Thioredoxin-interacting protein promotes high-glucose-induced macrovascular endothelial dysfunction

Abstract

Thioredoxin-interacting protein (TXNIP) emerges as a central regulator for glucose homeostasis, which goes awry in diabetic subjects. Endothelial dysfunction is considered the earliest detectable stage of cardiovascular disease (CVD), a major complication of diabetes. Here, we hypothesize that TXNIP may promote endothelial dysfunction seen in Type 1 diabetes mellitus (T1D). Using a T1D-like rat model, we found that diabetic rats showed significantly higher TXNIP mRNA and protein levels in peripheral blood, compared to their non-diabetic counterparts. Those changes were accompanied by decreased production of nitric oxide (NO) and vascular endothelial growth factor (VEGF), concurrent with increased expression of reactive oxygen species (ROS) and vascular cell adhesion molecule 1 (VCAM-1) in the aortic endothelium. In addition, TXNIP overexpression in primary human aortic endothelial cells (HAECs) induced by either high glucose or overexpression of carbohydrate response element binding protein (ChREBP), a major transcriptional activator of TXNIP, promoted early apoptosis and impaired NO bioactivity. The correlation between TXNIP expression levels and endothelial dysfunction suggests that TXNIP may be a potential biomarker for vascular complications in T1D patients.

1. Introduction

The American Heart Association (AHA) outlined in its 2011 Scientific Statement the traditional risk factors for cardiovascular disease (CVD) in children, which include constitutional factors (eg, family history of atherosclerosis, age, sex), behavioral/lifestyle factors (eg, nutrition/diet, physical inactivity, tobacco exposure, perinatal exposures), physiological factors (eg, blood pressure, lipids, obesity, glucose metabolism and insulin resistance), and medical diagnoses (eg, Type 1 and Type 2 diabetes mellitus, chronic/end-stage kidney disease) [1]. Although traditional risk factors are validated in many clinical trials for the diagnosis and management of CVD, they are not able to fully explain the increased morbidity and mortality from CVD seen in individuals with Type 1 diabetes (T1D) [2–5]. There is therefore a need to develop novel, reliable biomarkers that can identify patients at the highest risk for clinical CVD outcomes, as highlighted in both the 2005 Report of the National Heart, Lung, and Blood Institute–National Institute of Diabetes and Digestive and Kidney Diseases Working Group on Cardiovascular Complications of Type 1 Diabetes Mellitus, and the 2014 Scientific Statement from the American Heart Association and American Diabetes Association [6,7]. According to the 2011 AHA Scientific Statement on nontraditional risk factors and biomarkers for CVD applied to children and adolescents, biomarkers are defined as biological indicators for processes that are involved in developing a disease that may or may not be causal [1].

Endothelial dysfunction is a well-accepted central event that precedes the development of early atherosclerosis and occurs even in children with T1D [2]. This pathophysiological condition is characterized by an imbalance between antiatherosclerotic substances, the most characterized of which is nitric oxide (NO), and endothelium-dependent vasoconstricting factors, such as endothelin 1, prostanoids, and reactive oxygen species (ROS) [8,9]. Factors that can lead to endothelial dysfunction include hyperglycemia, increased circulating fatty acid levels, altered lipoproteins, and derivatives of glycation and oxidation [6]. Hyperglycemia itself reduces the bioavailability of NO, which may explain the impairment in endothelium-dependent vasodilatation shown in T1D subjects [6].

Thioredoxin-interacting protein (TXNIP), a central regulator for glucose homeostasis, has been found to play a role in the patho-physiological process of atherosclerosis associated with diabetes-induced endothelial dysfunction. TXNIP suppresses thioredoxin activity, promotes intracellular ROS production, and triggers early apoptosis in high-glucose treated primary human aortic endothelial cells (HAECs) [10,11]. In addition, T1D-related impairment of ischemia-mediated angiogenesis is endothelial TXNIP-dependent [12]. More importantly, recent population studies have indicated that TXNIP genetic variation is associated with diabetes and hypertension in the general Brazilian population [13], and that plasma levels of TXNIP are correlated with carotid artery intima-media thickness, an indicator of atherosclerosis in patients with early-state Type 2 diabetes (T2D) and impaired glucose tolerance [14]. Given that TXNIP is implicated in the pathogenesis of CVD in diabetic patients, this preclinical study was directed at exploring the potential of TXNIP as a novel biomarker for T1D-associated endothelial dysfunction.

2. Materials and methods

2.1. Animals

Male Sprague Dawley rats (weight 250–300 g, age 6–8 weeks) were purchased from Envigo (Indianapolis, USA). Diabetes was induced by a one-time intraperitoneal injection of streptozotocin (70 mg/kg; Sigma-Aldrich, St. Louis, USA), as we previously published [11], and defined as random blood glucose levels of greater than 250 mg/dL for three consecutive days. At the end of the four week treatment period, the rats were humanely euthanized. All the animal experiments were approved by the University of Missouri-Kansas City Institutional Animal Care Use Committee.

2.2. Cell culture

Primary human aortic endothelial cells (HAECs) were obtained from Lonza (Walkersville, USA) and cultured in EGM2 medium according to the manufacturer's instructions. Cells at passage 4–6 were used. Unless otherwise indicated, the glucose concentrations of 5.5 mM and 30 mM were defined as normal glucose and high glucose conditions in this study, respectively. After reaching confluence, HAECs were exposed to normal glucose (NG) or high glucose (HG) for 48 h with a daily change of culture media.

2.3. Transient transfection and adenoviral infection

For transient transfection, HAECs were grown in six-well plates and transfected with ChREBP (carbohydrate responsive element-binding protein) - or LacZ-expression plasmids as described previously [11]. Human ChREBP cDNA (Accession# BC012925) was cloned into the pcDNA3.1 vector for the overexpression of ChREBP in HAECs. Adenoviral infection was performed overnight in 80%-confluent cell culture with media (EBM-2; Lonza) containing 1% fetal bovine serum (FBS) and recombinant adenovirus at the desired multiplicity of infection (MOI), after which the virus-containing medium was removed and replaced by fresh EGM-2 medium (Lonza) for 24 h before treatments. Using these conditions, infection efficiency was typically at least 90%, as determined by GFP expression. The adenoviruses used in this study are the pre-packaged, ready-to-use adenoviruses Ad-GFP (Cat#1060) and Ad-GFP-h-TXNIP (Cat#ADV-226923) purchased from Vector Biolabs (Malvern, USA).

2.4. Western blot, real-time PCR and laboratory analyses

Western blot analysis was performed as described previously [11]. The PVDF membranes were incubated with anti-ChREBP (Cat#sc-21189; Santa Cruz Biotechnology, USA), anti-TXNIP (Cat#40–3700; Invitrogen, USA), anti-VEGF (Cat#ab46154; abcam, USA), anti-VCAM-1 (Cat#AF643; R&D Systems, Minneapolis, USA), anti-Cleaved Caspase-3 (Cat#9661S; Cell Signaling Technology, USA), or anti-β-Actin (Cat#A5441; Sigma-Aldrich) at 4°C overnight. Immunoreactive bands were quantified by the NIH ImageJ software. Total RNA was purified from the fresh, whole blood samples, according to the manual of the QIAamp RNA Blood Mini Kit (Qiagen, Hilden, Germany). Reverse transcription reactions and q-RT PCR were performed as described previously [11]. The primers used for TXNIP and β-actin were described previously [11]. TXNIP content in serum was assessed with a Thioredoxin Interacting Protein (TXNIP) ELISA Kit (Cat#ABIN825475, Antibodies-online, Atlanta, USA), following the manufacturer's instructions. The laboratory analyses of serum glucose, triglycerides, and cholesterol were performed in the Department of Pathology and Laboratory Medicine of Children's Mercy Hospital.

2.5. Terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) staining and caspase-3 activity assay

For TUNEL staining, HAECs were grown on cover slips in EGM2 medium. After the transient transfection, the medium was replaced by EBM-2 supplemented with 10% FBS with 5 mM or 30 mM glucose for 72 h with a daily medium change. Apoptosis was assessed by TUNEL staining using an In Situ Cell Death Detection kit from Roche (11684809910; Mannheim, Germany). The number of stained cells was assessed using the NIH ImageJ software. Percentage apoptosis was calculated from the number of TUNEL-positive cells divided by the total number of cells counted. Six images randomly obtained from two wells of a six-well plate were analyzed for each treatment group in each independent experiment.

For caspase-3 activity assay, the cells were harvested for the assay after the 24 h incubation in the EBM-2 supplemented with 10% FBS with 5 mM or 30 mM glucose. Caspase-3 activity was assessed in whole cell lysates with an EnzChek Caspase-3 Assay Kit #1 (Molecular Probes, Eugene, OR), following the manufacturer's instructions. Values for activity were normalized for cell protein.

2.6. Assessment of NO bioactivity

The transfected or adenovirus-infected HAECs were further exposed to 5 mM or 30 mM glucose for 48 h. NO bioactivity was assessed in whole cell lysates with a Direct cyclic GMP (cGMP) ELISA kit (Enzo Life Sciences, Farmingdale, USA), following the manufacturer's instructions. Values were normalized for cell protein.

2.7. Assessment of NO accumulation and intracellular reactive oxygen species (ROS) generation

Rat aortic rings (~3 mm in length) were incubated in PBS containing dihydroethidium (DHE) (10 μM; Molecular Probes, Eugene, OR) at 37 °C for 20 min for ROS accumulation assay. The rings were then rinsed with PBS three times and further incubated in PBS containing DAF-FM diacetate (10 μM; Molecular Probes, Eugene, OR) at 37 °C for 30 min for NO accumulation assay. The rings were then rinsed with PBS three times and replaced at 37 °C for 15 min, after which the rings were stimulated by Acetylcholine (10 μM; Sigma-Aldrich, USA) for 10 min and then snapped frozen with OCT embedding compound in liquid nitrogen. All the frozen rings were cut into 8 μm sections and imaged using a Zeiss LSM 510 Meta confocal microscope at excitation/emission maxima of 495/515 nm for NO or 518/605 nm for ROS.

2.8. Immunostaining and confocal microscopy

Frozen aortic ring sections were fixed in 4% paraformaldehyde for 10 min, permeabilized with 0.1% Triton X-100 for 10 min, and then blocked in Tris-buffered saline (TBS) containing 1% BSA for 1 h. Next, the sections were incubated with a rabbit anti-TXNIP (1:50), a goat anti-VCAM-1 (1:40), a rabbit anti-Cleaved Caspase-3 (1:100), or a goat anti-ChREBP antibody (1:50) in the blocking buffer overnight at 4 °C, followed by the further incubation with a Alexa Fluor 647 goat anti-rabbit IgG (1:1000; Invitrogen) or a Alexa Fluor 594 rabbit anti-goat IgG (1:1000; Invitrogen) for 1 h. Finally, the sections were counterstained with Hoechst 33342 (Invitrogen) and mounted in the ProLong antifade reagent (Invitrogen). Cell imaging was performed on a Zeiss LSM 510 Meta confocal microscope fitted with a PlanApo × 63 oil immersion objective.

2.9. Statistical analysis

Results represent means ± SEM. P values were calculated by 1-way ANOVA followed by post hoc analysis for data sets from triplicate experiments using GraphPad Prism 6. Statistical significance was set at p < 0.05.

3. Results

3.1. Circulating TXNIP concentrations correlate with endothelial dysfunction observed in T1D-like diabetic rats

Based on our previous study [11] we hypothesized that TXNIP promotes high-glucose-induced macrovascular endothelial dysfunction. To test that hypothesis, we examined whether an in vivo correlation exists between the circulating TXNIP level and endothelial dysfunction observed in the aorta.

The T1D-like diabetes in rats was induced by one-time intra-peritoneal injection of streptozotocin (STZ) without any further insulin treatment. Compared to the normal rats, these diabetic rats showed increased levels of triglycerides and cholesterol in serum after four weeks under hyperglycemia (Fig. 1A). As expected, higher TXNIP mRNA expression and TXNIP concentrations were found in whole blood and serum of the diabetic rats, respectively (Fig. 1A).

Fig. 1.

Circulating TXNIP concentrations correlate with endothelial dysfunction observed in a T1D-like rat model. (A), Fresh whole blood was collected from the saline-injected control (n = 5) and streptozotocin-induced diabetic (n = 6) groups after 4-week treatments. Serum glucose, triglycerides, and cholesterol were assessed by a chemistry laboratory (See Methods). TXNIP mRNA levels in whole blood and TXNIP content in serum were analyzed by qRT-PCR and a TXNIP ELISA Kit, respectively. qRT-PCR data were corrected for β-actin and quantified. (B), Fresh aortas were collected from the saline-injected control (n = 5) and streptozotocin-induced diabetic (n = 6) groups after 4-week treatments. Aortic rings were incubated in PBS containing dihydroethidium (10 μM) followed by DAF-FM diacetate (10 μM) for the assessments of reactive oxygen species (ROS) and nitric oxide (NO) accumulation, respectively. In situ expression analysis of TXNIP, VCAM-1, ChREBP, and cleaved caspase-3 was performed by immunofluorescence staining using frozen aortic ring sections. The endothelial monolayer is marked by white arrows in individual images. Representative images from three independent experiments are shown. Scale bar, 10 μm. (C) The protein levels of TXNIP, VEGF, ChREBP, VCAM-1 and cleaved caspase-3 were assessed by western blot analysis. Representative western blots from three independent experiments are shown. The protein levels corrected for β-actin and quantified using ImageJ software (NIH) are also shown. Data from three independent experiments represent means ± SEM (n = 3). *p < 0.01, versus the normal rats.

Accompanying these significant changes in circulating TXNIP were dramatic decreases in NO and VEGF (vascular endothelial growth factor) levels (Fig. 1B and C), and apparent increases in ROS, TXNIP, VCAM-1 (vascular cell adhesion protein 1) and cleaved caspase-3 productions in the endothelial monolayer of the aortas (Fig. 1B and C), which manifested characteristic endothelial dysfunction concurrent with impaired angiogenesis. Unexpectedly, that obvious induction of VCAM-1 and cleaved caspase-3 expression failed to be observed in whole aorta lysates where smooth muscle proteins predominated (Fig. 1C). This discrepancy implicates that endothelium is the major site responsible for the earliest detectable changes in the aorta under hyperglycemia.

3.2. TXNIP overexpression induced by ChREBP overexpression triggers endothelial dysfunction

Given that TXNIP mediates high-glucose-induced changes associated with endothelial dysfunction [11], and that ChREBP is a major transcription factor for TXNIP transcription in endothelial cells [11], we investigated whether TXNIP overexpression in HAECs can be achieved by ChREBP overexpression. As expected, TXNIP protein level was significantly increased in ChREBP-overexpressing HAECs created by transiently transfecting HAECs with a human-ChREBP-expressing construct, in a plasmid DNA concentration-dependent fashion (Fig. 2A). The ChREBP overexpression groups have greater percentage of TUNEL-positive cells under normal glucose conditions, which mimicked the high-glucose-treated, LacZ-transfected HAECs where TXNIP overexpression was induced by high glucose (Fig. 2A, B and C). We next performed an assay for caspase-3 activity, an early signal of apoptosis, to confirm that apoptosis was occurring. Consistent with the results of TUNEL staining, the cleaved caspase-3 levels and caspase-3 activities were notably increased in HAECs by both high glucose and ChREBP-overexpression (Fig. 2A and D), which indicates that TXNIP overexpression forces endothelial cells into early apoptosis. As a result of that early apoptosis, the cGMP level, a cellular indicator of NO bioactivity, was concomitantly decreased in both the high-glucose-treated HAECs and the ChREBP-overexpressing HAECs (Fig. 2E).

Fig. 2.

TXNIP overexpression accelerates early apoptosis and impairs NO bioactivity in HAECs. (A) to (E), Confluent HAECs were transiently transfected with a LacZ-expressing plasmid (control vector) or a human-ChREBP-expressing plasmid at different concentrations, and then exposed to 5.5 mM (NG) or 30 mM (HG) glucose for 24 h (D), 48 h (A and E) or 72 h (B and C). (A), The protein levels of TXNIP, ChREBP and cleaved caspase-3 were assessed by western blot analysis. (B), TUNEL staining of HAECs under various conditions. The enlarged images indicate the TUNEL-positive cells with fragmented nuclei. Representative images from three independent experiments are shown. Scale bar, 50 μm. (C) The percentage of TUNEL-positive cells was determined from six randomly selected fields (× 200) in two wells of a six-well plate for each group and was quantified using the ImageJ software. (D) Caspase-3 activities in HAECs under various conditions. (E) Intracellular cyclic GMP (cGMP) levels in HAECs under various conditions. (F) to (H), HAECs were infected with the adenoviruses encoding GFP (Ad-GFP; control vector) or human TXNIP (Ad-TXNIP) at different MOIs prior to their exposure to 5.5 mM (NG) or 30 mM glucose (HG) for 48 h (F and H) or 24 h (G). (F), The protein levels of TXNIP and cleaved caspase-3 were assessed by western blot analysis. (G) Caspase-3 activities in the infected HAECs. (H) cGMP levels in the infected HAECs. Representative western blots from three independent experiments are shown. The protein levels corrected for β-actin and quantified using the ImageJ software (NIH) are also shown here. Data from three independent experiments represent means ± SEM (n = 3). *p < 0.01, versus the control vector under normal glucose conditions.

3.3. TXNIP overexpression induced by TXNIP-encoding adenovirus infection triggers endothelial dysfunction

We further validated the effects of TXNIP overexpression by infecting HAECs with the adenovirus encoding human TXNIP (Ad-TXNIP) at different MOIs. Under normal glucose conditions, the Ad-TXNIP-expressing cells demonstrated a MOI-dependent increase through MOI 200 in cleaved caspase-3 levels and caspase-3 activities, showing similar phenotypes seen in high glucose-treated cells (Fig. 2F and G). Overexpression of Ad-TXNIP or exposure to high glucose also significantly reduced the cGMP level (Fig. 2H). All these data taken together verify the observed early apoptosis caused by ChREBP-overexpression-induced TXNIP overexpression.

As both increased endothelial apoptosis and impaired NO bioactivity contribute to endothelial dysfunction [15], all these findings suggest that TXNIP is directly related to endothelial function, and that endothelial dysfunction observed in a diabetes-like environment maintains a close association with TXNIP over-expression induced by high glucose.

4. Discussion

In this study, we established a correlation between blood TXNIP concentrations and endothelial dysfunction observed in a T1D like rat model. After four weeks of uncontrolled hyperglycemia, the diabetic rats showed evidence of vascular endothelial dysfunction, such as impaired NO bioactivity and oxidative stress. TXNIP was up-regulated in peripheral blood as well as in the aorta, including the aortic endothelium and smooth muscle cells. However, the up-regulation of a vascular inflammation marker, VCAM-1, and an early apoptosis marker, cleaved caspase-3, were only detected in the endothelial monolayer by in situ immunostaining but not in the whole aorta lysates by western blot analysis (Fig. 1B and C). This discrepancy may be explained by the clinical finding that endothelial dysfunction is the earliest detectable stage of CVD.

In the clinical setting, there is good evidence that most, if not all, cardiovascular risk factors are associated with endothelial dysfunction and it can be considered a barometer of the total risk burden [16]. Thus, the assessment of endothelial function is deemed important in identifying individuals at increased risk of future acute cardiovascular events [17]. By the use of EndoPAT, a finger plethysmography for the noninvasive assessment of peripheral microvascular endothelial function, T1D Exchange investigators detected lower mean reactive hyperemia-peripheral artery tonometry (RH-PAT) scores in T1D adolescents compared to their non-diabetic counterparts, indicating endothelial dysfunction correlated with T1D [18]. While EndoPAT is user-friendly with low inter-observer and intra-observer variability [9], the RH-PAT score is only partly dependent on NO [9,19]. Due to the complexity of endothelial dysfunction as a systemic condition [9,20], there is as yet no specific method that has been recommended in clinical guidelines for planning primary or secondary prevention of vascular disease [9]. Of note, endothelial function measurements should not rely on a single test but rather on the average of several tests [9]. Taking that guideline into account, the measurement of circulating endothelial dysfunction markers, including VCAM-1, intercellular adhesion molecule-1 (ICAM-1), and endothelial leukocyte adhesion molecule-1 (E-selectin), has been considered a good supplement to vascular imaging techniques and measures of blood flow. However, the evaluation of their additional predictive value to classical risk factors provided conflicting results [21,22]. There is therefore a critical need for new insights into the molecular regulation of endothelial dysfunction within endothelial cells, for the purpose of identifying novel markers that may serve as useful adjuncts to classical risk factors.

Circulating TXNIP may be such a potential biomarker, based on our findings and its demonstrated relation to PCOS (polycystic ovary syndrome), carotid intima-media thickness in patients with T2D, and risk of coronary artery disease (CAD) [14,23,24]. It also meets some of the 2011 AHA evaluation criteria of novel biomarkers for CVD in children [1]. As shown in this study, the peripheral blood TXNIP level correlates with endothelial dysfunction and can be measured either by real-time PCR or ELISA. The data are reliable, accurate, and reproducible, and the use of commercially available TXNIP ELISA kits is standardized, convenient, cost-effective and available for practical and widespread application. Moreover, TXNIP concentrations in plasma correlates with carotid intima-media thickness, an indicator of atherosclerosis, in patients with impaired glucose tolerance and early T2D [14]. Meanwhile, TXNIP can be down-regulated by metformin, a first-line therapy for T2D, in both HAECs and T1D-like rats [11]. These findings taken together implicate that TXNIP provides independent information on risk or prognosis of CVD in adults, accounts for a clinically significant part of CVD in adults, and is responsive to a treatment or intervention. However, to fully drive TXNIP into clinical application, it is necessary to have it meet every standard provided in the 2011 AHA Scientific Statement [1]. For that purpose, large-scale clinical trials and long-term follow-up are required to establish a solid relationship between blood TXNIP levels and preclinical vascular findings/disease. Furthermore, we need to determine whether blood TXNIP levels in childhood are consistent with those in adults, and to validate its sensitivity, specificity as well as responses to metformin treatment in adults for CVD events. Also, autopsy studies are still required to determine whether blood TXNIP levels correlate with atherosclerosis extent [1].

Overall, TXNIP is still far from its application to pre-diagnosis of CVD at this time. However, its great potential to become a biomarker has come to the fore. The role of TXNIP in the context of T1D-induced CVD deserves intensive scientific efforts, including both translational and clinical research.