Thioredoxin-Interacting Protein Is Stimulated by Glucose through a Carbohydrate Response Element and Induces ?-Cell Apoptosis
http://www.100md.com
内分泌学杂志 2005年第5期
Department of Medicine, University of Wisconsin-Madison, Madison, Wisconsin 53792
Address all correspondence to: Anath Shalev, M.D., University of Wisconsin-Madison, H4/526 Clinical Science Center, 600 Highland Avenue, Madison, Wisconsin 53792. E-mail: as7@medicine.wisc.edu.
Abstract
Recently, we identified thioredoxin-interacting protein (TXNIP) as the most dramatically glucose-induced gene in our human islet microarray study. TXNIP is a regulator of the cellular redox state, but its role in pancreatic ?-cells and the mechanism of its regulation by glucose remain unknown. We therefore generated a stable transfected ?-cell line (INS-1) overexpressing human TXNIP and found that TXNIP overexpression induced apoptosis as assessed by Bax, Bcl2, caspase-3, and cleaved caspase-9 as well as Hoechst staining. Interestingly, islets of insulin-resistant/diabetic mice (AZIP-F1, BTBRob/ob) demonstrated elevated TXNIP expression, suggesting that TXNIP may play a role in glucotoxicity and the ?-cell loss observed under these conditions. Furthermore, we found that glucose-induced TXNIP transcription is not dependent on glucose metabolism and is mediated by a distinct carbohydrate response element (ChoRE) in the human TXNIP promoter consisting of a perfect nonpalindromic repeat of two E-boxes. Transfection studies demonstrated that this ChoRE was necessary and sufficient to confer glucose responsiveness. Thus, TXNIP is a novel proapoptotic ?-cell gene elevated in insulin resistance/diabetes and up-regulated by glucose through a unique ChoRE and may link glucotoxicity and ?-cell apoptosis.
Introduction
THIOREDOXIN-INTERACTING PROTEIN (TXNIP) was first isolated from a 1,25-dihydroxyvitamin D3-treated HL-60 human promyelocytic cell line and therefore called vitamin D3-up-regulated gene 1 (VDUP1) (1, 2, 3). TXNIP binds to and inhibits thioredoxin and thereby modulates the cellular redox state (3, 4). The thioredoxin system reduces oxidized proteins by oxidizing the two cysteine residues of thioredoxin, which then are reduced back by the reduced nicotinamide adenine dinucleotide phosphate (NADPH)-dependent thioredoxin reductase. To date, two thioredoxin-binding proteins (TBP-1 and TBP-2) have been identified (3, 5). The protein p40phox was identified as TBP-1 and TXNIP/VDUP1 as TBP-2. In addition, TXNIP exerts antiproliferative effects and has been shown to render NIH3T3 cells and vascular smooth muscle cells more susceptible to oxidative stress and apoptosis (4, 6).
TXNIP was also identified as the gene spontaneously mutated in the Hyplip1 locus of HcB-19 mice, a variant C3H mouse strain, and originally thought to be responsible for familial combined hyperlipidemia (7, 8, 9). Although recent studies did not confirm this hypothesis in humans (10), the phenotype of these HcB-19 mice is characterized by elevated plasma triglycerides, ketone bodies, and free fatty acids. In addition, these mice suffer from selective insulin resistance but instead of progressing to diabetes exhibit hyperinsulinemia and mild hypoglycemia (11).
Interestingly, we identified TXNIP as the most dramatically up-regulated gene in an oligonucleotide microarray study looking at glucose effects on isolated human pancreatic islets (12). Previously, no TXNIP expression had been described in islets and the role of TXNIP in ?-cell biology remained unknown. On the other hand, ?-cell-specific overexpression of thioredoxin has been shown to prevent autoimmune and streptozotocin-induced diabetes by inhibiting ?-cell apoptosis (13). Considering that TXNIP inhibits thioredoxin and that islet TXNIP expression is highly regulated by glucose we hypothesized that this gene may have major implications for ?-cell biology especially in response to altered glucose homeostasis as seen in diabetes and insulin resistance.
Pancreatic ?-cells play a critical role in the pathogenesis of both type 1 and type 2 diabetes. ?-Cell loss in diabetes is mainly a result of apoptosis and is enhanced by hyperglycemia (14, 15, 16, 17). Although a number of transcription factors, such as pancreas duodenum homeobox-1 (PDX-1) and CCAAT/enhancer-binding protein-? (C/EBP-?), have been shown to be implicated in the toxic effects of glucose and ?-cell apoptosis (18, 19, 20, 21), the molecular mechanisms of this glucotoxicity are still not fully understood.
In the present study, we therefore aimed at analyzing the role of TXNIP in diabetes and ?-cell apoptosis and determining the mechanisms involved in the observed glucose-induced expression of TXNIP in islets. To this end, we measured TXNIP expression in pancreatic islets of insulin-resistant and diabetic mice, created a ?-cell line overexpressing human TXNIP, and assessed the effect of TXNIP on ?-cell apoptosis. In addition, we performed a detailed promoter analysis of the human TXNIP gene that led to the identification of a distinct carbohydrate response element (ChoRE) conferring the observed glucose responsiveness.
Materials and Methods
Primary islet and ?-cell culture
Human pancreatic islets were isolated as described previously (22) under the approved islet transplantation protocol of the National Institutes of Health (NIH) and were a generous gift of Dr. D. Harlan (National Institutes of Health, Bethesda, MD). Incubations of human islets were performed as previously described (12, 23). As a nonobese mouse model of insulin resistance, the lipoatrophic AZIP-F1 transgenic mice on C57BL/6 background were used (24). In these mice, a-zip, a dominant negative protein of CCAAT/enhancer-binding protein and Jun, is expressed under the aP2 promoter in adipose tissue, which leads to absence of white fat and severe insulin resistance. Islets were isolated as described previously (25) and were a generous gift of Drs. O. Gavrilova and C. Vinson. Islets of wild-type C57BL/6 and of BTBRob/ob mice, a model of obesity and diabetes, were kindly supplied by Dr. A. Attie. Wild-type mouse islets were incubated at 2.5 or 25 mM glucose for 70 h before RNA was extracted. HIT-T15 cells were grown in RPMI 1640 (Invitrogen, Carlsbad, CA) containing 11.1 mM glucose supplemented with 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin. INS-1 cells were maintained in RPMI 1640 (Invitrogen) supplemented with 10% FBS, 1% penicillin-streptomycin, 1 mM sodium pyruvate, 2 mM L-glutamine, 10 mM HEPES, and 0.05 mM 2-mercaptoethanol. After overnight culture in glucose-free and serum-free RPMI 1640 supplemented with 0.1% BSA and 5 mM glucose, INS-1 cells were incubated for 6 h at 2.5 mM glucose, with additional 22.5 mM glucose, 22.5 mM 2-deoxyglucose (not metabolized but phosphorylated by glucokinase) or 22.5 mM 3-O-methylglucose (not metabolized and not phosphorylated). Similarly, incubations were performed in the absence or presence of 2.5 mM cycloheximide to block protein synthesis (26) before RNA was extracted. Because INS-1 cells are typically grown at 11.1 mM glucose, the effect of 25 mM glucose on TXNIP expression was also assessed in comparison with 10 mM glucose as a more physiological glucose level for these cells.
Stable transfected INS-1 cells were maintained in the same RPMI as nontransfected INS-1 cells but supplemented with Geneticin (Invitrogen) 50 μg/ml.
To facilitate assessment of apoptosis, cells were transferred to serum-free RPMI containing 2.5 mM glucose because otherwise basal apoptosis rates are very low in INS-1 cells. Serum withdrawal has been shown to make cells more susceptible to apoptosis and was suggested to mimic the situation in type 1 and type 2 diabetes (27).
Quantitative real-time RT-PCR
RNA was extracted using TRIzol (Invitrogen) according to the manufacturer’s instructions and converted to cDNA with the First Strand cDNA synthesis kit (Roche, Indianapolis, IN). Quantitative real-time RT-PCR was performed on a Prism 7000 Sequence Detection System (Applied Biosystems, Foster City, CA). Human TXNIP was quantified with primers 9 and 10, mouse TXNIP with primers 11 and 13 and probe 12, and rat TXNIP with primers 13 and 14 and probe 12 (Table 1). Hamster TXNIP was quantified with primers 13 and 14 and specific amplification verified with a dissociation curve. Bcl2 was detected with primers 15 and 16, Bax with primers 17 and 18, and caspase-3 with primers 19 and 20 (Table 1). All samples were analyzed in triplicate and corrected for the18S ribosomal subunit (Applied Biosystems) run as an internal standard.
TABLE 1. Oligonucleotides used in this study
Apoptosis assays
The expression of three different apoptotic markers was assessed. Both Bax and Bcl2 belong to the same family of proteins containing two Bcl2 homology domains and are able to heterodimerize, but whereas Bcl2 inhibits apoptosis, Bax is proapoptotic. The ratio of Bax/Bcl2 therefore represents a good measure of the net effect on apoptosis (28). Because Bax and Bcl2 are early markers in the apoptosis cascade, we used as an additional measure caspase-3, a downstream marker that acts as an effector (17).
Protein levels of cleaved caspase-9 were assessed by Western blotting using the rat-specific cleaved caspase-9 antibody at 1:500 (Cell Sig Tech Asp353, Cell Signaling Technology, Inc., Beverly, MA). Whole-cell extract was prepared by lysing INS-1 cells in 50 mM HEPES buffer containing 1% Igepal CA 630 (Nonidet P40), 2 mM activated sodium orthovanadate, 100 mM sodium fluoride, 10 mM sodium pyrophosphate, 4 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, and 2 mM leupeptin. Samples were centrifuged, and protein concentration in the supernatant was quantified using the DC protein assay kit (Bio-Rad Laboratories, Hercules, CA). Twenty-five micrograms of protein were run on a 10–20% Tris-glycine gel and transferred to a polyvinylidene difluoride membrane. ?-Actin was detected using the Abcam 6276 antibody at 1:10,000 (Abcam, Cambridge, MA). Incubation with the primary antibodies was performed at 4 C overnight and with the secondary antibodies (horseradish peroxidase-conjugated antirabbit IgG for cleaved caspase-9 and antimouse IgG for ?-actin) for 1 h at 1:10,000 (Amersham, Little Chalfont, UK). Bands were visualized by autoradiography using ECL plus Western blotting detection reagents (Amersham).
In addition, apoptosis was assessed morphologically by Hoechst staining as described previously (29). To test susceptibility to hydrogen peroxide, cells were grown on 22-mm square glass cover slips in six-well plates and exposed to 15 μM H2O2 for 6 h and then grown overnight in regular growth media. Cells were washed with PBS and fixed in 10% formalin for 30 min. Apoptotic nuclei were detected by terminal uridine deoxynucleotidyl nick end labeling (TUNEL) staining using the DeadEndColorimetric TUNEL System (Promega, Madison, WI) according to the manufacturer’s instructions. Cells were visualized by counterstaining with PROTOCOL Gill hematoxylin Stain (Fisher Scientific, Pittsburgh, PA).
Plasmid construction
The human TXNIP promoter region including –1518 bp upstream of the ATG start codon (FL) was cloned from human genomic DNA using the Elongase Amplification System (Invitrogen) and primers 1 and 8 (Table 1). Promoter deletion mutants (D1–D5) were amplified using the forward primers 2–7 with the reverse primer 8 (Table 1). Primer 6 was used to create a PCR-generated mutation of the first E-box-like motif of the ChoRE site (mut D4). To construct the TXNIP promoter reporter constructs, the PCR amplicons were purified, cut with MluI and BglII, and ligated into the MluI, BglII precut pGL3 enhancer vector (Promega) upstream of the firefly luciferase reporter gene using T4 DNA ligase (Promega).
Oligonucleotides 21 and 22 (Table 1) containing a tandem repeat of the identified ChoRE and the 5' MluI and 3' BglII restriction sites were treated with T4 polynucleotide kinase (Promega) and annealed by incubating the reaction at room temperature for 1 h, then heated at 95 C for 5 min and allowed to cool down to room temperature on the bench top before transferring the reaction on ice. The generated double-stranded oligonucleotide was inserted into the MluI and BglII precut pGL3 control vector (Promega) upstream of the simian virus 40 (SV40) promoter and the firefly luciferase reporter gene as described above yielding the SV40-ChoRE plasmid.
To generate the human TXNIP expression plasmid, human TXNIP was cloned out of human islet cDNA using Elongase (Invitrogen) with primers 23 and 24 (Table 1), and the PCR product was subcloned into the pcDNA3.1/V5-His TOPO vector (Invitrogen) under the control of the constitutively active viral cytomegalovirus (CMV) promoter (CMV-hTXNIP). All constructs were verified by sequencing.
Transient transfection experiments
HIT-T15 cells were plated in 12-well plates and grown overnight to approximately 60% confluence. Cells were transfected with the various TXNIP promoter reporter constructs (0.5 μg DNA per well) in serum-free RPMI (Invitrogen) containing 2.5 mM glucose using Fugene 6 (Roche) according to the manufacturer’s instructions. To control for transfection efficiency, cells were cotransfected with 5 ng/well pRL-TK (Promega) control plasmid expressing the renilla luciferase reporter gene. Three hours after transfection, media were changed and cells were incubated in media containing 10% FBS, 1% penicillin-streptomycin, and 2.5 mM glucose for the control, additional 22.5 mM glucose for high glucose, or 22.5 mM 2-deoxyglucose or 22.5 mM fructose. The transcriptional activity of the D4 deletion construct was also assessed at 10 and 25 mM glucose. For transfection with the SV40-ChoRE plasmid (0.8 μg DNA per well) Lipofectamine Plus (Invitrogen) was used according to the supplier’s directions. Cells were harvested 24 h after transfection, and firefly as well as renilla luciferase activity was determined using the Dual Luciferase Assay Kit (Promega).
Generation of stable transfected cell lines
INS-1 cells were plated in a six-well plate and grown overnight to approximately 50% confluence. Cells were transfected using Lipofectamine Plus (Invitrogen) and 1.6 μg DNA per well in serum-free DMEM (Invitrogen) according to the manufacturer’s instructions, except that 20 μl of Plus Reagent was dissolved in 100 μl media and 12 μl Lipofectamine was dissolved in 100 μl media for each well. Three wells were transfected with CMV-hTXNIP, and three wells were transfected with CMV-LacZ, the pcDNA control plasmid (Invitrogen). Three hours later, the transfection medium was removed and replaced with RPMI supplemented with 10% FBS and 1% penicillin-streptomycin. Selection for stable transfected cells with 100 μg/ml Geneticin (Invitrogen) was started 2 d after transfection. After 2 wk, all Geneticin-sensitive cells had died and the dose of Geneticin was reduced to the maintenance dose of 50 μg/ml, which was used thereafter. Five weeks after transfection, these polyclonal, stable transfected cells were trypsinized and transferred to T-75 flasks.
Statistical analysis
Data are expressed as means ± SEM. P values were calculated by Student’s t tests or by one-way ANOVA for data sets of more than two groups.
Results
Glucose induces TXNIP expression in human and mouse islets
Isolated human islets contain 53.9 ± 2.5% ?-cells, 34.4 ± 2.5% -cells, and 10.4 ± 0.9% -cells as measured by immunohistochemistry (Powers, A. C., personal communication). Incubation of these islets at high glucose for 24 h increased TXNIP expression more than 5-fold compared with islets incubated at low glucose as measured by quantitative real-time RT-PCR (Fig. 1A) (mean cycle number at low glucose, 20.7 ± 0.16). This result is consistent with our previous finding of TXNIP being the top glucose-induced human islet gene found in our oligonucleotide microarray (12).
FIG. 1. Islet TXNIP expression measured by real-time RT-PCR. A, TXNIP expression in human islets incubated at high (16.7 mM) compared with low (1.7 mM) glucose for 24 h. Islets from three different donors were used; each donor served as its own control. B, TXNIP expression in islets isolated from six wild-type C57BL/6 mice incubated at high (25 mM) or low (2.5 mM) glucose for 70 h. C, TXNIP expression in islets of insulin-resistant AZIP-F1 transgenic mice (TG) compared with wild-type littermates (WT). Islets from six mice were pooled per group and analyzed in triplicate.
Similarly, TXNIP expression was more than 7-fold higher in mouse islets incubated at high glucose compared with low glucose (Fig. 1B) (mean cycle number at low glucose, 25.0 ± 0.18).
TXNIP is overexpressed in islets of insulin-resistant mice
To test whether TXNIP expression in pancreatic islets may be involved in diabetes and insulin resistance we measured mouse TXNIP expression in islets of nonobese insulin-resistant AZIP-F1 transgenic C57BL/6 mice (24). Interestingly, we found a highly significant more than 3-fold increase in mouse TXNIP expression in transgenic mice compared with wild-type littermates (Fig. 1C) (mean cycle number for wild-type mice, 25.3 ± 0.05). [Mean serum glucose concentrations of C57BL/6azip and C57BL/6 wild-type were 250 ± 16 mg/dl and 205 ± 7 mg/dl, respectively (25).] A smaller (1.6 ± 0.05-fold) but highly significant (P < 0.001) effect was observed in the obese, diabetic BTBRob/ob mice compared with lean BTBR wild-type animals. (Mean serum glucose levels in BTBRob/ob and lean BTBR wild-type animals are 494 ± 13 mg/dl and 121 ± 12 mg/dl, respectively.) The smaller effect observed in the diabetic BTBRob/ob compared with the C57BL/6azip mice may seem at first sight inconsistent with the concept of glucose-regulated TXNIP expression. However, because whole islets were used for these studies, this difference may be explained by a higher relative proportion of ?-cells in the C57BL/6azip islets compared with islets of BTBRob/ob mice that are characterized by significant ?-cell loss.
TXNIP is expressed in islet ?-cells, and overexpression induces ?-cell apoptosis
Using the INS-1 rat ?-cell line we detected significant endogenous TXNIP expression by real-time RT-PCR (24 cycles) confirming that TXNIP is expressed in islet ?-cells. To further investigate the role of ?-cell TXNIP expression, we generated a stable transfected INS-1 cell line overexpressing human TXNIP. Overexpression of human TXNIP compared with endogenous rat TXNIP was confirmed by quantitative real-time RT-PCR using species-specific primers (Table 1 and Fig. 2A). Interestingly, TXNIP overexpression led to a highly significant increase in ?-cell apoptosis as measured by Bax/Bcl2 ratio and caspase-3 expression (Fig. 2, B and C) as well as shown by the increased protein levels of cleaved (activated) caspase-9 (Fig. 3) and by Hoechst staining (Fig. 4). In addition, TXNIP overexpression made INS-1 ?-cells more susceptible to free radical-mediated stress conferred by hydrogen peroxide. TUNEL staining revealed 26.0% apoptotic cells in TXNIP-overexpressing ?-cells (CMV-hTXNIP) exposed to H2O2 compared with 13.6% in control cells (CMV-LacZ) (data not shown).
FIG. 2. TXNIP overexpression and ?-cell apoptosis. A, Overexpression of human TXNIP in stable transfected INS-1 ?-cells (CMV-hTXNIP) compared with LacZ control (CMV-LacZ) measured by real-time RT-PCR using species-specific primers: endogenous rat TXNIP (white) set at 1 and human TXNIP (black). B and C, TXNIP effects on Bax/Bcl2 ratio (B) and caspase-3 (C), as markers of apoptosis in CMV-hTXNIP and CMV-LacZ cells; one representative of two independent experiments run in triplicate is shown.
FIG. 3. Western blot of cleaved caspase-9. Control (CMV-LacZ) cells and INS-1 ?-cells overexpressing human TXNIP (CMV-hTXNIP) were serum starved for 48 h, and whole-cell extract was prepared as described in Materials and Methods. Twenty-five micrograms of protein were run per lane, and cleaved caspase-9 was detected using the specific Cell Sig Tech Asp353 antibody that exclusively binds to activated caspase-9. ?-Actin was visualized as a loading control. Bars represent fold change in cleaved caspase-9 protein levels as quantified by the difference in band intensity and corrected for ?-actin.
FIG. 4. Morphological assessment of apoptosis. A and B, Control (CMV-LacZ) cells (A) and INS-1 ?-cells (B) overexpressing human TXNIP (CMV-hTXNIP) were serum starved for 48 h, washed, fixed, and stained with Hoechst 33342 at 5 μg/ml for 30 min; white arrows point at apoptotic cell nuclei. C, Quantification of two independent experiments performed in triplicate (total of >8000 cells analyzed).
Glucose-induced TXNIP expression does not require de novo protein synthesis or glucose metabolism
To investigate whether the observed glucose induction of TXNIP is dependent on glucose metabolism, we also used the two nonmetabolizable sugars 2-deoxyglucose and 3-O-methylglucose and found that both led to an increase in TXNIP expression comparable to that observed with glucose (Fig. 5A). Although 2-deoxyglucose is not metabolized, it still is phosphorylated in analogy to glucose (26), whereas 3-O-methylglucose does not undergo phosphorylation by glucokinase. The retained stimulatory effect of 3-O-methylglucose therefore suggested that phosphorylation was not crucial for the observed glucose effect. To further address the question whether the glucose effect requires expression of other protein factors, we blocked overall protein synthesis with cycloheximide. Interestingly, the effects of glucose and 2-deoxyglucose were not inhibited (Fig. 5B), suggesting that the glucose effect is direct and does not require de novo protein synthesis. Taking into account the high glucose concentrations under which INS-1 cells are typically grown, we also used a more physiological baseline glucose concentration of 10 mM for these cells and observed very similar effects on TXNIP expression (Fig. 5C). In addition, we analyzed the effect of glucose on endogenous TXNIP expression in HIT-T15 ?-cells treated as for the transient transfection experiments. Incubation at high glucose for 24 h led to an 11.9 ± 3.5-fold increase in TXNIP expression compared with low glucose as measured by real-time RT-PCR (P = 0.007).
FIG. 5. Sugar-induced ?-cell TXNIP expression. A, INS-1 cells were incubated at 2.5 mM glucose, 25 mM glucose, 22.5 mM 2-deoxyglucose, or 22.5 mM 3-O-methylglucose for 6 h, and TXNIP expression was determined by real-time RT-PCR. B, INS-1 cells were treated as in A, but protein synthesis was blocked with 2.5 mM cycloheximide. C, INS-1 cells were incubated at 10 or 25 mM glucose for 6 h, and TXNIP expression was determined by real-time RT-PCR. One representative of two independent experiments run in triplicate is shown.
The human TXNIP promoter
To identify cis-acting factors responsible for the increase in TXNIP expression in response to glucose we analyzed the human TXNIP promoter by creating a number of reporter constructs encoding different deletions of the promoter and testing them in transfection studies (Fig. 6A). The full 1518-bp promoter sequence (FL) had strong promoter activity when transfected into HIT-T15 ?-cells as shown by comparison with the viral SV40 promoter (Fig. 6B).
FIG. 6. Deletion analysis of human TXNIP promoter. A, Schematic representation of full –1518 promoter sequence (FL) and deletion constructs (D1–D5). Numbers refer to base pairs upstream of start codon; black boxes, two E-boxes of ChoRE; mutD4, mutated ChoRE. B, Transcriptional activity of human TXNIP promoter in transfected HIT-T15 ?-cells. Luciferase activities are expressed as percentage of FL; SV40, constitutively active promoter; Empty V, promoterless vector. C, Glucose induction of human TXNIP transcription shown as percent increase in luciferase activity at high (25 mM) compared with low (2.5 mM) glucose (gluc). Bars represent means of three independent experiments performed in triplicate. D, Schematic representation of heterologous promoter constructs. LUC, luciferase. E, ChoRE-mediated glucose responsiveness shown as percent increase in luciferase activity at high vs. low glucose in HIT-T15 cells transfected with SV40 or SV40-ChoRE. Bars represent means of two independent experiments performed in triplicate.
The human TXNIP promoter contains a ChoRE consisting of two E-box-like motifs
To evaluate the effect of glucose on TXNIP transcription, parallel experiments were performed and cells incubated at low (2.5 mM) or high (25 mM) glucose after transfection. Although glucose had no effect on the SV40 promoter or the promoterless empty vector, it significantly induced the FL human TXNIP promoter as measured by reporter luciferase activity (P < 0.001) (Fig. 6C). An approximately 2-fold induction was also maintained for the deletion constructs D1–D4 (Fig. 6C). Likewise, a highly significant increase was observed when the effect of 25 mM glucose was compared with a higher 10 mM glucose baseline using the D4 construct (1.5 ± 0.1-fold increase; P < 0.001). Incubation of transfected HIT-T15 cells in the presence of 22.5 mM 2-deoxyglucose or 22.5 mM fructose resulted in a very similar 2-fold increase in TXNIP promoter activity (data not shown). On the other hand, incubation of cells at 1 μM insulin did not increase luciferase activity (mean fold change, 0.9 ± 0.1; P = 0.2). This is also consistent with the unchanged TXNIP expression in primary human islets incubated in the presence or absence of 10 μM insulin (mean fold change, 0.7 ± 0.6; P = 0.5). These findings suggest that the observed glucose induction of human TXNIP expression in ?-cells occurs at the transcriptional level, can be conferred by sugars other than glucose, and is not a result of an autocrine insulin effect.
Interestingly, deletion beyond 400 bp (D5) blunted the glucose effect almost completely (1.1 ± 0.05-fold). This suggested that a repeat of two E-box-like motifs at position –400 of the human TXNIP promoter [described in the mouse promoter as an upstream stimulatory factor (USF)/major late transcription factor repeat (2)] may serve as a ChoRE (Fig. 7), this especially because similar sequences have been shown previously to confer glucose responsiveness (30, 31, 32). In fact, mutation of the first E-box-like motif (mut D4) (Fig. 7) abolished all glucose response (Fig. 6C).
FIG. 7. ChoRE consisting of two E-boxes in TXNIP promoter. A, Consensus E-box sequence and typical conformation of two E-boxes (black) separated by 5 bp. B, Alignment of human, mouse, and rat TXNIP ChoRE. Arrows mark the critical direct CACG repeat separated by 7 bp. Numbers represent base pairs upstream of ATG start codon; D4, start of D4 deletion construct. C, Mutation construct of human ChoRE destroying the first E-box; mutD4, start of mutated D4 deletion construct.
The ChoRE identified in the human TXNIP promoter is sufficient to confer glucose responsiveness to a heterologous promoter
To investigate whether the ChoRE identified in the human TXNIP promoter is not only necessary but also sufficient for glucose-induced transactivation, we cloned it upstream of the heterologous SV40 promoter and performed transfection studies using HIT-T15 cells again in the presence of low and high glucose concentrations. A tandem repeat of the ChoRE sequence was enough to make the otherwise unresponsive SV40 promoter respond to glucose and induce glucose-mediated reporter gene transcription (Fig. 6, D and E).
Discussion
The results of this study demonstrate that TXNIP expression is elevated in islets of insulin-resistant and diabetic mice, that glucose induces TXNIP transcription through a newly identified ChoRE in the TXNIP promoter, and that TXNIP expression in turn enhances ?-cell apoptosis.
Apoptosis is the major form of pancreatic ?-cell loss in type 1 and type 2 diabetes (15, 16, 17). In addition, chronic hyperglycemia of diabetes is toxic to the ?-cell. This glucotoxicity is a gradual process characterized by increasing ?-cell dysfunction culminating in irreversible ?-cell loss by apoptosis (14) and resulting in a vicious cycle of destruction. Although the toxicity of glucose has been studied extensively, the molecular mechanisms connecting glucotoxicity and ?-cell loss have only begun to be unraveled (18, 19, 20, 21, 33, 34). Our findings of increased ?-cell TXNIP expression in response to glucose (Fig. 1) resulting in enhanced ?-cell apoptosis (Figs. 2–4) therefore shed new light on this process. They raise the possibility that TXNIP may represent a potential link between elevated glucose and ?-cell apoptosis. Currently, studies silencing TXNIP expression are on the way to prove this hypothesis. If confirmed, TXNIP or factors affecting its expression may serve as therapeutic targets to prevent ?-cell loss associated with diabetes. In fact, it is tempting to speculate that in the HcB-19 mice, the lack of functional TXNIP may result in decreased ?-cell apoptosis, which could at least in part explain why these mutant mice demonstrate fasting hyperinsulinemia and hypoglycemia and do not progress to diabetes despite their insulin resistance (11). On the other hand, in another study, no significant hyperinsulinemia was observed in TXNIP-mutant mice (35), which may seem inconsistent with the predicted increase in ?-cell mass. However, even in this recent work, a trend to higher insulin levels was observed in the TXNIP-mutant mice, and the difference between the two studies was thought to be a result of the age at which the mice were analyzed (35).
TXNIP is a negative regulator of thioredoxin (3, 4, 5), and interestingly, overexpression of thioredoxin in mouse pancreatic ?-cells prevented autoimmune and streptozotocin-induced diabetes by reducing apoptosis (13). It seems therefore likely that TXNIP exerts its proapoptotic effects on ?-cells at least in part by inhibiting thioredoxin and inducing oxidative stress as previously shown (4). Oxidative stress being a key element in ?-cell glucotoxicity and apoptosis (18, 21, 33) is also consistent with this hypothesis. In fact, TXNIP overexpression rendered INS-1 ?-cells more susceptible to apoptosis in response to oxidative stress induced by hydrogen peroxide (data not shown). In addition, vascular TXNIP expression has recently been shown to be elevated in streptozotocin-induced diabetes in rats and to induce vascular oxidative stress (36). However, TXNIP may also induce apoptosis through thioredoxin-mediated regulation of apoptosis signal regulating kinase-1 (ASK-1) (5). Taken together, the observation of glucose-induced TXNIP expression in non-?-cells (36, 37) and our findings of increased TXNIP expression in islets of different mouse models of diabetes and insulin resistance suggest that hyperglycemia induces TXNIP in different tissues where it exerts detrimental effects contributing to the pathogenesis and secondary complications of diabetes. It is also noteworthy that TXNIP is an early-response gene with rapid transcriptional regulation (38), suggesting that even short-term excursions in blood glucose could induce its expression.
Our results further demonstrated that the observed glucose effect on ?-cell TXNIP expression is direct (no inhibition by cycloheximide), that it can be mimicked by other sugars than glucose (stimulation by fructose), and that, unlike in most cases of glucose-regulated gene transcription, it does not require glucose metabolism (similar effects with 2-deoxyglucose and 3-O-methyglucose) (Fig. 5).
Promoter analysis revealed that glucose-stimulated TXNIP transcription requires only the 400 bp upstream of the TXNIP coding region (Fig. 6C). Right at –400 bp of the human TXNIP promoter we noticed a perfect repeat of two E-box-like motifs that are conserved in mouse (2) and rat. Mutation of the 5' E-box-like motif completely abolished glucose responsiveness (Fig. 6C). E-boxes have been shown to serve as the binding site for the basic helix-loop-helix leucine zipper family of transcription factors and a repeat of two E-boxes with 5-bp spacing has been found to be a typical feature of ChoRE motifs conferring glucose and carbohydrate responsiveness. In our case, the two E-boxes are also 5 bp apart but differ from the consensus E-box sequence of CANNTG in that they contain an A (italic) at the fifth position (CACGAG) and therefore are nonpalindromic (Fig. 7). Although the sequence of this E-box repeat is therefore clearly distinct from previously identified ChoREs, it still fulfills one critical feature containing a direct repeat of a CACG motif separated by 7 bp (30, 32).
ChoREs have been described in a number of rodent liver genes involved in glycolysis and lipid metabolism, i.e. L-type pyruvate kinase (L-PK) (39, 40), Spot 14 (S14) (30), and acetyl-CoA carboxylase (ACC) (32). In some cases, such glucose-regulated gene expression has been shown to be mediated by glucose-induced insulin secretion (32, 40) and to require accessory cis-acting factors (31). Neither seems to be the case for glucose-mediated human ?-cell TXNIP expression. In fact, the identified ChoRE was not only necessary but also sufficient to confer glucose responsiveness to a heterologous promoter (Fig. 6, D and E). Finding a ChoRE in the TXNIP promoter further demonstrates that glucose-induced gene expression through such elements is not restricted to glycolytic and lipogenic enzymes but may also play an important role in controlling cell proliferation or cell death.
Because in our case the identified ChoRE in the TXNIP promoter consists of two E-boxes, potential ChoRE-binding proteins include USF1/2. These proteins belong to the c-myc group of basic helix-loop-helix leucine zipper family of transcription factors that bind to E-boxes and have been implicated in glucose-mediated gene expression (41, 42). In fact, in the mouse TNXIP promoter, the two E-boxes were originally described as a USF/major late transcription factor repeat (2). Interestingly, preliminary studies in our laboratory indicate that USF1/2 indeed bind to the identified ChoRE in the human TXNIP promoter and induce TXNIP transcription (data not shown). In this context, it is particularly intriguing that familial combined hyperlipidemia has most recently been linked with USF-1 rather than with TXNIP as previously thought (10). It is therefore tempting to speculate that USF-1 acts through TXNIP, which would explain some of the phenotypic similarities between TXNIP-mutant HcB-19 mice and human familial combined hyperlipidemia that led originally to the belief of TXNIP being associated with this disease.
In conclusion, we have identified TXNIP as a novel proapoptotic ?-cell gene and demonstrated that its expression is induced by glucose and elevated in diabetes and insulin resistance. TXNIP may therefore represent a link between glucotoxicity and ?-cell apoptosis, two critical mechanisms in the pathogenesis of diabetes. In addition, we have identified a unique ChoRE in the human TXNIP promoter that is necessary and sufficient for glucose-induced TXNIP transcription and thereby provides new insight into the molecular mechanisms of glucose-regulated ?-cell gene expression.
References
Chen KS, DeLuca HF 1994 Isolation and characterization of a novel cDNA from HL-60 cells treated with 1,25-dihydroxyvitamin D-3. Biochim Biophys Acta 1219:26–32
Ludwig DL, Kotanides H, Le T, Chavkin D, Bohlen P, Witte L 2001 Cloning, genetic characterization, and chromosomal mapping of the mouse VDUP1 gene. Gene 269:103–112
Nishiyama A, Matsui M, Iwata S, Hirota K, Masutani H, Nakamura H, Takagi Y, Sono H, Gon Y, Yodoi J 1999 Identification of thioredoxin-binding protein-2/vitamin D(3) up-regulated protein 1 as a negative regulator of thioredoxin function and expression. J Biol Chem 274:21645–21650
Junn E, Han SH, Im JY, Yang Y, Cho EW, Um HD, Kim DK, Lee KW, Han PL, Rhee SG, Choi I 2000 Vitamin D3 up-regulated protein 1 mediates oxidative stress via suppressing the thioredoxin function. J Immunol 164:6287–6295
Nishiyama A, Masutani H, Nakamura H, Nishinaka Y, Yodoi J 2001 Redox regulation by thioredoxin and thioredoxin-binding proteins. IUBMB Life 52:29–33
Schulze PC, De Keulenaer GW, Yoshioka J, Kassik KA, Lee RT 2002 Vitamin D3-up-regulated protein-1 (VDUP-1) regulates redox-dependent vascular smooth muscle cell proliferation through interaction with thioredoxin. Circ Res 91:689–695
Bodnar JS, Chatterjee A, Castellani LW, Ross DA, Ohmen J, Cavalcoli J, Wu C, Dains KM, Catanese J, Chu M, Sheth SS, Charugundla K, Demant P, West DB, de Jong P, Lusis AJ 2002 Positional cloning of the combined hyperlipidemia gene Hyplip1. Nat Genet 30:110–116
Castellani LW, Weinreb A, Bodnar J, Goto AM, Doolittle M, Mehrabian M, Demant P, Lusis AJ 1998 Mapping a gene for combined hyperlipidaemia in a mutant mouse strain. Nat Genet 18:374–377
Pajukanta P, Bodnar JS, Sallinen R, Chu M, Airaksinen T, Xiao Q, Castellani LW, Sheth SS, Wessman M, Palotie A, Sinsheimer JS, Demant P, Lusis AJ, Peltonen L 2001 Fine mapping of Hyplip1 and the human homolog, a potential locus for FCHL. Mamm Genome 12:238–245
Pajukanta P, Lilja HE, Sinsheimer JS, Cantor RM, Lusis AJ, Gentile M, Duan XJ, Soro-Paavonen A, Naukkarinen J, Saarela J, Laakso M, Ehnholm C, Taskinen MR, Peltonen L 2004 Familial combined hyperlipidemia is associated with upstream transcription factor 1 (USF1). Nat Genet 36:371–376
Hui TY, Sheth SS, Diffley JM, Potter DW, Lusis AJ, Attie AD, Davis RA 2004 Mice lacking thioredoxin-interacting protein provide evidence linking cellular redox state to appropriate response to nutritional signals. J Biol Chem 279:24387–24393
Shalev A, Pise-Masison CA, Radonovich M, Hoffmann SC, Hirshberg B, Brady JN, Harlan DM 2002 Oligonucleotide microarray analysis of intact human pancreatic islets: identification of glucose-responsive genes and a highly regulated TGF? signaling pathway. Endocrinology 143:3695–3698
Hotta M, Tashiro F, Ikegami H, Niwa H, Ogihara T, Yodoi J, Miyazaki J 1998 Pancreatic ?-cell-specific expression of thioredoxin, an antioxidative and antiapoptotic protein, prevents autoimmune and streptozotocin-induced diabetes. J Exp Med 188:1445–1451
Bonner-Weir S 2000 Life and death of the pancreatic ?-cells. Trends Endocrinol Metab 11:375–378
Pick A, Clark J, Kubstrup C, Levisetti M, Pugh W, Bonner-Weir S, Polonsky KS 1998 Role of apoptosis in failure of ?-cell mass compensation for insulin resistance and ?-cell defects in the male Zucker diabetic fatty rat. Diabetes 47:358–364
Poitout V, Robertson RP 2002 Secondary ?-cell failure in type 2 diabetes: a convergence of glucotoxicity and lipotoxicity. Endocrinology 143:339–342
Mathis D, Vence L, Benoist C 2001 ?-Cell death during progression to diabetes. Nature 414:792–798
Robertson RP, Harmon J, Tran PO, Tanaka Y, Takahashi H 2003 Glucose toxicity in ?-cells: type 2 diabetes, good radicals gone bad, and the glutathione connection. Diabetes 52:581–587
Seufert J, Weir GC, Habener JF 1998 Differential expression of the insulin gene transcriptional repressor CCAAT/enhancer-binding protein ? and transactivator islet duodenum homeobox-1 in rat pancreatic ?-cells during the development of diabetes mellitus. J Clin Invest 101:2528–2539
Maedler K, Storling J, Sturis J, Zuellig RA, Spinas GA, Arkhammar PO, Mandrup-Poulsen T, Donath MY 2004 Glucose- and interleukin-1?-induced ?-cell apoptosis requires Ca2+ influx and extracellular signal-regulated kinase (ERK) 1/2 activation and is prevented by a sulfonylurea receptor 1/inwardly rectifying K+ channel 6.2 (SUR/Kir6.2) selective potassium channel opener in human islets. Diabetes 53:1706–1713
Kajimoto Y, Kaneto H 2004 Role of oxidative stress in pancreatic ?-cell dysfunction. Ann NY Acad Sci 1011:168–176
Ricordi C, Lacy PE, Scharp DW 1989 Automated islet isolation from human pancreas. Diabetes 38(Suppl 1):140–142
Shalev A, Blair PJ, Hoffmann SC, Hirshberg B, Peculis BA, Harlan DM 2002 A proinsulin gene splice variant with increased translation efficiency is expressed in human pancreatic islets. Endocrinology 143:2541–2547
Reitman ML, Mason MM, Moitra J, Gavrilova O, Marcus-Samuels B, Eckhaus M, Vinson C 1999 Transgenic mice lacking white fat: models for understanding human lipoatrophic diabetes. Ann NY Acad Sci 892:289–296
Minn AH, Patterson NB, Pack S, Hoffmann SC, Gavrilova O, Vinson C, Harlan DM, Shalev A 2003 Resistin is expressed in pancreatic islets. Biochem Biophys Res Commun 310:641–645
Roduit R, Morin J, Masse F, Segall L, Roche E, Newgard CB, Assimacopoulos-Jeannet F, Prentki M 2000 Glucose down-regulates the expression of the peroxisome proliferator-activated receptor-gene in the pancreatic ?-cell. J Biol Chem 275:35799–35806
Maestre I, Jordan J, Calvo S, Reig JA, Cena V, Soria B, Prentki M, Roche E 2003 Mitochondrial dysfunction is involved in apoptosis induced by serum withdrawal and fatty acids in the ?-cell line INS-1. Endocrinology 144:335–345
Oltvai ZN, Korsmeyer SJ 1994 Checkpoints of dueling dimers foil death wishes. Cell 79:189–192
Wrede CE, Dickson LM, Lingohr MK, Briaud I, Rhodes CJ 2002 Protein kinase B/Akt prevents fatty acid-induced apoptosis in pancreatic ?-cells (INS-1). J Biol Chem 277:49676–49684
Koo SH, Towle HC 2000 Glucose regulation of mouse S(14) gene expression in hepatocytes. Involvement of a novel transcription factor complex. J Biol Chem 275:5200–5207
Portois L, Tastenoy M, Viollet B, Svoboda M 2002 Functional analysis of the glucose response element of the rat glucagon receptor gene in insulin-producing INS-1 cells. Biochim Biophys Acta 1574:175–186
O’Callaghan BL, Koo SH, Wu Y, Freake HC, Towle HC 2001 Glucose regulation of the acetyl-CoA carboxylase promoter PI in rat hepatocytes. J Biol Chem 276:16033–16039
Tanaka Y, Gleason CE, Tran PO, Harmon JS, Robertson RP 1999 Prevention of glucose toxicity in HIT-T15 cells and Zucker diabetic fatty rats by antioxidants. Proc Natl Acad Sci USA 96:10857–10862
Harmon JS, Gleason CE, Tanaka Y, Oseid EA, Hunter-Berger KK, Robertson RP 1999 In vivo prevention of hyperglycemia also prevents glucotoxic effects on PDX-1 and insulin gene expression. Diabetes 48:1995–2000
Sheth SS, Castellani LW, Chari S, Wagg C, Thipphavong CK, Bodnar JS, Tontonoz P, Attie AD, Lopaschuk GD, Lusis AJ 2005 Thioredoxin-interacting protein deficiency disrupts the fasting-feeding metabolic transition. J Lipid Res 46:123–134
Schulze PC, Yoshioka J, Takahashi T, He Z, King GL, Lee RT 2004 Hyperglycemia promotes oxidative stress through inhibition of thioredoxin function by thioredoxin-interacting protein. J Biol Chem 279:30369–30374
Hirota T, Okano T, Kokame K, Shirotani-Ikejima H, Miyata T, Fukada Y 2002 Glucose down-regulates Per1 and Per2 mRNA levels and induces circadian gene expression in cultured Rat-1 fibroblasts. J Biol Chem 277:44244–44251
Saitoh T, Tanaka S, Koike T 2001 Rapid induction and Ca(2+) influx-mediated suppression of vitamin D3 up-regulated protein 1 (VDUP1) mRNA in cerebellar granule neurons undergoing apoptosis. J Neurochem 78:1267–1276
Hasegawa J, Osatomi K, Wu RF, Uyeda K 1999 A novel factor binding to the glucose response elements of liver pyruvate kinase and fatty acid synthase genes. J Biol Chem 274:1100–1107
Vaulont S, Vasseur-Cognet M, Kahn A 2000 Glucose regulation of gene transcription. J Biol Chem 275:31555–31558
Lefrancois-Martinez AM, Martinez A, Antoine B, Raymondjean M, Kahn A 1995 Upstream stimulatory factor proteins are major components of the glucose response complex of the L-type pyruvate kinase gene promoter. J Biol Chem 270:2640–2643
Vallet VS, Henrion AA, Bucchini D, Casado M, Raymondjean M, Kahn A, Vaulont S 1997 Glucose-dependent liver gene expression in upstream stimulatory factor 2 –/– mice. J Biol Chem 272:21944–21949(Alexandra H. Minn, Christ)
Address all correspondence to: Anath Shalev, M.D., University of Wisconsin-Madison, H4/526 Clinical Science Center, 600 Highland Avenue, Madison, Wisconsin 53792. E-mail: as7@medicine.wisc.edu.
Abstract
Recently, we identified thioredoxin-interacting protein (TXNIP) as the most dramatically glucose-induced gene in our human islet microarray study. TXNIP is a regulator of the cellular redox state, but its role in pancreatic ?-cells and the mechanism of its regulation by glucose remain unknown. We therefore generated a stable transfected ?-cell line (INS-1) overexpressing human TXNIP and found that TXNIP overexpression induced apoptosis as assessed by Bax, Bcl2, caspase-3, and cleaved caspase-9 as well as Hoechst staining. Interestingly, islets of insulin-resistant/diabetic mice (AZIP-F1, BTBRob/ob) demonstrated elevated TXNIP expression, suggesting that TXNIP may play a role in glucotoxicity and the ?-cell loss observed under these conditions. Furthermore, we found that glucose-induced TXNIP transcription is not dependent on glucose metabolism and is mediated by a distinct carbohydrate response element (ChoRE) in the human TXNIP promoter consisting of a perfect nonpalindromic repeat of two E-boxes. Transfection studies demonstrated that this ChoRE was necessary and sufficient to confer glucose responsiveness. Thus, TXNIP is a novel proapoptotic ?-cell gene elevated in insulin resistance/diabetes and up-regulated by glucose through a unique ChoRE and may link glucotoxicity and ?-cell apoptosis.
Introduction
THIOREDOXIN-INTERACTING PROTEIN (TXNIP) was first isolated from a 1,25-dihydroxyvitamin D3-treated HL-60 human promyelocytic cell line and therefore called vitamin D3-up-regulated gene 1 (VDUP1) (1, 2, 3). TXNIP binds to and inhibits thioredoxin and thereby modulates the cellular redox state (3, 4). The thioredoxin system reduces oxidized proteins by oxidizing the two cysteine residues of thioredoxin, which then are reduced back by the reduced nicotinamide adenine dinucleotide phosphate (NADPH)-dependent thioredoxin reductase. To date, two thioredoxin-binding proteins (TBP-1 and TBP-2) have been identified (3, 5). The protein p40phox was identified as TBP-1 and TXNIP/VDUP1 as TBP-2. In addition, TXNIP exerts antiproliferative effects and has been shown to render NIH3T3 cells and vascular smooth muscle cells more susceptible to oxidative stress and apoptosis (4, 6).
TXNIP was also identified as the gene spontaneously mutated in the Hyplip1 locus of HcB-19 mice, a variant C3H mouse strain, and originally thought to be responsible for familial combined hyperlipidemia (7, 8, 9). Although recent studies did not confirm this hypothesis in humans (10), the phenotype of these HcB-19 mice is characterized by elevated plasma triglycerides, ketone bodies, and free fatty acids. In addition, these mice suffer from selective insulin resistance but instead of progressing to diabetes exhibit hyperinsulinemia and mild hypoglycemia (11).
Interestingly, we identified TXNIP as the most dramatically up-regulated gene in an oligonucleotide microarray study looking at glucose effects on isolated human pancreatic islets (12). Previously, no TXNIP expression had been described in islets and the role of TXNIP in ?-cell biology remained unknown. On the other hand, ?-cell-specific overexpression of thioredoxin has been shown to prevent autoimmune and streptozotocin-induced diabetes by inhibiting ?-cell apoptosis (13). Considering that TXNIP inhibits thioredoxin and that islet TXNIP expression is highly regulated by glucose we hypothesized that this gene may have major implications for ?-cell biology especially in response to altered glucose homeostasis as seen in diabetes and insulin resistance.
Pancreatic ?-cells play a critical role in the pathogenesis of both type 1 and type 2 diabetes. ?-Cell loss in diabetes is mainly a result of apoptosis and is enhanced by hyperglycemia (14, 15, 16, 17). Although a number of transcription factors, such as pancreas duodenum homeobox-1 (PDX-1) and CCAAT/enhancer-binding protein-? (C/EBP-?), have been shown to be implicated in the toxic effects of glucose and ?-cell apoptosis (18, 19, 20, 21), the molecular mechanisms of this glucotoxicity are still not fully understood.
In the present study, we therefore aimed at analyzing the role of TXNIP in diabetes and ?-cell apoptosis and determining the mechanisms involved in the observed glucose-induced expression of TXNIP in islets. To this end, we measured TXNIP expression in pancreatic islets of insulin-resistant and diabetic mice, created a ?-cell line overexpressing human TXNIP, and assessed the effect of TXNIP on ?-cell apoptosis. In addition, we performed a detailed promoter analysis of the human TXNIP gene that led to the identification of a distinct carbohydrate response element (ChoRE) conferring the observed glucose responsiveness.
Materials and Methods
Primary islet and ?-cell culture
Human pancreatic islets were isolated as described previously (22) under the approved islet transplantation protocol of the National Institutes of Health (NIH) and were a generous gift of Dr. D. Harlan (National Institutes of Health, Bethesda, MD). Incubations of human islets were performed as previously described (12, 23). As a nonobese mouse model of insulin resistance, the lipoatrophic AZIP-F1 transgenic mice on C57BL/6 background were used (24). In these mice, a-zip, a dominant negative protein of CCAAT/enhancer-binding protein and Jun, is expressed under the aP2 promoter in adipose tissue, which leads to absence of white fat and severe insulin resistance. Islets were isolated as described previously (25) and were a generous gift of Drs. O. Gavrilova and C. Vinson. Islets of wild-type C57BL/6 and of BTBRob/ob mice, a model of obesity and diabetes, were kindly supplied by Dr. A. Attie. Wild-type mouse islets were incubated at 2.5 or 25 mM glucose for 70 h before RNA was extracted. HIT-T15 cells were grown in RPMI 1640 (Invitrogen, Carlsbad, CA) containing 11.1 mM glucose supplemented with 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin. INS-1 cells were maintained in RPMI 1640 (Invitrogen) supplemented with 10% FBS, 1% penicillin-streptomycin, 1 mM sodium pyruvate, 2 mM L-glutamine, 10 mM HEPES, and 0.05 mM 2-mercaptoethanol. After overnight culture in glucose-free and serum-free RPMI 1640 supplemented with 0.1% BSA and 5 mM glucose, INS-1 cells were incubated for 6 h at 2.5 mM glucose, with additional 22.5 mM glucose, 22.5 mM 2-deoxyglucose (not metabolized but phosphorylated by glucokinase) or 22.5 mM 3-O-methylglucose (not metabolized and not phosphorylated). Similarly, incubations were performed in the absence or presence of 2.5 mM cycloheximide to block protein synthesis (26) before RNA was extracted. Because INS-1 cells are typically grown at 11.1 mM glucose, the effect of 25 mM glucose on TXNIP expression was also assessed in comparison with 10 mM glucose as a more physiological glucose level for these cells.
Stable transfected INS-1 cells were maintained in the same RPMI as nontransfected INS-1 cells but supplemented with Geneticin (Invitrogen) 50 μg/ml.
To facilitate assessment of apoptosis, cells were transferred to serum-free RPMI containing 2.5 mM glucose because otherwise basal apoptosis rates are very low in INS-1 cells. Serum withdrawal has been shown to make cells more susceptible to apoptosis and was suggested to mimic the situation in type 1 and type 2 diabetes (27).
Quantitative real-time RT-PCR
RNA was extracted using TRIzol (Invitrogen) according to the manufacturer’s instructions and converted to cDNA with the First Strand cDNA synthesis kit (Roche, Indianapolis, IN). Quantitative real-time RT-PCR was performed on a Prism 7000 Sequence Detection System (Applied Biosystems, Foster City, CA). Human TXNIP was quantified with primers 9 and 10, mouse TXNIP with primers 11 and 13 and probe 12, and rat TXNIP with primers 13 and 14 and probe 12 (Table 1). Hamster TXNIP was quantified with primers 13 and 14 and specific amplification verified with a dissociation curve. Bcl2 was detected with primers 15 and 16, Bax with primers 17 and 18, and caspase-3 with primers 19 and 20 (Table 1). All samples were analyzed in triplicate and corrected for the18S ribosomal subunit (Applied Biosystems) run as an internal standard.
TABLE 1. Oligonucleotides used in this study
Apoptosis assays
The expression of three different apoptotic markers was assessed. Both Bax and Bcl2 belong to the same family of proteins containing two Bcl2 homology domains and are able to heterodimerize, but whereas Bcl2 inhibits apoptosis, Bax is proapoptotic. The ratio of Bax/Bcl2 therefore represents a good measure of the net effect on apoptosis (28). Because Bax and Bcl2 are early markers in the apoptosis cascade, we used as an additional measure caspase-3, a downstream marker that acts as an effector (17).
Protein levels of cleaved caspase-9 were assessed by Western blotting using the rat-specific cleaved caspase-9 antibody at 1:500 (Cell Sig Tech Asp353, Cell Signaling Technology, Inc., Beverly, MA). Whole-cell extract was prepared by lysing INS-1 cells in 50 mM HEPES buffer containing 1% Igepal CA 630 (Nonidet P40), 2 mM activated sodium orthovanadate, 100 mM sodium fluoride, 10 mM sodium pyrophosphate, 4 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, and 2 mM leupeptin. Samples were centrifuged, and protein concentration in the supernatant was quantified using the DC protein assay kit (Bio-Rad Laboratories, Hercules, CA). Twenty-five micrograms of protein were run on a 10–20% Tris-glycine gel and transferred to a polyvinylidene difluoride membrane. ?-Actin was detected using the Abcam 6276 antibody at 1:10,000 (Abcam, Cambridge, MA). Incubation with the primary antibodies was performed at 4 C overnight and with the secondary antibodies (horseradish peroxidase-conjugated antirabbit IgG for cleaved caspase-9 and antimouse IgG for ?-actin) for 1 h at 1:10,000 (Amersham, Little Chalfont, UK). Bands were visualized by autoradiography using ECL plus Western blotting detection reagents (Amersham).
In addition, apoptosis was assessed morphologically by Hoechst staining as described previously (29). To test susceptibility to hydrogen peroxide, cells were grown on 22-mm square glass cover slips in six-well plates and exposed to 15 μM H2O2 for 6 h and then grown overnight in regular growth media. Cells were washed with PBS and fixed in 10% formalin for 30 min. Apoptotic nuclei were detected by terminal uridine deoxynucleotidyl nick end labeling (TUNEL) staining using the DeadEndColorimetric TUNEL System (Promega, Madison, WI) according to the manufacturer’s instructions. Cells were visualized by counterstaining with PROTOCOL Gill hematoxylin Stain (Fisher Scientific, Pittsburgh, PA).
Plasmid construction
The human TXNIP promoter region including –1518 bp upstream of the ATG start codon (FL) was cloned from human genomic DNA using the Elongase Amplification System (Invitrogen) and primers 1 and 8 (Table 1). Promoter deletion mutants (D1–D5) were amplified using the forward primers 2–7 with the reverse primer 8 (Table 1). Primer 6 was used to create a PCR-generated mutation of the first E-box-like motif of the ChoRE site (mut D4). To construct the TXNIP promoter reporter constructs, the PCR amplicons were purified, cut with MluI and BglII, and ligated into the MluI, BglII precut pGL3 enhancer vector (Promega) upstream of the firefly luciferase reporter gene using T4 DNA ligase (Promega).
Oligonucleotides 21 and 22 (Table 1) containing a tandem repeat of the identified ChoRE and the 5' MluI and 3' BglII restriction sites were treated with T4 polynucleotide kinase (Promega) and annealed by incubating the reaction at room temperature for 1 h, then heated at 95 C for 5 min and allowed to cool down to room temperature on the bench top before transferring the reaction on ice. The generated double-stranded oligonucleotide was inserted into the MluI and BglII precut pGL3 control vector (Promega) upstream of the simian virus 40 (SV40) promoter and the firefly luciferase reporter gene as described above yielding the SV40-ChoRE plasmid.
To generate the human TXNIP expression plasmid, human TXNIP was cloned out of human islet cDNA using Elongase (Invitrogen) with primers 23 and 24 (Table 1), and the PCR product was subcloned into the pcDNA3.1/V5-His TOPO vector (Invitrogen) under the control of the constitutively active viral cytomegalovirus (CMV) promoter (CMV-hTXNIP). All constructs were verified by sequencing.
Transient transfection experiments
HIT-T15 cells were plated in 12-well plates and grown overnight to approximately 60% confluence. Cells were transfected with the various TXNIP promoter reporter constructs (0.5 μg DNA per well) in serum-free RPMI (Invitrogen) containing 2.5 mM glucose using Fugene 6 (Roche) according to the manufacturer’s instructions. To control for transfection efficiency, cells were cotransfected with 5 ng/well pRL-TK (Promega) control plasmid expressing the renilla luciferase reporter gene. Three hours after transfection, media were changed and cells were incubated in media containing 10% FBS, 1% penicillin-streptomycin, and 2.5 mM glucose for the control, additional 22.5 mM glucose for high glucose, or 22.5 mM 2-deoxyglucose or 22.5 mM fructose. The transcriptional activity of the D4 deletion construct was also assessed at 10 and 25 mM glucose. For transfection with the SV40-ChoRE plasmid (0.8 μg DNA per well) Lipofectamine Plus (Invitrogen) was used according to the supplier’s directions. Cells were harvested 24 h after transfection, and firefly as well as renilla luciferase activity was determined using the Dual Luciferase Assay Kit (Promega).
Generation of stable transfected cell lines
INS-1 cells were plated in a six-well plate and grown overnight to approximately 50% confluence. Cells were transfected using Lipofectamine Plus (Invitrogen) and 1.6 μg DNA per well in serum-free DMEM (Invitrogen) according to the manufacturer’s instructions, except that 20 μl of Plus Reagent was dissolved in 100 μl media and 12 μl Lipofectamine was dissolved in 100 μl media for each well. Three wells were transfected with CMV-hTXNIP, and three wells were transfected with CMV-LacZ, the pcDNA control plasmid (Invitrogen). Three hours later, the transfection medium was removed and replaced with RPMI supplemented with 10% FBS and 1% penicillin-streptomycin. Selection for stable transfected cells with 100 μg/ml Geneticin (Invitrogen) was started 2 d after transfection. After 2 wk, all Geneticin-sensitive cells had died and the dose of Geneticin was reduced to the maintenance dose of 50 μg/ml, which was used thereafter. Five weeks after transfection, these polyclonal, stable transfected cells were trypsinized and transferred to T-75 flasks.
Statistical analysis
Data are expressed as means ± SEM. P values were calculated by Student’s t tests or by one-way ANOVA for data sets of more than two groups.
Results
Glucose induces TXNIP expression in human and mouse islets
Isolated human islets contain 53.9 ± 2.5% ?-cells, 34.4 ± 2.5% -cells, and 10.4 ± 0.9% -cells as measured by immunohistochemistry (Powers, A. C., personal communication). Incubation of these islets at high glucose for 24 h increased TXNIP expression more than 5-fold compared with islets incubated at low glucose as measured by quantitative real-time RT-PCR (Fig. 1A) (mean cycle number at low glucose, 20.7 ± 0.16). This result is consistent with our previous finding of TXNIP being the top glucose-induced human islet gene found in our oligonucleotide microarray (12).
FIG. 1. Islet TXNIP expression measured by real-time RT-PCR. A, TXNIP expression in human islets incubated at high (16.7 mM) compared with low (1.7 mM) glucose for 24 h. Islets from three different donors were used; each donor served as its own control. B, TXNIP expression in islets isolated from six wild-type C57BL/6 mice incubated at high (25 mM) or low (2.5 mM) glucose for 70 h. C, TXNIP expression in islets of insulin-resistant AZIP-F1 transgenic mice (TG) compared with wild-type littermates (WT). Islets from six mice were pooled per group and analyzed in triplicate.
Similarly, TXNIP expression was more than 7-fold higher in mouse islets incubated at high glucose compared with low glucose (Fig. 1B) (mean cycle number at low glucose, 25.0 ± 0.18).
TXNIP is overexpressed in islets of insulin-resistant mice
To test whether TXNIP expression in pancreatic islets may be involved in diabetes and insulin resistance we measured mouse TXNIP expression in islets of nonobese insulin-resistant AZIP-F1 transgenic C57BL/6 mice (24). Interestingly, we found a highly significant more than 3-fold increase in mouse TXNIP expression in transgenic mice compared with wild-type littermates (Fig. 1C) (mean cycle number for wild-type mice, 25.3 ± 0.05). [Mean serum glucose concentrations of C57BL/6azip and C57BL/6 wild-type were 250 ± 16 mg/dl and 205 ± 7 mg/dl, respectively (25).] A smaller (1.6 ± 0.05-fold) but highly significant (P < 0.001) effect was observed in the obese, diabetic BTBRob/ob mice compared with lean BTBR wild-type animals. (Mean serum glucose levels in BTBRob/ob and lean BTBR wild-type animals are 494 ± 13 mg/dl and 121 ± 12 mg/dl, respectively.) The smaller effect observed in the diabetic BTBRob/ob compared with the C57BL/6azip mice may seem at first sight inconsistent with the concept of glucose-regulated TXNIP expression. However, because whole islets were used for these studies, this difference may be explained by a higher relative proportion of ?-cells in the C57BL/6azip islets compared with islets of BTBRob/ob mice that are characterized by significant ?-cell loss.
TXNIP is expressed in islet ?-cells, and overexpression induces ?-cell apoptosis
Using the INS-1 rat ?-cell line we detected significant endogenous TXNIP expression by real-time RT-PCR (24 cycles) confirming that TXNIP is expressed in islet ?-cells. To further investigate the role of ?-cell TXNIP expression, we generated a stable transfected INS-1 cell line overexpressing human TXNIP. Overexpression of human TXNIP compared with endogenous rat TXNIP was confirmed by quantitative real-time RT-PCR using species-specific primers (Table 1 and Fig. 2A). Interestingly, TXNIP overexpression led to a highly significant increase in ?-cell apoptosis as measured by Bax/Bcl2 ratio and caspase-3 expression (Fig. 2, B and C) as well as shown by the increased protein levels of cleaved (activated) caspase-9 (Fig. 3) and by Hoechst staining (Fig. 4). In addition, TXNIP overexpression made INS-1 ?-cells more susceptible to free radical-mediated stress conferred by hydrogen peroxide. TUNEL staining revealed 26.0% apoptotic cells in TXNIP-overexpressing ?-cells (CMV-hTXNIP) exposed to H2O2 compared with 13.6% in control cells (CMV-LacZ) (data not shown).
FIG. 2. TXNIP overexpression and ?-cell apoptosis. A, Overexpression of human TXNIP in stable transfected INS-1 ?-cells (CMV-hTXNIP) compared with LacZ control (CMV-LacZ) measured by real-time RT-PCR using species-specific primers: endogenous rat TXNIP (white) set at 1 and human TXNIP (black). B and C, TXNIP effects on Bax/Bcl2 ratio (B) and caspase-3 (C), as markers of apoptosis in CMV-hTXNIP and CMV-LacZ cells; one representative of two independent experiments run in triplicate is shown.
FIG. 3. Western blot of cleaved caspase-9. Control (CMV-LacZ) cells and INS-1 ?-cells overexpressing human TXNIP (CMV-hTXNIP) were serum starved for 48 h, and whole-cell extract was prepared as described in Materials and Methods. Twenty-five micrograms of protein were run per lane, and cleaved caspase-9 was detected using the specific Cell Sig Tech Asp353 antibody that exclusively binds to activated caspase-9. ?-Actin was visualized as a loading control. Bars represent fold change in cleaved caspase-9 protein levels as quantified by the difference in band intensity and corrected for ?-actin.
FIG. 4. Morphological assessment of apoptosis. A and B, Control (CMV-LacZ) cells (A) and INS-1 ?-cells (B) overexpressing human TXNIP (CMV-hTXNIP) were serum starved for 48 h, washed, fixed, and stained with Hoechst 33342 at 5 μg/ml for 30 min; white arrows point at apoptotic cell nuclei. C, Quantification of two independent experiments performed in triplicate (total of >8000 cells analyzed).
Glucose-induced TXNIP expression does not require de novo protein synthesis or glucose metabolism
To investigate whether the observed glucose induction of TXNIP is dependent on glucose metabolism, we also used the two nonmetabolizable sugars 2-deoxyglucose and 3-O-methylglucose and found that both led to an increase in TXNIP expression comparable to that observed with glucose (Fig. 5A). Although 2-deoxyglucose is not metabolized, it still is phosphorylated in analogy to glucose (26), whereas 3-O-methylglucose does not undergo phosphorylation by glucokinase. The retained stimulatory effect of 3-O-methylglucose therefore suggested that phosphorylation was not crucial for the observed glucose effect. To further address the question whether the glucose effect requires expression of other protein factors, we blocked overall protein synthesis with cycloheximide. Interestingly, the effects of glucose and 2-deoxyglucose were not inhibited (Fig. 5B), suggesting that the glucose effect is direct and does not require de novo protein synthesis. Taking into account the high glucose concentrations under which INS-1 cells are typically grown, we also used a more physiological baseline glucose concentration of 10 mM for these cells and observed very similar effects on TXNIP expression (Fig. 5C). In addition, we analyzed the effect of glucose on endogenous TXNIP expression in HIT-T15 ?-cells treated as for the transient transfection experiments. Incubation at high glucose for 24 h led to an 11.9 ± 3.5-fold increase in TXNIP expression compared with low glucose as measured by real-time RT-PCR (P = 0.007).
FIG. 5. Sugar-induced ?-cell TXNIP expression. A, INS-1 cells were incubated at 2.5 mM glucose, 25 mM glucose, 22.5 mM 2-deoxyglucose, or 22.5 mM 3-O-methylglucose for 6 h, and TXNIP expression was determined by real-time RT-PCR. B, INS-1 cells were treated as in A, but protein synthesis was blocked with 2.5 mM cycloheximide. C, INS-1 cells were incubated at 10 or 25 mM glucose for 6 h, and TXNIP expression was determined by real-time RT-PCR. One representative of two independent experiments run in triplicate is shown.
The human TXNIP promoter
To identify cis-acting factors responsible for the increase in TXNIP expression in response to glucose we analyzed the human TXNIP promoter by creating a number of reporter constructs encoding different deletions of the promoter and testing them in transfection studies (Fig. 6A). The full 1518-bp promoter sequence (FL) had strong promoter activity when transfected into HIT-T15 ?-cells as shown by comparison with the viral SV40 promoter (Fig. 6B).
FIG. 6. Deletion analysis of human TXNIP promoter. A, Schematic representation of full –1518 promoter sequence (FL) and deletion constructs (D1–D5). Numbers refer to base pairs upstream of start codon; black boxes, two E-boxes of ChoRE; mutD4, mutated ChoRE. B, Transcriptional activity of human TXNIP promoter in transfected HIT-T15 ?-cells. Luciferase activities are expressed as percentage of FL; SV40, constitutively active promoter; Empty V, promoterless vector. C, Glucose induction of human TXNIP transcription shown as percent increase in luciferase activity at high (25 mM) compared with low (2.5 mM) glucose (gluc). Bars represent means of three independent experiments performed in triplicate. D, Schematic representation of heterologous promoter constructs. LUC, luciferase. E, ChoRE-mediated glucose responsiveness shown as percent increase in luciferase activity at high vs. low glucose in HIT-T15 cells transfected with SV40 or SV40-ChoRE. Bars represent means of two independent experiments performed in triplicate.
The human TXNIP promoter contains a ChoRE consisting of two E-box-like motifs
To evaluate the effect of glucose on TXNIP transcription, parallel experiments were performed and cells incubated at low (2.5 mM) or high (25 mM) glucose after transfection. Although glucose had no effect on the SV40 promoter or the promoterless empty vector, it significantly induced the FL human TXNIP promoter as measured by reporter luciferase activity (P < 0.001) (Fig. 6C). An approximately 2-fold induction was also maintained for the deletion constructs D1–D4 (Fig. 6C). Likewise, a highly significant increase was observed when the effect of 25 mM glucose was compared with a higher 10 mM glucose baseline using the D4 construct (1.5 ± 0.1-fold increase; P < 0.001). Incubation of transfected HIT-T15 cells in the presence of 22.5 mM 2-deoxyglucose or 22.5 mM fructose resulted in a very similar 2-fold increase in TXNIP promoter activity (data not shown). On the other hand, incubation of cells at 1 μM insulin did not increase luciferase activity (mean fold change, 0.9 ± 0.1; P = 0.2). This is also consistent with the unchanged TXNIP expression in primary human islets incubated in the presence or absence of 10 μM insulin (mean fold change, 0.7 ± 0.6; P = 0.5). These findings suggest that the observed glucose induction of human TXNIP expression in ?-cells occurs at the transcriptional level, can be conferred by sugars other than glucose, and is not a result of an autocrine insulin effect.
Interestingly, deletion beyond 400 bp (D5) blunted the glucose effect almost completely (1.1 ± 0.05-fold). This suggested that a repeat of two E-box-like motifs at position –400 of the human TXNIP promoter [described in the mouse promoter as an upstream stimulatory factor (USF)/major late transcription factor repeat (2)] may serve as a ChoRE (Fig. 7), this especially because similar sequences have been shown previously to confer glucose responsiveness (30, 31, 32). In fact, mutation of the first E-box-like motif (mut D4) (Fig. 7) abolished all glucose response (Fig. 6C).
FIG. 7. ChoRE consisting of two E-boxes in TXNIP promoter. A, Consensus E-box sequence and typical conformation of two E-boxes (black) separated by 5 bp. B, Alignment of human, mouse, and rat TXNIP ChoRE. Arrows mark the critical direct CACG repeat separated by 7 bp. Numbers represent base pairs upstream of ATG start codon; D4, start of D4 deletion construct. C, Mutation construct of human ChoRE destroying the first E-box; mutD4, start of mutated D4 deletion construct.
The ChoRE identified in the human TXNIP promoter is sufficient to confer glucose responsiveness to a heterologous promoter
To investigate whether the ChoRE identified in the human TXNIP promoter is not only necessary but also sufficient for glucose-induced transactivation, we cloned it upstream of the heterologous SV40 promoter and performed transfection studies using HIT-T15 cells again in the presence of low and high glucose concentrations. A tandem repeat of the ChoRE sequence was enough to make the otherwise unresponsive SV40 promoter respond to glucose and induce glucose-mediated reporter gene transcription (Fig. 6, D and E).
Discussion
The results of this study demonstrate that TXNIP expression is elevated in islets of insulin-resistant and diabetic mice, that glucose induces TXNIP transcription through a newly identified ChoRE in the TXNIP promoter, and that TXNIP expression in turn enhances ?-cell apoptosis.
Apoptosis is the major form of pancreatic ?-cell loss in type 1 and type 2 diabetes (15, 16, 17). In addition, chronic hyperglycemia of diabetes is toxic to the ?-cell. This glucotoxicity is a gradual process characterized by increasing ?-cell dysfunction culminating in irreversible ?-cell loss by apoptosis (14) and resulting in a vicious cycle of destruction. Although the toxicity of glucose has been studied extensively, the molecular mechanisms connecting glucotoxicity and ?-cell loss have only begun to be unraveled (18, 19, 20, 21, 33, 34). Our findings of increased ?-cell TXNIP expression in response to glucose (Fig. 1) resulting in enhanced ?-cell apoptosis (Figs. 2–4) therefore shed new light on this process. They raise the possibility that TXNIP may represent a potential link between elevated glucose and ?-cell apoptosis. Currently, studies silencing TXNIP expression are on the way to prove this hypothesis. If confirmed, TXNIP or factors affecting its expression may serve as therapeutic targets to prevent ?-cell loss associated with diabetes. In fact, it is tempting to speculate that in the HcB-19 mice, the lack of functional TXNIP may result in decreased ?-cell apoptosis, which could at least in part explain why these mutant mice demonstrate fasting hyperinsulinemia and hypoglycemia and do not progress to diabetes despite their insulin resistance (11). On the other hand, in another study, no significant hyperinsulinemia was observed in TXNIP-mutant mice (35), which may seem inconsistent with the predicted increase in ?-cell mass. However, even in this recent work, a trend to higher insulin levels was observed in the TXNIP-mutant mice, and the difference between the two studies was thought to be a result of the age at which the mice were analyzed (35).
TXNIP is a negative regulator of thioredoxin (3, 4, 5), and interestingly, overexpression of thioredoxin in mouse pancreatic ?-cells prevented autoimmune and streptozotocin-induced diabetes by reducing apoptosis (13). It seems therefore likely that TXNIP exerts its proapoptotic effects on ?-cells at least in part by inhibiting thioredoxin and inducing oxidative stress as previously shown (4). Oxidative stress being a key element in ?-cell glucotoxicity and apoptosis (18, 21, 33) is also consistent with this hypothesis. In fact, TXNIP overexpression rendered INS-1 ?-cells more susceptible to apoptosis in response to oxidative stress induced by hydrogen peroxide (data not shown). In addition, vascular TXNIP expression has recently been shown to be elevated in streptozotocin-induced diabetes in rats and to induce vascular oxidative stress (36). However, TXNIP may also induce apoptosis through thioredoxin-mediated regulation of apoptosis signal regulating kinase-1 (ASK-1) (5). Taken together, the observation of glucose-induced TXNIP expression in non-?-cells (36, 37) and our findings of increased TXNIP expression in islets of different mouse models of diabetes and insulin resistance suggest that hyperglycemia induces TXNIP in different tissues where it exerts detrimental effects contributing to the pathogenesis and secondary complications of diabetes. It is also noteworthy that TXNIP is an early-response gene with rapid transcriptional regulation (38), suggesting that even short-term excursions in blood glucose could induce its expression.
Our results further demonstrated that the observed glucose effect on ?-cell TXNIP expression is direct (no inhibition by cycloheximide), that it can be mimicked by other sugars than glucose (stimulation by fructose), and that, unlike in most cases of glucose-regulated gene transcription, it does not require glucose metabolism (similar effects with 2-deoxyglucose and 3-O-methyglucose) (Fig. 5).
Promoter analysis revealed that glucose-stimulated TXNIP transcription requires only the 400 bp upstream of the TXNIP coding region (Fig. 6C). Right at –400 bp of the human TXNIP promoter we noticed a perfect repeat of two E-box-like motifs that are conserved in mouse (2) and rat. Mutation of the 5' E-box-like motif completely abolished glucose responsiveness (Fig. 6C). E-boxes have been shown to serve as the binding site for the basic helix-loop-helix leucine zipper family of transcription factors and a repeat of two E-boxes with 5-bp spacing has been found to be a typical feature of ChoRE motifs conferring glucose and carbohydrate responsiveness. In our case, the two E-boxes are also 5 bp apart but differ from the consensus E-box sequence of CANNTG in that they contain an A (italic) at the fifth position (CACGAG) and therefore are nonpalindromic (Fig. 7). Although the sequence of this E-box repeat is therefore clearly distinct from previously identified ChoREs, it still fulfills one critical feature containing a direct repeat of a CACG motif separated by 7 bp (30, 32).
ChoREs have been described in a number of rodent liver genes involved in glycolysis and lipid metabolism, i.e. L-type pyruvate kinase (L-PK) (39, 40), Spot 14 (S14) (30), and acetyl-CoA carboxylase (ACC) (32). In some cases, such glucose-regulated gene expression has been shown to be mediated by glucose-induced insulin secretion (32, 40) and to require accessory cis-acting factors (31). Neither seems to be the case for glucose-mediated human ?-cell TXNIP expression. In fact, the identified ChoRE was not only necessary but also sufficient to confer glucose responsiveness to a heterologous promoter (Fig. 6, D and E). Finding a ChoRE in the TXNIP promoter further demonstrates that glucose-induced gene expression through such elements is not restricted to glycolytic and lipogenic enzymes but may also play an important role in controlling cell proliferation or cell death.
Because in our case the identified ChoRE in the TXNIP promoter consists of two E-boxes, potential ChoRE-binding proteins include USF1/2. These proteins belong to the c-myc group of basic helix-loop-helix leucine zipper family of transcription factors that bind to E-boxes and have been implicated in glucose-mediated gene expression (41, 42). In fact, in the mouse TNXIP promoter, the two E-boxes were originally described as a USF/major late transcription factor repeat (2). Interestingly, preliminary studies in our laboratory indicate that USF1/2 indeed bind to the identified ChoRE in the human TXNIP promoter and induce TXNIP transcription (data not shown). In this context, it is particularly intriguing that familial combined hyperlipidemia has most recently been linked with USF-1 rather than with TXNIP as previously thought (10). It is therefore tempting to speculate that USF-1 acts through TXNIP, which would explain some of the phenotypic similarities between TXNIP-mutant HcB-19 mice and human familial combined hyperlipidemia that led originally to the belief of TXNIP being associated with this disease.
In conclusion, we have identified TXNIP as a novel proapoptotic ?-cell gene and demonstrated that its expression is induced by glucose and elevated in diabetes and insulin resistance. TXNIP may therefore represent a link between glucotoxicity and ?-cell apoptosis, two critical mechanisms in the pathogenesis of diabetes. In addition, we have identified a unique ChoRE in the human TXNIP promoter that is necessary and sufficient for glucose-induced TXNIP transcription and thereby provides new insight into the molecular mechanisms of glucose-regulated ?-cell gene expression.
References
Chen KS, DeLuca HF 1994 Isolation and characterization of a novel cDNA from HL-60 cells treated with 1,25-dihydroxyvitamin D-3. Biochim Biophys Acta 1219:26–32
Ludwig DL, Kotanides H, Le T, Chavkin D, Bohlen P, Witte L 2001 Cloning, genetic characterization, and chromosomal mapping of the mouse VDUP1 gene. Gene 269:103–112
Nishiyama A, Matsui M, Iwata S, Hirota K, Masutani H, Nakamura H, Takagi Y, Sono H, Gon Y, Yodoi J 1999 Identification of thioredoxin-binding protein-2/vitamin D(3) up-regulated protein 1 as a negative regulator of thioredoxin function and expression. J Biol Chem 274:21645–21650
Junn E, Han SH, Im JY, Yang Y, Cho EW, Um HD, Kim DK, Lee KW, Han PL, Rhee SG, Choi I 2000 Vitamin D3 up-regulated protein 1 mediates oxidative stress via suppressing the thioredoxin function. J Immunol 164:6287–6295
Nishiyama A, Masutani H, Nakamura H, Nishinaka Y, Yodoi J 2001 Redox regulation by thioredoxin and thioredoxin-binding proteins. IUBMB Life 52:29–33
Schulze PC, De Keulenaer GW, Yoshioka J, Kassik KA, Lee RT 2002 Vitamin D3-up-regulated protein-1 (VDUP-1) regulates redox-dependent vascular smooth muscle cell proliferation through interaction with thioredoxin. Circ Res 91:689–695
Bodnar JS, Chatterjee A, Castellani LW, Ross DA, Ohmen J, Cavalcoli J, Wu C, Dains KM, Catanese J, Chu M, Sheth SS, Charugundla K, Demant P, West DB, de Jong P, Lusis AJ 2002 Positional cloning of the combined hyperlipidemia gene Hyplip1. Nat Genet 30:110–116
Castellani LW, Weinreb A, Bodnar J, Goto AM, Doolittle M, Mehrabian M, Demant P, Lusis AJ 1998 Mapping a gene for combined hyperlipidaemia in a mutant mouse strain. Nat Genet 18:374–377
Pajukanta P, Bodnar JS, Sallinen R, Chu M, Airaksinen T, Xiao Q, Castellani LW, Sheth SS, Wessman M, Palotie A, Sinsheimer JS, Demant P, Lusis AJ, Peltonen L 2001 Fine mapping of Hyplip1 and the human homolog, a potential locus for FCHL. Mamm Genome 12:238–245
Pajukanta P, Lilja HE, Sinsheimer JS, Cantor RM, Lusis AJ, Gentile M, Duan XJ, Soro-Paavonen A, Naukkarinen J, Saarela J, Laakso M, Ehnholm C, Taskinen MR, Peltonen L 2004 Familial combined hyperlipidemia is associated with upstream transcription factor 1 (USF1). Nat Genet 36:371–376
Hui TY, Sheth SS, Diffley JM, Potter DW, Lusis AJ, Attie AD, Davis RA 2004 Mice lacking thioredoxin-interacting protein provide evidence linking cellular redox state to appropriate response to nutritional signals. J Biol Chem 279:24387–24393
Shalev A, Pise-Masison CA, Radonovich M, Hoffmann SC, Hirshberg B, Brady JN, Harlan DM 2002 Oligonucleotide microarray analysis of intact human pancreatic islets: identification of glucose-responsive genes and a highly regulated TGF? signaling pathway. Endocrinology 143:3695–3698
Hotta M, Tashiro F, Ikegami H, Niwa H, Ogihara T, Yodoi J, Miyazaki J 1998 Pancreatic ?-cell-specific expression of thioredoxin, an antioxidative and antiapoptotic protein, prevents autoimmune and streptozotocin-induced diabetes. J Exp Med 188:1445–1451
Bonner-Weir S 2000 Life and death of the pancreatic ?-cells. Trends Endocrinol Metab 11:375–378
Pick A, Clark J, Kubstrup C, Levisetti M, Pugh W, Bonner-Weir S, Polonsky KS 1998 Role of apoptosis in failure of ?-cell mass compensation for insulin resistance and ?-cell defects in the male Zucker diabetic fatty rat. Diabetes 47:358–364
Poitout V, Robertson RP 2002 Secondary ?-cell failure in type 2 diabetes: a convergence of glucotoxicity and lipotoxicity. Endocrinology 143:339–342
Mathis D, Vence L, Benoist C 2001 ?-Cell death during progression to diabetes. Nature 414:792–798
Robertson RP, Harmon J, Tran PO, Tanaka Y, Takahashi H 2003 Glucose toxicity in ?-cells: type 2 diabetes, good radicals gone bad, and the glutathione connection. Diabetes 52:581–587
Seufert J, Weir GC, Habener JF 1998 Differential expression of the insulin gene transcriptional repressor CCAAT/enhancer-binding protein ? and transactivator islet duodenum homeobox-1 in rat pancreatic ?-cells during the development of diabetes mellitus. J Clin Invest 101:2528–2539
Maedler K, Storling J, Sturis J, Zuellig RA, Spinas GA, Arkhammar PO, Mandrup-Poulsen T, Donath MY 2004 Glucose- and interleukin-1?-induced ?-cell apoptosis requires Ca2+ influx and extracellular signal-regulated kinase (ERK) 1/2 activation and is prevented by a sulfonylurea receptor 1/inwardly rectifying K+ channel 6.2 (SUR/Kir6.2) selective potassium channel opener in human islets. Diabetes 53:1706–1713
Kajimoto Y, Kaneto H 2004 Role of oxidative stress in pancreatic ?-cell dysfunction. Ann NY Acad Sci 1011:168–176
Ricordi C, Lacy PE, Scharp DW 1989 Automated islet isolation from human pancreas. Diabetes 38(Suppl 1):140–142
Shalev A, Blair PJ, Hoffmann SC, Hirshberg B, Peculis BA, Harlan DM 2002 A proinsulin gene splice variant with increased translation efficiency is expressed in human pancreatic islets. Endocrinology 143:2541–2547
Reitman ML, Mason MM, Moitra J, Gavrilova O, Marcus-Samuels B, Eckhaus M, Vinson C 1999 Transgenic mice lacking white fat: models for understanding human lipoatrophic diabetes. Ann NY Acad Sci 892:289–296
Minn AH, Patterson NB, Pack S, Hoffmann SC, Gavrilova O, Vinson C, Harlan DM, Shalev A 2003 Resistin is expressed in pancreatic islets. Biochem Biophys Res Commun 310:641–645
Roduit R, Morin J, Masse F, Segall L, Roche E, Newgard CB, Assimacopoulos-Jeannet F, Prentki M 2000 Glucose down-regulates the expression of the peroxisome proliferator-activated receptor-gene in the pancreatic ?-cell. J Biol Chem 275:35799–35806
Maestre I, Jordan J, Calvo S, Reig JA, Cena V, Soria B, Prentki M, Roche E 2003 Mitochondrial dysfunction is involved in apoptosis induced by serum withdrawal and fatty acids in the ?-cell line INS-1. Endocrinology 144:335–345
Oltvai ZN, Korsmeyer SJ 1994 Checkpoints of dueling dimers foil death wishes. Cell 79:189–192
Wrede CE, Dickson LM, Lingohr MK, Briaud I, Rhodes CJ 2002 Protein kinase B/Akt prevents fatty acid-induced apoptosis in pancreatic ?-cells (INS-1). J Biol Chem 277:49676–49684
Koo SH, Towle HC 2000 Glucose regulation of mouse S(14) gene expression in hepatocytes. Involvement of a novel transcription factor complex. J Biol Chem 275:5200–5207
Portois L, Tastenoy M, Viollet B, Svoboda M 2002 Functional analysis of the glucose response element of the rat glucagon receptor gene in insulin-producing INS-1 cells. Biochim Biophys Acta 1574:175–186
O’Callaghan BL, Koo SH, Wu Y, Freake HC, Towle HC 2001 Glucose regulation of the acetyl-CoA carboxylase promoter PI in rat hepatocytes. J Biol Chem 276:16033–16039
Tanaka Y, Gleason CE, Tran PO, Harmon JS, Robertson RP 1999 Prevention of glucose toxicity in HIT-T15 cells and Zucker diabetic fatty rats by antioxidants. Proc Natl Acad Sci USA 96:10857–10862
Harmon JS, Gleason CE, Tanaka Y, Oseid EA, Hunter-Berger KK, Robertson RP 1999 In vivo prevention of hyperglycemia also prevents glucotoxic effects on PDX-1 and insulin gene expression. Diabetes 48:1995–2000
Sheth SS, Castellani LW, Chari S, Wagg C, Thipphavong CK, Bodnar JS, Tontonoz P, Attie AD, Lopaschuk GD, Lusis AJ 2005 Thioredoxin-interacting protein deficiency disrupts the fasting-feeding metabolic transition. J Lipid Res 46:123–134
Schulze PC, Yoshioka J, Takahashi T, He Z, King GL, Lee RT 2004 Hyperglycemia promotes oxidative stress through inhibition of thioredoxin function by thioredoxin-interacting protein. J Biol Chem 279:30369–30374
Hirota T, Okano T, Kokame K, Shirotani-Ikejima H, Miyata T, Fukada Y 2002 Glucose down-regulates Per1 and Per2 mRNA levels and induces circadian gene expression in cultured Rat-1 fibroblasts. J Biol Chem 277:44244–44251
Saitoh T, Tanaka S, Koike T 2001 Rapid induction and Ca(2+) influx-mediated suppression of vitamin D3 up-regulated protein 1 (VDUP1) mRNA in cerebellar granule neurons undergoing apoptosis. J Neurochem 78:1267–1276
Hasegawa J, Osatomi K, Wu RF, Uyeda K 1999 A novel factor binding to the glucose response elements of liver pyruvate kinase and fatty acid synthase genes. J Biol Chem 274:1100–1107
Vaulont S, Vasseur-Cognet M, Kahn A 2000 Glucose regulation of gene transcription. J Biol Chem 275:31555–31558
Lefrancois-Martinez AM, Martinez A, Antoine B, Raymondjean M, Kahn A 1995 Upstream stimulatory factor proteins are major components of the glucose response complex of the L-type pyruvate kinase gene promoter. J Biol Chem 270:2640–2643
Vallet VS, Henrion AA, Bucchini D, Casado M, Raymondjean M, Kahn A, Vaulont S 1997 Glucose-dependent liver gene expression in upstream stimulatory factor 2 –/– mice. J Biol Chem 272:21944–21949(Alexandra H. Minn, Christ)