perifusion was done by stimulation with 3 mmol/l glucose for 20 min and pulses for 10 min with either 10 mmol/l glucose or 10 mmol/l glucose with 10 nmol/l GLP-1. Flow rate was 0.5 ml/min. The dead space was taken into account when expressing the results as a function of time. Insulin content of the perifused β-cells was measured by sonificating the Biogel P2 containing the cells in 2 mmol/l acetic acid/0.25%BSA.

Glucose oxidation was measured as conversion of [U-¹⁴C]-labeled glucose into ¹⁴CO₂ (11). The escape of glucose carbon from mitochondrial oxidation (cataplerosis) was measured radiometrically (11). After 2-h labeling in 100 μl Earle’s Hepes medium containing [U-¹⁴C]-labeled glucose (2.5 μCi/tube), the cells were washed five times in cold buffer and sonicated in 1 mol/l perchloric acid. Cataplerosis was measured, using a β-scintillation counter, as moles of ¹⁴C-labeled carbon-atoms that were precipitated by perchloric acid and that were solubilized by adding water (0.2 mol/l NaCl), chloroform and methanol (volume ratios 1:2:1) (12) to the precipitate. Whereas the water/methanol phase contains peptides synthesized from glucose-derived amino acids (11), the bottom organic phase contains acyl CoAs and lipids that are synthesized from glucose carbon (12).

To explore the significance and regulation of cataplerosis via citrate in insulin release, we have performed a molecular and physiological study of this pathway in FACS-purified �-cells that were cultured for 24 h at 3 mmol/l glucose (G3 cells) and 10 mmol/l glucose (G10 cells). As can be seen in Fig. 1A, the secretory response to a subsequent acute rise in extracellular glucose was severely suppressed in G3 cells (∼75%) as compared with G10 cells. In perifusion experiments, the secretory defect was detected during both the first and early second phase of glucose-stimulated insulin release and it could not be abolished by adding the gluco-incretin hormone glucagon-like peptide-1 (7�37)amide (GLP-1; Fig. 1B). The difference in insulin secretory responsiveness from G3 and G10 cells was not due to changes in cellular viability and insulin content (23 � 2 vs. 20 � 3 pg/cell, G3 vs. G10 cells). Quantitative differences between G3 and G10 cells were detected at the level of metabolic flux, measured by incubating the cells for 2 h with [U-¹⁴C]-glucose. These differences were present at both the level of ¹⁴CO₂ production (glucose oxidation; Fig. 2A) and cataplerosis, which was detected as accumulation of ¹⁴C-labeled intermediates in the cells (Fig. 2B). These metabolic fluxes are expected to be relevant for, respectively, the K_ATP-dependent pathway and cataplerosis (1� 4, 11, 12). Messenger RNA profiling of these cells using Affymetrix GeneChips led to the observations that 76 mRNAs were upregulated in high glucose and that 103 mRNAs were more abundant in cells cultured at low glucose. A molecular correlation to the metabolic changes in Fig. 2 is shown in Fig. 3A, which summarizes the regulated transcripts that are involved in metabolic signaling pathways. Upregulation of mRNA encoding GLUT2 and mitochondrially encoded subunits of the respiratory chain was observed in G10 cells, while on the contrary, mRNA corresponding to the inactivating kinase of the pyruvate dehydrogenase complex was upregulated fourfold in G3 cells. Amplification of signaling caused by enhanced glucose oxidation in G10 cells may be expected from upregulated mRNA encoding both subunits of the K_ATP channels and the calcium sensor protein calmodulin (Fig. 3A). A molecular correlation to the measured cataplerotic flux (Fig. 2B) was also present at the level of microarray analysis (Fig. 3A), because ACL, a cataplerotic enzyme, was upregulated between three- and fourfold at the mRNA level. The resulting citrate efflux may be used in the cells for several purposes. First, the produced acetyl CoA produced by ACL could serve as precursor for lipid biosynthesis. In agreement with this idea, we identified 12 lipogenic transcripts that were upregulated in G10 cells encoding enzymes for de novo synthesis of acyl CoAs and metabolites on the mevalonate pathway. The specific upregulated mRNAs in acyl-group synthesis encode fatty acid synthase (FAS), stearoyl CoA desaturase-2 (as well as a homologous protein), NADH-cytochrome b5-reductase, and acyl CoA hydrolase. In the mevalonate pathway, higher abundance in G10 cells was detected of mRNA encoding HMG-CoA synthase, HMG-CoA reductase, isopentenyl pyrophosphate (IPPP)/dimethylallylpyrophosphate isomerase, farnesylpyrophosphate synthase, squalene epoxidase, as well as 7-dehydrocholesterol reductase. Second, the enhanced citrate efflux could serve to accelerate the citrate pyruvate cycle (3), which reconverts the produced oxaloacetate into malate and pyruvate that re-enters the mitochondria (anaplerosis). The metabolic flux through this cycle was recently observed to correlate well to glucose-regulated insulin release (13). Both malic enzyme and pyruvate carboxylase are highly expressed in �-cells (11,14). Together with glucose-6-phosphate dehydrogenase, malic enzyme produces cycoplasmic NADPH that is required for reductive biosynthesis. Whereas transcript encoding malic enzyme was not upregulated in G10 cells, the glucose-6-phosphate dehydrogenase mRNA was elevated consistently (2.1-fold higher that in G3 cells).

Glucose stimulation of stearoyl CoA desaturase-gene expression was also noted at the protein level (Fig. 3B). Because, on the one hand, the antiserum does not discriminate between the different stearoyl CoA desaturase isoforms and, on the other hand, SCD1 probes on the rat U34A chip scored below the detection threshold in G3 and G10 cells), we interpret the protein changes in Fig. 3B to reflect changes in SCD2 and/or SCD2-related mRNA. Higher abundance in G3 cells was observed for transcripts encoding mitochondrial long-chain acyl CoA dehydrogenase and CDP-diacylglycerol synthase, two enzymes that are expected to use substrate from the pool of available acyl CoAs and mitochondiral creatinine kinase, which may buffer cellular ATP. Incubation in 3 mmol/l glucose increased the abundance of mRNA encoding NAPD⁺-dependent isocitrate dehydrogenase (2.1 � 0.1-fold increase), a change that could divert mitochondrial citrate from cataplerosis and citrate/pyruvate cycling in G3 cells. Finally, 24-h culture in 10 mmol/l glucose switched lipoprotein receptor gene expression from VLDL receptors (3.0 � 1.1-fold decrease) to LDL receptors (5.5 � 1.7-fold increase), contributing to the cellular possibilities to enlarge the pool of cholesterol in G10 cells.

Transcriptional regulation of genes involved in lipid and glucose metabolism is coordinated by members of the SREBP and PPAR families. Consistent with this knowledge, we observed upregulation of SREBP1c mRNA in G10 cells, using both Affymetrix U34A microarrays (2.7 � 0.6-fold difference with G3 cells) and RT-PCR analysis (Fig. 3C). SREBP2 mRNA (probe not on U34A array) was very low in RT-PCR analysis (data not shown). In liver, transcription of the SREBP1c gene depends on a permissive action of insulin (15); furthermore, in primary cultures of rat hepatocytes, SREBP1c-gene transcription is directly enhanced by glucose (16) and dominant-negative SREBP1c suppresses glucose stimulation of lipogenic gene expression in liver (16). Because the insulin levels that had accumulated at the end of the 24-h tissue culture were 5- to 10-fold higher in the G10 culture than in the G3 culture (data not shown), it is conceivable that the observed stimulation is the combination of direct glucose signaling (14) and autocrine insulin signaling, as was reported to exist in vitro (17). It is known that PPAR-α and PPAR-� are expressed in �-cells (18) and that glucose represses PPAR-α gene transcription (19). Because PPAR-� and PPAR-γ probes were below the detection limit, and because PPAR-α probes were not on the U34A microarray, we assessed PPAR-mRNA abundance via RT-PCR analysis. Consistent upregulation of PPAR-α and PPAR-� mRNA was observed in G3 cells (Fig. 3C) while PPAR-γ mRNA expression level remained below detection limits after 35 PCR cycles (data not shown). The effect of low glucose on PPAR-α expression (2.5-fold higher level in G3 cells than in G10 cells; P < 0.001) may well be responsible for the observed upregulation of mRNA encoding both PDH kinase and mitochondrial long-chain acyl CoA dehydrogenase (Fig. 3A). The observed PPAR-α�dependent metabolic adaptation may explain (at least in part) the switch from glucose oxidation to fatty acid oxidation that is known to occur when cells are chronically exposed to low glucose (19).

The coordinated glucose-regulated change in cataplerosis of citrate, induction of the glucose 6-phosphate dehydrogenase gene, and lipogenic gene transcription may be an accidental correlate to the changes in the secretory phenotype. To test the idea that continuous cataplerotic flux is necessary for normal stimulation of the insulin secretory response, we measured the acute regulation of insulin release from rat �-cells with or without pharmacological inhibition of ACL. We tested this site of intervention, since ACL represents the cataplerotic enzyme that is common for both citrate/pyruvate cycling and acyl CoA/cholesterol production (Fig. 3A). For this purpose, we used two different, chemically unrelated, ACL inhibitors: radicicol, a macrolide antibiotic (9) and (-)-hydroxycitrate, a citrate analog (10); both compounds effectively blocked enzymatic activity in extracts from purified rat �-cells (data not shown). As illustrated in Fig. 4, radicicol lowered insulin release stimulated by 10 mmol/l glucose and GLP-1 by 44 � 13% (P < 0.01). Moreover, (-)-hydroxycitrate (2 mmol/l) reduced the amount insulin release stimulated by 10 mmol/l glucose alone (48 � 6% inhibition; P = 0.002) or 10 mmol/l glucose plus GLP-1 (28 � 6% inhibition; P < 0.02). Therefore, in agreement with earlier observations in the intact perfused pancreas (10), it can be concluded from our data that the acute metabolic flux through the cataplerotic ACL reaction is critical for normal stimulation of insulin release by glucose and GLP-1. Orchestrated upregulation of genes involved in acyl CoA and cholesterol synthesis and the effects of ACL inhibition on insulin release both suggest that a rapid flux through ACL is essential for insulin release. ACL both delivers acetyl CoA for lipid biosysnthesis (e.g., cholesterol, terpenes, or acyl groups) and brings oxaloacetate into the citrate/pyruvate cycle, which generates NAPDH in concert with NAPDH produced by flux through the glucose 6-phosphate dehydrogenase step. Direct measurement of pyruvate cycling in insulin-producing cells via nuclear magnetic resonance (NMR) spectroscopy (13) indicates that the metabolic flux through the citrate/pyruvate cycle is closely linked to insulin release. Pancreatic islet secretory granules have been observed to contain numerous cholesterol-rich complexes (20). The participation of cholesterol in the process of granule fusion was observed in neuroendocrine PC-12 cells, where cholesterol was observed to be responsible for the concentration of the SNARE protein SNAP-25 (21). The glucose-induced upregulation of genes involved in acyl CoA synthesis and fatty acid desaturation may also be linked to granule biogenesis and/or exocytosis. Acyl CoAs have been proposed to be directly involved in the regulation of exocytosis (4,22), possibly because they act as substrates for protein lipidation. Another possibility is that glucose regulation of cholesterol, phospholipid, and sphingolipid synthesis alters the composition of membrane domains that are critical for exocytosis, for instance by influencing membrane fluidity or lipid raft density. Regulation of membrane fluidity via SREBP and/or SCD gene transcription has been proposed to be important in other eukaryotic systems (23). In this context, it is interesting to mention that Δ9-stearoyl desaturase exists in liver as a short-lived cellular protein whose abundance is sensitive to nutritional signals (24).

In conclusion, our present analysis of part of the glucose-induced transcriptome in FACS-purified rat β-cells identifies a coordinated glucose-regulated gene cluster that may be involved in citrate/pyruvate cycling, NADPH formation, lipid signaling, granule biogenesis, and/or maintenance of lipid membrane domains into an exocytosis-competent state. It should be further investigated if these results can be extrapolated to regulatory mechanisms that operate in human β-cells. If so, they can aid in the identification of β-cell abnormalities in diabetic patients and in the identification of new targets for oral antidiabetic drugs.