Inhibition of Agouti-Related Peptide Expression by Glucose in a Clonal Hypothalamic Neuronal Cell Line Is Mediated by Glycolysis, Not Oxidative Phosphorylation

The regulation of neuroendocrine electrical activity and gene expression by glucose is mediated through several distinct metabolic pathways. Many studies have implicated AMP and ATP as key metabolites mediating neuroendocrine responses to glucose, especially through their effects on AMP-activated protein kinase (AMPK), but other studies have suggested that glycolysis, and in particular the cytoplasmic conversion of nicotinamide adenine dinucleotide (NAD+) to reduced NAD (NADH), may play a more important role than oxidative phosphorylation for some effects of glucose. To address these molecular mechanisms further, we have examined the regulation of agouti-related peptide (AgRP) in a clonal hypothalamic cell line, N-38. AgRP expression was induced monotonically as glucose concentrations decreased from 10 to 0.5 mm glucose and with increasing concentrations of glycolytic inhibitors. However, neither pyruvate nor 3-β-hydroxybutyrate mimicked the effect of glucose to reduce AgRP mRNA, but on the contrary, produced the opposite effect of glucose and actually increased AgRP mRNA. Nevertheless, 3β-hydroxybutyrate mimicked the effect of glucose to increase ATP and to decrease AMPK phosphorylation. Similarly, inhibition of AMPK by RNA interference increased, and activation of AMPK decreased, AgRP mRNA. Additional studies demonstrated that neither the hexosamine nor the pentose/carbohydrate response element-binding protein pathways mediate the effects of glucose on AgRP expression. These studies do not support that either ATP or AMPK mediate effects of glucose on AgRP in this hypothalamic cell line but support a role for glycolysis and, in particular, NADH. These studies support that cytoplasmic or nuclear NADH, uniquely produced by glucose metabolism, mediates effects of glucose on AgRP expression.

HYPOTHALAMIC NEURONS, IN particular neurons that express agouti-related peptide (AgRP), play a key role in energy balance (1). Hypothalamic AgRP was originally identified by homology to the agouti gene, whose ectopic expression leads to an obese phenotype in Ay/a mutant mice (2), suggesting a role for AgRP in the regulation of energy balance. Of particular interest, transgenic expression of AgRP causes hyperphagia and obesity through antagonism of melanocortin receptors in the central nervous system (3). Similarly, intracerebroventricular infusion of AgRP causes increased food intake and weight gain (4), and a 48-h fast stimulates hypothalamic AgRP expression (5). Furthermore, hypothalamic AgRP expression is up-regulated in leptin-deficient obese mice (6) and is down-regulated by leptin treatment (5). Nevertheless, leptin is not the only factor that regulates AgRP, because fasting induces AgRP in leptin-resistant db/db mice (5) and in streptozotocin-induced diabetic mice independent of plasma leptin and insulin (7). One factor plausibly mediating some leptin-independent effects of fasting is glucose (7), which has long been thought to regulate energy homeostasis through central mechanisms (8). Indeed, Lee et al. (9) reported that AgRP gene expression in several cell lines is inhibited by glucose, apparently mediated by AMP-activated protein kinase (AMPK). These results are consistent with a more recent study that also demonstrated that glucose regulates AMPK phosphorylation in a hypothalamic neuronal cell line (10). On the other hand, other studies have implicated nonmitochondrial reduced nicotinamide adenine dinucleotide (NADH), a unique signature of glucose metabolism, in mediating electrical (11, 12) and molecular (13) responses to glucose. In the present study based on the hypothalamic neuronal cell line, N-38, generated from primary fetal mouse hypothalamic neurons (14), effects of glucose on AgRP gene expression were independent of ATP, AMPK, and several other potential mediators, consistent with a role for nonmitochondrial NADH.

Materials and Methods

Cell culture and treatments

The N-38 hypothalamic cell line was selected from several hypothalamic neuronal cell lines described previously (14). Normally, the adherent cells were maintained in 75-cm² tissue culture flasks at 37 C with 5% CO₂ and grown in 4.5 g/liter (25 mm) d-glucose-DMEM (Cellgro of Mediatech, Herndon, VA), supplemented with 10% heat-inactivated fetal bovine serum and 1% penicillin-streptomycin. Glucose-free DMEM (Invitrogen, Carlsbad, CA) and 25 mm glucose-DMEM were mixed to make the media at several glucose concentrations. Before treatment, N-38 cells were seeded into six- or 12-well plates and initially cultured in 25 mm glucose-DMEM for 1–2 d. For low-to-high glucose switch studies, cells were exposed to 0.5 mm glucose medium for 1 d before being exposed to increasing glucose concentrations. At about 90% confluence, the cells were subject to various glucose concentrations and/or the treatments of drugs and incubated for 24 h before being harvested for total RNA extraction.

All drugs (2-deoxyglucose, sodium pyruvate, sodium lactate, sodium iodoacetate, sodium 3-hydroxybutyrate, and glucosamine) were purchased from Sigma-Aldrich (St. Louis, MO) except that 5-aminoimidazole-4-carboxamide ribonucleoside (AICAR) was purchased from Toronto Research Chemical (Ontario, Canada). They were dissolved in sterile PBS or DMSO to make 1 m or 1 mm (iodoacetate) stock solutions, which were then filter sterilized. Upon treatment, aliquots from stock solutions were diluted to corresponding culture media.

Quantitative real-time PCR (qRT-PCR)

Total RNA was isolated using TRIzol reagent (Invitrogen, Carlsbad, CA), preceded by washing cells with prewarmed 1× PBS, using a protocol for qRT-PCR as described (15). Briefly, 5 μg total RNA was converted to cDNA, with 100 pg cDNA used for each individual reaction in a 40-cycle three-step PCR using the ABI Prism 7900 thermocycler (Applied Biosystems, Foster City, CA). The PCR master mix contained 1× PCR buffer (20 mm Tris, pH 8.4, and 50 mm KCl), 5 mm MgCl₂, 200 μm dNTPs, 0.5× SYBR Green (Molecular Probes, Eugene, OR), 200 nm each primer pair, and 0.25 U Platinum Taq (Invitrogen). Amplicon size and reaction specificity were confirmed by agarose gel electrophoresis. Cyclophilin, α-tubulin, and RPS11 were used as housekeeping genes for normalization. The primer sequences for AgRP, TXNIP, FAS, L-PK, and the housekeeping genes are provided on request. Data are expressed as Ct (threshold cycle) values obtained from ABI SDS software package, and the relative quantification of mRNA levels (fold regulation, FR), which were normalized to housekeeping genes, was determined by the ΔΔCt method.

ATP quantification

N-38 cells were seeded to Nunc 96-well cell culture plates with a density of 1 × 10⁴ cells per well (100 μl). On the next day, when the culture became 50–70% confluent, cells were refreshed with media of various glucose concentrations with or without drugs of various doses. After a 12-h incubation, media were removed and cells were washed with cold 1× PBS. Cell lysates were prepared using diluted Luciferase Cell Culture Lysis 5× Reagent (Promega, Madison, WI) and transferred to solid white microplates (Costar; Corning Inc., Corning, NY) for ATP assays and to clear flat-bottom microplates (Nunc) for protein assays. Protein concentration for each well was determined using the BCA Protein Assay Kit (Pierce Biotechnology Inc., Rockford, IL). Cellular ATP concentrations per well were measured using ATP Determination Kit (Molecular Probes, Invitrogen) in an LMaxII384 microplate luminometer (Molecular Devices Corp., Sunnyvale, CA), according to the manufacturer’s instructions. Data were analyzed using SOFTmax ProV software for LMax, and the ATP concentration was calculated using the ATP standard curve, followed by normalization to protein concentration (pmol ATP/μg protein).

AMPKα[pT172] ELISA

N-38 cells were seeded to 12-well culture plates with a density of about 2 × 10⁵ cells per well. On the next day, when the culture became 70–80% confluent, cells were refreshed with 0.5 or 10 mm glucose-DMEM with or without lactate or 3-hydroxybutyrate (3-HOB). At various time points, media were aspirated and cells were washed with cold 1× PBS. Cell lysates were prepared using Biosource Cell Extraction Buffer (Biosource International, Camarillo, CA) plus protease inhibitor cocktail (Sigma, catalog no. P2714) and 1 mm phenylmethylsulfonyl fluoride, according to the manufacturer’s guidelines. Protein concentration was determined using the BCA Protein Assay Kit (Pierce). The level of AMPKα protein phosphorylated at Thr172 was quantified using the Biosource AMPKα[pT172] immunoassay kit (catalog no. KHO0651). Absorbance of each well on the assay plate was determined at 450 nm on a Beckman Coulter AD340 micro-ELISA plate reader. AMPKα[pT172] concentrations were calculated from the standard curve and normalized to total protein concentration (units per milligram protein).

siRNA transfection

SiRNA reagents targeted to AMPK α2-subunit (gene name, mouse PRKAA2; accession no. NM_178143) were purchased from Dharmacon (Lafayette, CO) or QIAGEN (Valencia, CA). Either BLOCK-iT Red Fluorescent Oligo from Invitrogen or siCONTROL nontargeting pool from Dharmacon was used as a negative control. N-38 cells were reverse transfected in a 12-well format using Lipofectamine RNAiMAX Transfection Reagent (Invitrogen) according to the manufacturer’ protocol. Cells were incubated for 48 h after transfection and then refreshed in 0.5 or 10 mm glucose media for 24 h before being harvested to assay for gene knockdown and check AgRP mRNA level by qRT-PCR.

Sample preparation for DNA microarray

N-38 cells were seeded to six-well culture plates with a density of about 1.5 × 10⁵ cells per well and fasted in 2 mm glucose media for 48 h on the next day, followed by the switch to 2 or 10 mm glucose media for 4 or 24 h. Total RNA was extracted, and 24 samples (six replicates per group) were sent to the Microarray Core Facility for processing.

Statistical analysis

Data were presented as mean ± sem (n = 4–8 replicates per group). Statistical significance between two groups was determined in Excel using Student’s t test with the significance level set at P < 0.05. Multiple comparisons were carried out with JMP software on a Macintosh computer system using one-way ANOVA followed by Tukey-Kramer honestly significant difference (HSD) test.

Results

Glucose inhibits AgRP gene expression and increases ATP in hypothalamic cell line N-38

The clonal cell line N-38 was selected for further study on the basis of its expression of glucokinase and AgRP and the relatively robust effects of glucose on AgRP gene expression. Normally, the cells are maintained in DMEM with 25 mm glucose, but when glucose levels were reduced for 24 h, AgRP expression was induced as glucose concentration decreased (Fig. 1A). Conversely, if cells were maintained at 0.5 mm glucose for 24 h, and then glucose concentrations increased, AgRP mRNA was inhibited as glucose concentration increased. Furthermore, cellular ATP was similarly dependent on glucose concentration (Fig. 1B), consistent with the hypothesis that AMP-dependent kinase, whose activity is regulated by AMP/ATP ratios, mediate effects of glucose on AgRP gene expression. However, ATP and AgRP were dissociated by other experiments, as described below.

Pyruvate and 2-deoxy-glucose (2-DG) block inhibition of AgRP expression by glucose

It has been reported that in the N1E-115 neuronal cell line, supplementation of pyruvate into 2.5 mm glucose resulted in a reduction of AgRP expression (9). We therefore examined whether pyruvate or lactate could substitute for glucose to suppress AgRP expression at 0.5 or 2 mm glucose in the N-38 hypothalamic neuronal cell line. As shown in Fig. 2, A and B, neither pyruvate nor lactate (10 mm) added to 0.5 or 2 mm glucose mimicked the effect of 10 mm glucose to inhibit AgRP expression. On the contrary, addition of 10 mm pyruvate to 10 mm glucose blocked the inhibition of AgRP expression by glucose. Similarly, addition of 10 mm 2-DG, an inhibitor of glycolysis, also blocked the inhibition of AgRP expression by 10 mm glucose and increased AgRP mRNA significantly (2.23 ± 0.28, P = 0.004). Addition of pyruvate to 2-DG produced a slightly additive effect to further induce AgRP expression (2.65 ± 0.26, P = 0.0004). Because pyruvate and lactate would be expected to produce ATP and mimic the effects of glucose on AMPK, we next examined the cellular ATP levels in the conditions described above. As shown in Fig. 2C, lactate mimicked the effect of glucose on cellular ATP, though, interestingly, pyruvate did not and, in fact, reduced ATP in the presence of 10 mm glucose, as did 2-DG. Thus, the results from pyruvate or 2-DG supplementation are consistent with but do not prove the hypothesis that ATP/AMPK mediates effects of glucose on AgRP expression, whereas the results from lactate would appear to rule out this hypothesis.

Inhibitor of glyceraldehyde 3-phosphate dehydrogenase (GAPDH) blocks inhibition of AgRP expression by glucose

Because glucose inhibited AgRP expression but pyruvate and lactate did not, we hypothesized that the metabolic signal mediating the molecular effects of glucose is generated by glycolysis. Sodium iodoacetate, a specific inhibitor of GAPDH, has been used to block glycolysis at the only step that produces cytoplasmic NADH (11). To further investigate the role of glycolysis in mediating molecular effects of glucose, iodoacetate was added to 10 mm glucose. As shown in Fig. 3, at 10 mm glucose, AgRP expression was induced (Fig. 3A), and ATP concentration was reduced (Fig. 3B) as a function of iodoacetate concentration.

The ketone body 3-HOB induces AgRP expression, increases ATP, and inhibits AMPK

Because the neurons can metabolize ketone bodies to produce ATP, if effects of glucose on AgRP gene expression are mediated by oxidative phosphorylation, ketone bodies should mimic the effects of glucose on AgRP gene expression. However, on the contrary, AgRP expression was actually dose-dependently induced by 3-HOB from the already elevated levels observed at 0.5 mm glucose (Fig. 4A) (a similar induction by 3-HOB was also observed at 10 mm glucose).

Nevertheless, 3-HOB also dose-dependently induced ATP (Fig. 4B). These results, as with the lactate results, strongly argue against a role for ATP in mediating the effects of glucose on AgRP gene expression. To assess whether a similar conclusion could be drawn regarding the role of AMPK, we directly measured AMPK phosphorylation. As shown in Fig. 5A, switching from 10 to 0.5 mm glucose induced AMPK phosphorylation, as expected. As shown in Fig. 5B, both lactate and 3-HOB inhibited AMPK phosphorylation, as expected, strongly supporting the conclusion that AMPK does not mediate the effects of glucose on AgRP gene expression.

Direct inhibition of AMPK by RNA interference induces, rather than inhibits, AgRP gene expression

To address more directly the role of AMPK in mediating effects of glucose, we assessed whether direct inhibition by RNA interference would mimic the effect of glucose to inhibit AgRP. Because the AMPKα2 is the only isoform implicated in mediating effects of glucose, we focused on this isoform. As shown in Fig. 6, double-stranded RNA complementary to the AMPKα2 isoform successfully reduced expression of that gene by more than 70% at both low and high glucose concentrations. However, rather than mimicking the effect of glucose to inhibit AgRP expression, inhibiting AMPK actually induced AgRP expression at 10 mm and had no effect at 0.5 mm glucose, possibly because expression was already maximal at that concentration. Consistent with these results, activation of AMPK by 1 or 2 mm AICAR inhibited, rather than induced, AgRP mRNA (Fig. 6C).

Glucosamine does not block the effect of glucose on AgRP but does induce several genes regulated by the pentose or hexosamine pathway

Many effects of glucose on gene expression are mediated by the hexosamine pathway, which is activated by glucosamine (16). To assess the role of the hexosamine pathway in mediating effects of glucose on AgRP, we incubated cells at 10 mm glucose with several doses of glucosamine. As shown in Fig. 7, glucosamine failed to mimic (or block) the effects of glucose to inhibit AgRP. However, glucosamine did induce liver-type pyruvate kinase (L-PK), fatty acid synthase (FAS), and thioredoxin-interacting protein (TXNIP), consistent with previous reports that FAS (17) and TXNIP (18) are induced through the hexosamine pathway. Interestingly, Txnip can also be activated by inhibitors of glycolysis by inducing the carbohydrate response element-binding protein (ChREBP) transcriptional complex dependent on the pentose pathway (19). As shown in Fig. 8, we corroborated this observation, in that 2-DG, pyruvate, and glucose strikingly induce Txnip, but 3-HOB neither induces nor inhibits Txnip. Taken together, we conclude that the hexosamine pathway and the pentose pathway mediate effects of glucose on Txnip, but not on AgRP, whose regulation by glucose appears to be mediated exclusively by glycolysis.

Discussion

The main purpose of the present study was to assess the metabolic pathways that mediate the effects of glucose to regulate hypothalamic gene expression. A precedent for such an analysis was a report that AgRP is regulated by glucose in two neuronal cell lines (9) and in a study examining effects of glucose and leptin in a similar hypothalamic cell line (10). The N-38 line was chosen for further study because it expresses glucokinase as well as AgRP and among cell lines tested was most sensitive to glucose. It is of some interest that AgRP was sensitive to glucose concentrations in the range observed in blood and that are optimal for glucokinase-mediated glucose sensing but higher than the range of glucose generally observed in brain parenchyma. Electrical activity of at least some hypothalamic neurons (as well insulin secretion from pancreatic β-cells) is also sensitive to these concentrations of glucose, suggesting that the N-38 neuronal cell line is plausibly similar to this population of hypothalamic neurons.

2-DG, an inhibitor of glycolysis, increases appetite in humans and other species, as reviewed (8). Here we demonstrate that 2-DG robustly blocked the inhibition of AgRP by 10 mm glucose and even induced AgRP above levels observed at 0.5 mm glucose (Fig. 2). Similarly, iodoacetate, a specific inhibitor of GAPDH (11), has similar effects on AgRP mRNA and cellular ATP level as 2-DG (Fig. 3). In contrast, d-glucosamine had no effect on AgRP mRNA expression, although it did induce TXNIP and FAS, which are induced by the hexosamine pathway (17, 18) (Fig. 7). It should be noted that glucosamine also inhibits glucokinase and some electrical effects of glucose (11), raising the possibility that effects of glucose in this cell line are not mediated by glucokinase. Taken together, these studies demonstrate that the inhibition of AgRP expression by glucose is mediated by glucose metabolism but not by the hexosamine pathway.

Consistent with our previous observations on electrical activity in hypothalamic neurons (11) and similar studies in pancreatic β-cells (20, 21), pyruvate failed to mimic effects of glucose and, indeed, antagonized the effect of glucose on AgRP gene expression (Fig. 2). These results are consistent with other studies demonstrating that pyruvate mimicked the effect of 2-DG to block glucose-regulated gene expression (13, 22). Because pyruvate inhibits glycolysis, these results suggest that glycolysis, not oxidative phosphorylation, may mediate some effects of glucose on AgRP gene expression. Consistent with this hypothesis, lactate mimicked the effect of glucose on ATP production but failed to mimic the effect of glucose on AgRP expression (Fig. 2). Similarly, the ketone body 3-HOB mimicked the effect of glucose on ATP production but, like pyruvate, actually increased, rather than decreased, expression of AgRP (Fig. 4). Consistent with effects on ATP, both lactate and 3-HOB mimicked the effect of glucose to inhibit the phosphorylation of AMPK (Fig. 5). Furthermore, direct inhibition of AMPK by RNA interference failed to inhibit AgRP expression, as would be predicted if AMPK mediated the effect of glucose and, in fact, like pyruvate and 3-HOB, actually stimulated AgRP gene expression (Fig. 6). Conversely, activation of AMPK by AICAR inhibited AgRP gene expression. Taken together, these studies appear to rule out a role for oxidative phosphorylation or AMPK in mediating the effects of glucose on AgRP expression.

In contrast with the present results, Lee et al. (9) reported that both inhibition of AMPK and addition of pyruvate decreased AgRP expression in two neuronal cell lines and in an ex vivo hypothalamic preparation. The most likely reason for the marked differences between present results and the results by Lee et al. (9) is the differences in the cell lines studied. It should be noted that in the paper by Lee et al. (9), the inhibition of AgRP expression by pyruvate and the induction of AgRP expression by AICAR were shown only in NIE-115 cells, whereas the results from the inhibition of AgRP with dominant-negative AMPK were shown only in GT1-7 cells. N1E-115 cells were not derived from hypothalamic neurons, in contrast to the N-38 cells. Although the GT1-7 cell line was derived from hypothalamic neurons, this cell line was derived on the basis of expression of the reproductive neuropeptide GnRH, which in vivo does not coexpress AgRP and is not involved in energy balance. With regard to the effects of pyruvate on an ex vivo hypothalamic preparation, we have carried out similar studies in hypothalamic slices and, while replicating the effects of glucose on hypothalamic gene expression, have not been able to replicate the effect of pyruvate. (Yang, X.J., M. Iijima, and C.V. Mobbs, unpublished). Thus, the present studies demonstrate that effects of glucose on AgRP expression do not necessarily involve changes in ATP or AMPK activity in a cell line relatively similar to endogenous AgRP-expressing neurons. However, as with any in vitro system, the applicability of these results to the in vivo condition remains to be determined.

Another pathway by which glucose can regulate gene expression is through the pentose pathway, via the activation of the ChREBP. One gene that appears to be induced by this pathway is Txnip, which is induced by glucose as well as inhibitors of by activation of the ChREBP pathway, which is normally activated by the pentose pathway (19). We corroborated this interesting effect with several inhibitors of glycolysis, including pyruvate (Fig. 8). Because the activation of TXNIP by inhibitors of glycolysis is thought to be mediated by activation of ChREBP, whereas these inhibitors blocked the effect of glucose to inhibit AgRP expression, these results strongly argue against a role for ChREBP or the pentose pathway in mediating effects of glucose on AgRP gene expression. Consistent with this conclusion, the AgRP promoter region does not have recognizable carbohydrate response element sequences.

It is of some interest that pyruvate, but not lactate, inhibits effects of glucose on AgRP mRNA. The simplest explanation for the different effects of lactate and pyruvate is that when pyruvate is in excess, it is partially metabolized to lactate, consuming NADH; this does not occur when lactate is metabolized to pyruvate. These results support a role for nonmitochondrial NADH in mediating effects of glucose on electrical (11, 12) and molecular (13) responses to glucose. In addition, the result with 3-HOB might also support a role for NADH in mediating molecular responses to glucose, because a recent paper (23) has suggested that 3-HOB significantly reduces cytoplasmic NADH levels. We have attempted to measure cellular NADH levels in our various experimental conditions with commercial NAD(H) quantification kits, but the results were inconclusive. We also measured cellular NAD(P)H level using fluorometry (data shown in the supplemental figure, published as supplemental data on The Endocrine Society’s Journals Online web site at http://endo.endojournals.org). Although these results are consistent with the NADH hypothesis, it should be noted that this method does not distinguish NADH and NADPH. Therefore, the NADH hypothesis still remains to be proven.

In conclusion, inhibitors of glycolysis block the effects of glucose to inhibit AgRP but mimic the effects of glucose on genes regulated by the ChREBP pathway. Furthermore, activation of the hexosamine pathway mimics the effect of glucose on genes induced by the hexosamine pathway but not AgRP. Similarly, several metabolites that mimic the effect of glucose on ATP production and AMPK phosphorylation fail to mimic the effect of glucose on AgRP gene expression, whereas inhibition of AMPK gene expression induces, and activation of AMPK activity inhibits, AgRP expression, the opposite profile as would be expected if AMPK mediated effects of glucose. Taken together, these results clearly indicate that a product of glycolysis, but not oxidative phosphorylation, the pentose pathway, or the hexosamine pathway, mediates effects of glucose on AgRP gene expression in the hypothalamic cell line N-38. We hypothesize that the production of NADH by nuclear GAPDH, through the regulation of the transcription factor C-terminal binding protein (CtBP), mediates effects of glucose on AgRP, as has been demonstrated for the regulation of neuronal brain-derived neurotrophic factor (BDNF) (13). In turn, we hypothesize that C-terminal binding protein represses AgRP gene expression by inhibiting the activity of cAMP response element-binding protein-binding protein (CBP), as previously reported (24), because there are several prominent cAMP response element-binding protein (CREB) binding elements in the AgRP promoter. These results and hypotheses are summarized in Fig. 9.

Acknowledgments

Disclosure Statement: The authors have nothing to disclose.

Abbreviations:

 
• AgRP

  Agouti-related peptide

 
• AICAR

  5-aminoimidazole-4-carboxamide ribonucleoside

 
• AMPK

  AMP-activated protein kinase

 
• ChREBP

  carbohydrate response element-binding protein

 
• FAS

  fatty acid synthase

 
• GAPDH

  glyceraldehyde 3-phosphate dehydrogenase

 
• 3-HOB

  3-hydroxybutyrate

 
• HSD

  honestly significant difference

 
• L-PK

  liver-type pyruvate kinase

 
• NADH

  reduced nicotinamide adenine dinucleotide

 
• qRT-PCR

  quantitative real-time PCR

 
• TXNIP

  thioredoxin-interacting protein.