Thioredoxin-Interacting Protein Stimulates Its Own Expression via a Positive Feedback Loop

Abstract

Thioredoxin-interacting protein (TXNIP) has emerged as a key regulator of important cellular processes including redox state, inflammation, and apoptosis and plays a particularly critical role in pancreatic β-cell biology and diabetes development. High glucose and diabetes induce TXNIP expression, whereas inhibition of TXNIP expression or TXNIP deficiency protects against pancreatic β-cell apoptosis and diabetes. We now have discovered that TXNIP stimulates its own expression by promoting dephosphorylation and nuclear translocation of its transcription factor, carbohydrate response element-binding protein (ChREBP), resulting in a positive feedback loop as well as regulation of other ChREBP target genes playing important roles in glucose and lipid metabolism. Considering the detrimental effects of elevated TXNIP in β-cell biology, this novel pathway sheds new light onto the vicious cycle of increased TXNIP, leading to even more TXNIP expression, oxidative stress, inflammation, β-cell apoptosis, and diabetes progression. Moreover, the results demonstrate, for the first time, that TXNIP modulates ChREBP activity and thereby uncover a previously unappreciated link between TXNIP signaling and cell metabolism.

Thioredoxin-interacting protein (TXNIP) has emerged as a key regulator of important cellular processes including redox state (1–5) and apoptosis (6–10), is involved in inflammasome activation (11), inflammation, and endoplasmic reticulum stress (12, 13), and plays a particularly critical role in pancreatic β-cell biology (6–9). Initially identified as the top glucose-induced gene in a human islet microarray study, we found that β-cell expression of TXNIP is increased in diabetes (7, 8, 14). Moreover, TXNIP overexpression induces β-cell apoptosis and is essential for glucotoxicity-induced β-cell death (7), whereas lack of TXNIP promotes endogenous β-cell survival and prevents type 1 and type 2 diabetes (8, 9, 15). Glucose-induced TXNIP expression is mediated via a nonpalindromic E-box motif consisting of a repeat of CACGAG sequences in the TXNIP promoter that serves as the binding site for the carbohydrate response element-binding protein (ChREBP) (16).

The basic helix-loop-helix transcription factor ChREBP has been recognized as the main transcription factor mediating glucose-induced gene expression in liver (17) as well as in pancreatic β-cells (16), whereas its paralog, MondoA, has been shown to do the same in muscle (18), and a recently discovered isoform, ChREBP-β, has been described in adipocytes (19). To date, ChREBP/MondoA remains the only bona fide nutrient/glucose-responsive transcription factor known to control various target genes involved in glucose and lipid metabolism (13, 14). The activity of ChREBP is regulated by its phosphorylation status and cellular localization and, in particular, dephosphorylation of Ser196 (as observed in response to glucose) allows ChREBP to enter the nucleus and transactivate its target genes such as liver-type pyruvate kinase (L-PK) or TXNIP (17, 20, 21).

Surprisingly, we now have discovered that TXNIP can induce its own transcription, and the present studies were aimed at identifying the mechanisms involved. In addition to revealing the processes controlling this positive feedback loop and promoting TXNIP expression, we discovered that TXNIP enhances ChREBP dephosphorylation and nuclear localization and induces ChREBP target gene expression, uncovering a novel cross talk mechanism between these two key signaling pathways.

Materials and Methods

Cell culture and islet isolation

INS-1 cells were grown in RPMI 1640 medium (Invitrogen) supplemented with 10% fetal bovine serum, 1% penicillin/streptomycin, 1 mM sodium pyruvate, 2 mM l-glutamine, 10 mM HEPES, and 0.05 mM 2-mercaptoethanol. The INS-hTXNIP cell line stably transfected with a human (h) TXNIP expression plasmid and the control INS-LacZ cell line overexpressing LacZ have been described previously (9).

Mouse islets were isolated from male 8-week-old, wild-type C57BL/6 mice (The Jackson Laboratory) or from obese, insulin-resistant, and diabetic BTBRlep^ob/ob (BTBRob/ob) and control BTBRlep^+/+ (BTBRlean) mice (8) by collagenase digestion as described previously (7, 9). All mouse studies were approved by the University of Alabama at Birmingham Animal Care and Use Committee.

Quantitative real-time RT-PCR

Total RNA was extracted using RNeasy kit (QIAGEN) according to the manufacturer's instructions. RNA (1 μg) was reverse transcribed to cDNA using the first strand cDNA synthesis kit (Roche).

Quantitative real-time PCR was performed on a Prism 7000 Sequence Detection System using Sybergreen (Applied Biosystems). The various species-specific TXNIP primers (9) are listed in Supplemental Table 1, and their specificities are shown in Supplemental Table 2. (Of note, rat and mouse primers were not able to detect endogenous or transfected in, hTXNIP. In contrast, specially designed rat/human primers detected both endogenous rat and transfected in hTXNIP.) All samples were corrected for the 18S ribosomal subunit (Applied Biosystems) run as an internal standard.

Plasmid construction and transient transfection assays

Construction of the various TXNIP promoter and simian virus 40-carbohydrate response element (ChoRE) reporter plasmids were described previously (9) as were the TXNIP and ChREBP overexpression plasmids (9, 16).

C57BL/6 mouse islets (∼200) were transfected with hTXNIP expression or control LacZ plasmids (1 μg/mL) using DharmaFECT Duo transfection reagent (5 μL/2 mL medium in a suspension tissue culture dish) and harvested 48 hours later for RNA extraction.

INS-hTXNIP and INS-LacZ cells were grown in 12-well plates and transfected with the full-length TXNIP promoter reporter plasmid or control simian virus 40 plasmid (0.4 μg/well) using DharmaFECT Duo transfection reagent 1 (1 μL/well; Dharmacon/Thermo Scientific). To determine the TXNIP promoter region responsible for the TXNIP effect, INS-1 cells were transfected with different TXNIP promoter deletion reporter plasmids (0.4 μg/well) together with pRL-TK control plasmid (20 ng/well; Promega Corp) using DharmaFECT Duo. Cells were harvested 24 hours after transfection, and luciferase activity was determined by Dual Luciferase assay kit (Promega Corp).

Chromatin immunoprecipitation (ChIP)

ChIP assays were performed as previously detailed (16). In brief, 500 μg of cross-linked INS-1 protein extracts (by BCA protein assay) were incubated overnight at 4°C with 4 μg of goat anti-ChREBP (sc-21189) antibodies or normal goat IgG (sc-2028) (Santa Cruz Biotechnology). Immune complexes were captured with 50 μL of 50% protein A-Sepharose agrose slurry (Sigma). DNA fragments were purified using a Qiagen PCR purification kit and quantified by real-time PCR with primers as described previously (16) and listed in Supplemental Table 1.

RNA interference

INS-1 cells were grown in 6-well plates and transfected with 3 specific small interfering RNA oligos for rat TXNIP (siTXNIP; Dharmacon; siGENOME SMAPRTpool gene identification no. 117514) rat ChREBP (siChREBP; Invitrogen; Mlxxipl-RSS301430) or scrambled oligo (0.1 μM) (Dharmacon, D-001810–01–20) using DharmaFECT transfection reagent (Dharmacon/Thermo Scientific) (5 μL per well) as described previously (16, 22). The final concentration of oligos used was 25 nM. Cells were harvested after 48 hours for RNA and protein extraction.

Cell fractionation, Western blotting, and immunohistochemistry

Nuclear and cytoplasmic protein extracts were prepared as described previously (7, 23). The rabbit anti-ChREBP IgG (1:200; sc-33764), and goat antirabbit IgG (1:5000, sc-2004) antibodies were used (Santa Cruz Biotechnology). Phosphorylated ChREBP was detected as described previously (24) using a specific phospho-ChREBP antibody detecting only ChREBP phosphorylated at Ser196, the site critical for controlling nuclear translocation (25) (generous gift of Dr C. Postic, Paris, France). Phosphorylated 5′-AMP-activated protein kinase (phospho-AMPK) and total AMPK were detected using the phospho-AMPK (1:1000; Cell Signaling Technology; catalog no. 2535) and total AMPK (1:1000; Cell Signaling Technology; catalog no. 5831) antibodies, respectively. β-Actin (1:5000; Abcam; antibody 3280) was run as a control. Bands were visualized by ECL Plus (Amersham GE Health) and quantified by ImageQuant. Immunohistochemistry for insulin and ChREBP was performed as previously described (8, 23).

Statistical analysis

Student's t tests were used to calculate the significance of a difference between 2 groups. For data sets of more than 2 groups, we performed one-way-ANOVA calculations.

Results and Discussion

TXNIP induces its own transcription

Using our rat INS-1 β-cells stably transfected with a hTXNIP expression plasmid (INS-hTXNIP) we were able to achieve effective overexpression (Supplemental Figure 1, A and B) with an increase in total TXNIP mRNA and TXNIP protein levels comparable to the approximately 4-fold increase observed in diabetes (22) assuring physiological relevance of this model. Taking advantage of this unique model and the ability to differentially measure human, rat, or mouse TXNIP mRNA in a species-specific manner (Supplemental Tables 1 and 2), we discovered that TXNIP overexpression induces endogenous TXNIP mRNA expression (Figure 1A). Transient transfection of hTXNIP into native rat INS-1 β-cells or rat H9C2 cardiomyocytes also resulted in a significant increase in endogenous rat TXNIP (Supplemental Figure 2, A and B) demonstrating that the effect was not restricted to our stable cell lines or to β-cells and also occurred in other cell types. Moreover, we confirmed these findings in primary mouse islets transiently transfected with our human TXNIP expression plasmid (Figure 1B). In addition, we found that this TXNIP induction in response to TXNIP overexpression occurred at the transcriptional level as demonstrated by the increase in TXNIP promoter-driven luciferase activity (Figure 2A), but not of a heterologous promoter (Figure 2B). Because TXNIP has not been found to bind to DNA and act as a transcription factor, the observed effects likely required involvement of other factors, and we therefore set out to identify the cis- and trans-acting factors conferring TXNIP-induced TXNIP transcription.

Figure 1.

TXNIP effects on β-cell TXNIP expression. A, Rat INS-1 β-cells overexpressing hTXNIP or control (LacZ) were assessed for endogenous rat TXNIP expression using quantitative RT-PCR and species-specific rat primers. B, Isolated, primary wild-type (C57BL/6J) mouse islets were transfected with hTXNIP or LacZ control plasmid and 48 hours later endogenous mouse TXNIP was measured by quantitative RT-PCR and mouse-specific primers. Bars represent means ± SEM; n = 3–4.

Figure 2.

Analysis of TXNIP promoter activity and cis-acting element conferring TXNIP effects. TXNIP effects on hTXNIP promoter activity (panel A) or simian virus 40 (SV40) promoter activity (panel B) as assessed by luciferase activity in INS-LacZ and INS-hTXNIP cells transfected with the respective reporter plasmids. C, Promoter deletion study using the full-length hTXNIP promoter (FL) and a number of deletions/mutations (D1–D5) transfected into INS-hTXNIP and INS-LacZ cells. Black boxes represent E-box motif ChoRE. Bars represent mean fold change in luciferase activity ± SEM of at least 3 independent experiments as compared with INS-LacZ control cells. mut, mutant; N.S., nonsignificant.

An E-box ChoRE motif in the TXNIP promoter is required for TXNIP-induced TXNIP transcription

To identify the promoter region responsible for TXNIP-induced TXNIP transcription, we performed promoter deletion studies and found a highly significant induction of the hTXNIP promoter unless the E-box motif of the ChoRE was mutated or deleted (mutD4, D5) (Figure 2C). (Although the small effects of TXNIP on mutD4 and D5 were not significant, we cannot exclude the possibility that TXNIP may also have some minor effects on basal transcription.) Moreover, this ChoRE was able to confer TXNIP responsiveness to a heterologous promoter demonstrating that this E-box motif is not only necessary but also sufficient for the observed TXNIP effects (Supplemental Figure 2C). Of note, this is the same E-box repeat we previously identified to be responsible for glucose-induced TXNIP expression and to serve as the binding site for ChREBP (9, 16). Obviously, this raised the possibility that ChREBP might also be mediating the TXNIP effects.

TXNIP increases ChREBP binding to the TXNIP (and L-PK) promoters

Indeed, we found that TXNIP overexpression increases ChREBP occupancy of the TXNIP promoter in vivo as demonstrated by ChIP assays, whereas no binding was observed to the glyceraldehyde 3-phosphate dehydrogenase internal control, and the IgG control showed no enrichment confirming the specificity of these ChIP assays (Figure 3). Furthermore, TXNIP also induced ChREBP binding to the promoter of the bona fide ChREBP target gene, L-PK (17, 26) (Figure 3B) and increased L-PK expression (Figure 3D) demonstrating that the effects are not restricted to TXNIP, but also include other ChREBP target genes. These findings provide additional support for the idea that ChREBP is mediating the observed downstream effects of TXNIP on gene expression (in addition to serving as the main upstream regulator of β-cell TXNIP expression.)

Figure 3.

ChREBP binding to target gene promoters in response to TXNIP. INS-LacZ and INS-hTXNIP cells were maintained in regular growth medium and cross linked, and ChIP assays were performed as described in Materials and Methods. TXNIP effects on ChREBP occupancy of the TXNIP (panel A) and L-PK (panel B) promoter. C, IgG negative control ChIP. D, TXNIP effects on L-PK mRNA level as measured by quantitative RT-PCR. Bars represent mean fold change ± SEM of at least 3 independent experiments. GAPDH, glyceraldehyde 3-phosphate dehydrogenase.

TXNIP inhibits ChREBP phosphorylation and promotes nuclear ChREBP

The transcriptional activity of ChREBP is primarily regulated by its cellular localization, nuclear entry, and posttranslational modification (17, 20, 21, 25). Therefore, we next investigated whether TXNIP affected nuclear ChREBP levels. Indeed, TXNIP promoted nuclear localization of ChREBP as shown by the increased ChREBP levels in nuclear cell fractions, the increased nuclear/cytoplasmic ratio, and immunohistochemistry showing increased nuclear staining for ChREBP in INS-hTXNIP cells (Figure 4A–C and Supplemental Figure 4A). Furthermore, transfection of INS-1 cells with TXNIP small interfering RNAs not only resulted in effective TXNIP knockdown (Supplemental Figure 3A), but also led to a dramatic reduction in nuclear ChREBP and its nuclear/cytoplasmic ratio (Supplemental Figure 3, B and C), further confirming the important role TXNIP plays in regulating nuclear ChREBP localization. Translocation of ChREBP into the nucleus has been shown to require a number of dephosphorylation steps (20, 21), in particular dephosphorylation of Ser196 (25). We therefore also investigated whether TXNIP decreased ChREBP phosphorylation and indeed found a clear decline in Ser196 phosphorylation in response to TXNIP (Figure 4D and Supplemental Figure 4B).

Figure 4.

TXNIP effects on ChREBP phosphorylation and nuclear localization. INS-LacZ and INS-hTXNIP cells were maintained in regular growth medium prior to fractionation into nuclear and cytoplasmic protein extracts and measurement of ChREBP protein levels by Western blotting. TXNIP effects on nuclear ChREBP protein levels (panel A) and nuclear to cytoplasmic ChREBP ratio (panel B). C, Visualization of ChREBP (green) and insulin (red) in INS-LacZ and INS-hTXNIP cells by fluorescent immunohistochemistry (×100). D, Assessment of phosphorylated ChREBP (p-Ser196-ChREBP) in INS-LacZ and INS-hTXNIP cell extracts and quantification of the ratio of phosphorylated ChREBP to total ChREBP. E, Islets of diabetic and obese BTBR ob/ob mice with a known increase in TXNIP levels were assessed for ChREBP expression and compared with islets of control BTBR lean mice using quantitative RT-PCR. F, Immunohistochemistry of pancreatic cross sections and visualization of ChREBP protein (blue) in a control BTBR lean islet and a diabetic BTBR ob/ob islet (×40); higher magnification of respective nuclei (insets). Bars represent means ± SEM of at least 3 independent experiments.

The phosphorylation status of ChREBP has been shown to be mainly regulated by protein phosphatase 2A (PP2A) (27) and by AMPK (28, 29). We therefore first used the protein serine/threonine phosphatase inhibitor okadaic acid and indeed found that it was able to counteract the TXNIP-induced expression of endogenous TXNIP as well as of L-PK (Supplemental Figure 5, A and B) raising the possibility that activation of PP2A was involved in the observed TXNIP effects. However, analysis of PP2A activity revealed no difference in response to TXNIP overexpression (Supplemental Figure 5C) making it very unlikely that PP2A was mediating TXNIP-induced TXNIP expression. We therefore next investigated the possibility that AMPK might be involved. In fact, we found that the AMPK activator 5-amino-1-β-d-ribofuranosyl-imidazole-4-carboxamide completely blunted TXNIP-induced TXNIP expression and L-PK expression (Figure 5, A and B). Moreover, if TXNIP were conferring its effects via AMPK, we would expect that TXNIP would lead to decreased phosphorylation and activation of AMPK and that is exactly what we observed (Figure 5C and Supplemental Figure 4C). These findings suggest that TXNIP promotes ChREBP-mediated transcription (including TXNIP and L-PK) through inhibition of AMPK phosphorylation/activation and thereby provide mechanistic insight into this novel pathway. However, because TXNIP is not a phosphatase, its inhibitory effects on AMPK phosphorylation are likely to be indirect and to involve yet additional factors (eg, inhibition of upstream AMPK-kinase (30), activation of protein phosphatases such as PP1-R6 (31) and/or allosteric modulation (30)). Interestingly, the level of TXNIP-induced TXNIP expression and L-PK expression was also significantly reduced in response to siChREBP as compared with scrambled control (Supplemental Figure 6). This further supports the notion of ChREBP activity playing an important role in the observed TXNIP effects on TXNIP transcription. Furthermore, we also explored the in vivo effects of increased TXNIP on ChREBP using diabetic BTBRob/ob mice. We have previously shown that BTBRob/ob islets have a 4-fold increase in TXNIP expression as compared with control BTBRlean islets (22). Now we found that this is accompanied by a significant increase in ChREBP in BTBRob/ob islets (Figure 4, E and F). These results are consistent with a recent report of increased ChREBP in diabetic human β-cells (32) as well as with the suggested role of ChREBP in our newly discovered pathway and provide an intriguing pathophysiological link to diabetes.

Figure 5.

Role of AMPK in TXNIP-induced gene expression. Effects of the AMPK activator 5-amino-1-β-d-ribofuranosyl-imidazole-4-carboxamide (AICAR) (1 mM for 24 hours), on TXNIP-induced TXNIP expression (panel A) and TXNIP-induced L-PK expression (panel B) as assessed by quantitative RT-PCR in INS-LacZ and INS-hTXNIP cells. C, TXNIP effects on phosphorylated AMPK as assessed by Western blotting and compared with total AMPK; β-actin is shown as a loading control. Bars represent means ± SEM of 3 independent experiments. N.S., nonsignificant.

In peripheral tissues such as muscle and adipose tissue, TXNIP has been shown to inhibit glucose uptake via inhibition of glucose transporter 1 (GLUT1) (33–35), but, as such, is less likely to have any major suppressive effects on glucose uptake and flux in pancreatic β-cells. In fact, if it did, one would expect a negative feedback on TXNIP expression because glucose is a strong stimulus for TXNIP transcription. However, our data demonstrate that TXNIP overexpression results in increased ChREBP-mediated TXNIP transcription in pancreatic β-cells suggesting a positive feedback loop. These results indicate that TXNIP overexpression can bypass and override any potential minor effects from decreased intracellular glucose, most likely via the observed inhibition of AMPK and activation of ChREBP-mediated transcription. Together, these findings suggest that TXNIP induces its own expression by promoting dephosphorylation, nuclear translocation, and DNA binding of its transcription factor ChREBP. Moreover, the results demonstrate, for the first time, that TXNIP inhibits AMPK phosphorylation/activation, modulates ChREBP activity and thereby regulates other ChREBP target genes such as L-PK, revealing a previously unappreciated cross talk between these important signaling pathways.

In summary, we have discovered the existence of a positive feedback loop regulating TXNIP expression that is active in pancreatic β-cells and primary islets as well as in other cell types (Figure 1 and Supplemental Figure 2). Taking into account the crucial role that elevated TXNIP levels play in β-cell glucose toxicity (7), endoplasmic reticulum stress (12, 13), apoptosis (6–9), and the pathogenesis of diabetes (8), these findings support the notion that TXNIP levels rise over time not only as a result of elevated blood glucose levels and/or endoplasmic reticulum stress, but also as part of a vicious cycle by which increased TXNIP levels lead to more TXNIP expression and thereby amplify the associated detrimental effects on β-cell biology including oxidative stress, inflammation, and ultimately β-cell death and disease progression. Moreover, we have identified the mechanism conferring this feedback loop and found that TXNIP promotes the activity of its own transcription factor, ChREBP, which in turn induces transcription of other ChREBP target genes playing important roles in glucose and lipid metabolism. This suggests that this novel pathway has implications even beyond TXNIP expression and reveals a thus-far unrecognized link between TXNIP signaling and cell metabolism.

Additional material

Supplementary data supplied by authors.