Tumor Necrosis Factor-α–Induced Iron Sequestration and Oxidative Stress in Human Endothelial Cells

Atherosclerosis now is generally recognized as a chronic inflammatory condition, and inflammatory cytokines such as tumor necrosis factor (TNF)-α have been associated with the development of atherosclerotic lesions and consequent cardiovascular events.^1,2 Dysfunction and loss of vascular endothelial cells, which provide a nonthrombogenic surface and a permeability barrier, occur early in atherosclerosis.³

Several lines of evidence has suggested that TNF-α–induced cell injury is mediated through its ability to promote intracellular reactive oxygen species (ROS) formation.^4,5 Among these species, superoxide anion (O₂⁻) and hydrogen peroxide (H₂O₂) are not very reactive, and usually are neutralized by an elaborate antioxidant defense system. However, transition metal-catalyzed Haber-Weiss reaction can transform O₂⁻ and H₂O₂ to hydroxyl radical, which are extremely powerful oxidizing species.^6,7

Iron is an essential element required for biochemical reactions subserving a wide variety of functions in cells and organisms.⁸ However, free iron, possibly the most important transition metal in biologic systems, can act as an electron donor.^6,9 Excessive intracellular accumulation of iron therefore could amplify the damaging effect of oxidative stress in inflammatory conditions, leading to cell injury.

The recent identification of iron transport proteins has rapidly expanded our knowledge of molecular aspects of iron processing, especially in the reticuloendothelial system and small intestine.^10,11 Such proteins include transferrin receptor (TfR), divalent metal transporter 1 (DMT1) and ferroportin 1 (FP1), which tightly regulate intracellular iron. TfR localizes to the plasma membrane, binds transferrin, and delivers transferrin-bound iron into cells by endocytosis.¹² DMT1 is a transmembrane protein centrally involved in absorption of ferrous iron and intracellular delivery of iron.¹³ As a collaborating counterpart to TfR and DMT1, FP1 can export iron across the basolateral membrane and donate the metal to transferrin and to the circulation after oxidation by a membrane-bound ferroxidase.¹⁴

Ludwiczek et al have demonstrated effect of TNF-α or other cytotoxic agents on iron transport proteins in human monocytic cell lines.¹⁵ We have reported that dysregulation of iron transport proteins, such as induction of TfR and suppression of FP1, led to intracellular iron excess in polymorphonuclear leukocytes (PNMLs) from hemodialysis (HD) patients.¹⁶ Because HD patients often show hypercytokinemia, we suspected that dysregulation of iron transport proteins in PMNL might be a consequence of markedly increased serum concentrations of TNF-α and interleukin (IL)-6. These iron transport proteins presumably are expressed ubiquitously and regulated similarly throughout the body. Our preliminary experiments demonstrated the presence of iron transport proteins in human umbilical vein endothelial cells (HUVEC). We therefore hypothesized that TNF-α could cause dysregulation of iron transport proteins and iron accumulation in these cells and vascular endothelium in general, where it might promote iron-dependent cellular oxidative damage.

In the present study, we sought to characterize the TNF-α on iron metabolism in HUVECs, including expression of iron transport proteins as well as intracellular iron regulators. TNF-α–induced hydroxyl radical production also was assessed in the presence or absence of iron.

Methods

Materials

MCDB131 medium were purchased from GIBCO (Grand Island, NY). Recombinant human TNF-α, human transferrin, and apo-transferrin were purchased from Sigma Chemical (St. Louis, Mo). ⁵⁹FeCl₃ (ferric chloride) and ⁵⁹Fe-citrate (ferrous citrate) were obtained from Perkin Elmer Life Sciences (Boston, Mass).

Cell Culture

HUVECs were isolated by collagen digestion and cultured in modified MCDB131 medium containing 10% fetal calf serum as previously described.¹⁷ Passages 3 to 8 of these cells were used for the experiments. Cultures of subconfluent cells were treated with 10 or 50 ng/mL TNF-α in fresh medium.

Relative Quantitation of Transporter mRNAs by TaqMan Real-Time Polymerase Chain Reaction

Total RNA was isolated from HUVECs using an RNA extraction reagent (Isogen-LS; Nippon Gene, Tokyo, Japan), according to the manufacturer’s instruction. For determining amounts of TfR, DMT1, and FP1 mRNAs in HUVECs, relative quantitation was performed using a polymerase chain reaction (PCR) with TaqMan polymerase and an ABI PRISM 7900HT sequence detection system (Applied Biosystems, Foster City, Calif). TaqMan real-time PCR primers and TaqMan probes were designed using Primer Express software (Applied Biosystems).

Amplification was performed in MicroAmp optical 384-well reaction plates (Applied Biosystems) in an ABI PRISM 7900HT sequence detection system. Amplification conditions were 2 minutes at 50°C and 10 minutes at 95°C, followed by 50 cycles of 15 seconds at 95°C and 1 minute at 58°C.

To correct for variation in total RNA content and unequal reverse transcription efficiency, TfR, DMT1, and FP1 quantities were normalized to the amount of mRNA encoding GAPDH.

Immunocytochemistry

The immunostaining protocol was essentially as described by Waheed et al.¹⁸ Briefly, HUVECs were cultured on glass coverslips and fixed in 4% paraformaldehyde. After blocking, cells were incubated with rabbit anti-mouse DMT1 or FP1 IgG (Alpha Diagnostics, San Antonio, Tex) or mouse anti-human TfR IgG (Dako, Glostrup, Denmark), diluted 1:50 in phosphate-buffered saline (PBS). Bound IgG was detected with a 1:100 dilution of fluorescein isothiocyanate (FITC) conjugated swine anti-rabbit IgG or FITC conjugated rabbit anti-mouse IgG (Dako), and photographed with an LSM510 microscope (Carl Zeiss, Jena, Germany) equipped for chemiluminescence.

SDS-PAGE and Western Blotting

HUVECs were lysed in RIPA lysis buffer (Upstate, Lake Placid, NY) containing protease inhibitor cocktail (Complete; Roche Diagnostics, Tokyo, Japan). Cytosolic and nuclear extracts were isolated accordingly. Fifty micrograms of cytosolic protein was subjected to SDS-PAGE under reducing conditions according to Laemmli,¹⁹ and then transferred to a nitrocellulose membrane (Hybond; Amersham Biosciences, Buckinghamshire, UK). Membranes were blocked with 0.05% Tween 20 and 5% skim milk in PBS, and then incubated with rabbit anti-mouse DMT1 or FP1 IgG or mouse anti-human TfR IgG diluted 1:1000 in PBS, containing 0.05% Tween 20. The secondary reaction was performed using HRP-conjugated donkey anti-rabbit IgG or a HRP-conjugated sheep anti-mouse IgG (Amersham Biosciences) diluted 1:2000. Bound antibodies were visualized by ECL detection reagents (Amersham Biosciences) and quantified by an LAS-1000 Lumino image analyzer (Fuji Film, Tokyo, Japan).

Iron Uptake and Release Studies

Labeled diferric transferrin was prepared as described previously.¹² Before starting treatment, the medium was replaced with modified MCDB131 medium containing 2% fetal calf serum; HUVEC then were allowed to adjust to these conditions. To study uptake of transferrin-bound iron (TBI) or nontransferrin-bound iron (NTBI), either 12.5 μg/mL of ⁵⁹Fe-labeled transferrin or 5 μmol/L of ⁵⁹Fe-citrate was added to the medium as described previously.¹⁵ After treatment, radioactivity in the cell pellet was determined by a gamma counter and normalized by total cell protein.

For iron release experiments, cells treated with the appropriate additives for 24 hours were incubated with a 1:1 mixture of ⁵⁹Fe-labeled transferrin and ⁵⁹Fe-citrate for 4 hours, then the release of radiolabeled iron was determined for up to 4 hours. The radioactivity of supernatant and cell fraction was counted, and the relative percentage of total radioactivity in supernatant was calculated.

Electrophoretic Mobility Shift Assay

The binding activity of iron regulatory protein (IRP) 1 to the iron responsive element (IRE) was measured by electrophoretic mobility shift assay using an RNA Gel Shift Kit (Fermentas Life Science, Hanover, Md) according to the manufacturer’s instructions. ³²P-labeled IRE RNA probe (59 nucleotides) was synthesized using as a control template pTZ19/EcoRI containing the human ferritin IRE sequence by in vitro transcription with T7 RNA polymerase. This transcript was incubated with 5 μg of cytosolic extracts for 30 minutes at room temperature. The total amount of IRP1 present in active and inactive (aconitase) forms in each sample was assessed using 2-mercaptoethenol (ME) at a final concentration of 2%. Analysis of RNA-protein complex was performed by nondenaturating gel electrophoresis and subsequent autoradiography.

Measurement of Aconitase Activity

Aconitase activity was measured by a coupled assay in which formation of NADPH was monitored spectrophotometrically. Under appropriate conditions, the rate NADPH production is proportional to aconitase activity. One milliunit of aconitase activity was defined as the amount catalyzing formation of 1 nmol of isocitrate per minute as previously described.²⁰ In brief, activity was measured in 1.0 mL of a reaction mixture containing 50 mmol/L Tris-HCl at pH 7.4, 5 mmol/L sodium citrate, 0.6 mmol/L MnCl₂, 0.2 mmol/L NADP⁺, 2 U of isocitrate dehydrogenase, and 50 μg of extract protein. The linear change in absorbance was determined at 340 nm (ε₃₄₀=6.22 mmol/L per cm).

Quantifying Hydroxyl Radical Production

Demonstration of hydroxyl radical production during iron challenge used a salicylate trap method involving high-performance liquid chromatography (HPLC) with electrochemical detection, as described previously in the literature.²¹

HUVEC were reseeded in 6-well cell culture clusters with modified MCDB131 medium containing 2% fetal calf serum, and allowed to adhere for 2 hours. After treatment, cells were incubated with reaction buffer containing 10 mmol/L Na salicylate, 1.2 mmol/L MgCl₂, and 1.2 mmol/L CaCl₂ in PBS for 30 minutes. Samples were diluted 1:1 with the HPLC mobile phase (30 mmol/L citrate/acetate buffer at pH 3.6/19% methanol) and immediately analyzed for specific hydroxylation products (2,3-, 2,4- and 2,5-dihydroxybenzoic acids; DHBA). The total DHBA concentrations were normalized by total cell protein.

Statistical Analysis

All values are presented as mean±SEM. Comparisons between groups were performed with unpaired t tests and factorial ANOVA with post hoc comparison using the Bonferroni correction. P<0.05 was considered to indicate statistical significance. Statistical calculations were performed with a personal computer using StatView for Windows (version 5.0; SAS Institute, Cary, NC).

Results

Effects of TNF-α on TfR, DMT1, and FP1 mRNAs in HUVEC

The TfR/GAPDH and DMT1/GAPDH mRNA ratios were significantly increased in HUVEC treated with TNF-α in a dose-dependent manner (TfR; control, 82.2±3.8%; TNF-α 10 ng/mL, 111.0±8.7%; TNF-α 50 ng/mL, 157.0±4.5%, P<0.05; DMT1; control, 103.2±5.1%; TNF-α 10 ng/mL, 169.2±5.3%; TNF-α 50 ng/mL, 250.4±4.3%; P<0.05). In contrast, the FP1/GAPDH mRNA ratio dose-dependently decreased (91.8±4.9%, 36.4±7.0%, and 28.6±6.0%, respectively; Figure 1).

• Download figure
• Download PowerPoint

Immunocytochemistry for TfR, DMT1, and FP1

Immunocytochemical staining of HUVEC demonstrated augmentation of immunoreactivity for TfR and DMT1 protein in TNF-α-treated HUVEC. Staining for both proteins was more intense in HUVEC treated with 50 ng/mL TNF-α than those treated with 10 ng/mL TNF-α (Figure 2A and 2B). In contrast, cellular FP1 staining was attenuated by TNF-α treatment; in particular hardly any FP1 staining was visible in 50 ng/mL TNF-α-treated HUVEC (Figure 2C).

• Download figure
• Download PowerPoint

TfR, DMT1, and FP1 Protein Content

Content of TfR, DMT1, and FP1 protein was semiquantified by Western analysis. As shown in Figure 3, bands representing TfR, DMT1, and FP1 proteins were recognized at ≈95 kDa, 65 kDa, and 60 kDa, respectively. After a 24-hour exposure of HUVEC to 10 or 50 ng/mL TNF-α, cellular expression of the TfR protein dose-dependently increased 1.4-fold or 1.8-fold, respectively (Figure 3A), whereas DMT1 protein also increased (1.9- or 2.6-fold, respectively; Figure 3B). In contrast, expression of FP1 protein was substantially reduced in TNF-α–treated HUVEC (Figure 3C).

• Download figure
• Download PowerPoint

Effects of TNF-α on Iron Uptake and Release in Endothelial Cells

For the purpose of examining the effect of TNF-α on uptake of TBI as well as NTBI, we respectively measured uptake of ⁵⁹Fe-labeled transferrin and ⁵⁹Fe-citrate into HUVEC. The 24-hour exposure of HUVECs to 10 ng/mL TNF-α resulted in a significant increase (27.5%) in ⁵⁹Fe-labeled transferrin uptake compared with cells without treatment. Treatment with 50 ng/mL TNF-α led to a more pronounced increase (68.3%) in cellular uptake of ⁵⁹Fe-labeled transferrin (Figure 4A). Treatment of HUVECs with TNF-α also increased uptake of ⁵⁹Fe-citrate in a dose-dependent manner (Figure 4B). Moreover, iron release from TNF-α-treated cells was substantially decreased (control, 101.5±2.8%; TNF-α 10 ng/mL, 43.6±8.8%; TNF-α 50 ng/mL, 40.9±2.5%; P<0.05) (Figure 4C).

• Download figure
• Download PowerPoint

Binding Activity of IRP to IRE

IRE-binding activity of IRP was analyzed by an RNA electrophoretic mobility shift assay in extracts of TNF-α–treated cells. Treatment with TNF-α increased IRE binding activity in a dose- and time-dependent manner [Figure 5A; 2-ME (−)]. When cell extracts were incubated with 2% 2-ME, inactive IRP1 was activated to the high-affinity RNA-binding form, resulting in a value representing total IRP1.²² With 2-ME, no major differences in IRP1 binding to IRE were observed between control and TNF-α–treated cells [Figure 5A, 2-ME (+)].

• Download figure
• Download PowerPoint

TNF-α–Induced Inactivation of Aconitase

In the absence of TNF-α, aconitase activity of HUVEC did not change during a 24-hour period. Treatment of HUVECs with 10 ng/mL TNF-α decreased aconitase activity in a time-dependent manner over 24 hours. The decline in aconitase activity was more prominent in cells treated with 50 ng/mL TNF-α (Figure 5B).

Iron and TNF-α–Induced Hydroxyl Radical Production

When the medium did not contain TNF-α, addition of transferrin (50 μmol/L) or ferrous sulfate (FeSO₄; 50 μmol/L) did not affect DHBA production; in the absence of iron, addition of 10 ng/mL TNF-α did not affect production of DHBA. However, addition of 50 ng/mL TNF-α was associated with a significant increase in DHBA production, which might be caused by uptake of iron from culture medium containing 2% FCS (≈0.6 to 0.8 μmol/L). When iron was present, either as transferrin or as FeSO_4, the effect of 50 ng/mL TNF-α on DHBA production in iron-treated cells was more prominent than that of 10 ng/mL TNF-α (Figure 6).

• Download figure
• Download PowerPoint

Discussion

In the present study we demonstrated for the first time that iron transporters including TfR, DMT1, and FP1 coexist in HUVEC, whereas TNF-α–induced synthesis of iron import proteins (TfR and DMT1) and reduced that of the export protein FP1. These effects would be expected to cause accumulation of iron in the endothelial cells. These changes in iron transporters are likely to be regulated posttranscriptionally, as we noted that TNF-α enhanced binding of IRP1 to IRE. Another striking observation was that synthesis of hydroxyl radical increased significantly only in the presence of both iron and TNF-α.

Cellular iron homeostasis is accomplished by synchronized regulation of iron import proteins and iron export protein. Expression of these iron transporters is coordinately and reciprocally controlled at a posttranscriptional level by interactions between IRP and RNA stem-loop structures known as IREs, which are found within untranslated regions (UTR) of mRNAs of iron transporters. TfR and DMT1 mRNA contain multiple IREs and a single IRE in the 3′ UTR, respectively. By contrast, FP1 mRNA contains a single IRE in the 5′UTR.^10,13,14,23

IRP1, a central cytoplasmic regulator of cellular iron metabolism, is regulated by unusual iron–sulfur cluster switching and then by a bifunctional protein. In cells with abundant iron, IRP1 forms a cuboid 4Fe-4S cluster, which loses its IRE-binding activity. Moreover, this cluster converts IRP1 to a cytosolic aconitase. In iron-starved cells, however, the cluster disassembles and IRP1 acquires IRE-binding activity. IRP/IRE interactions are believed to result in stabilization of otherwise unstable TfR and DMT1 mRNAs, whereas inhibiting translation of the mRNA encoding FP1.^10,24,25 As a result, iron-starved cells accentuate their capacity to take up TBI and NTBI while minimizing iron release, whereas iron-replete cells decrease iron uptake and increase iron release.

We presently observed that IRE-binding activity of IRP1 increased in a time-dependent manner in TNF-α–treated HUVECs (Figure 5A), in conjunction with augmentation of expression of TfR and DMT1 but repression of FP1. Thus, the changes mimicked the state of cellular iron starvation. Earlier studies have shown that elevation of intracellular ROS, including O₂⁻ and H₂O₂, trigger dissociation of a single iron atom from the 4Fe-4S cluster, resulting in conversion of cytosolic aconitase to IRE-binding protein.^22,26,27 Furthermore, ROS production by endothelial cells is known to be increased by TNF-α.^4,5 Thus, we suspected that TNF-α could induce binding activity of IRP1 to IRE via the intracellular ROS production, thus and modulating expression of iron transport proteins. Further investigation might be necessary for clarifying early mechanisms leading to the TNF-α effect on IRP1 protein.

We also demonstrated that aconitase activity decreased significantly in TNF-α–treated cells (Figure 5B), representing a mirror effect of enhancement of IRP1 binding activity with IRE. In TNF-α–treated HUVECs, inactivation of aconitase may in turn inhibit the citric acid cycle and consequent energy production, which could result in endothelial damage.

We presently observed that combined exposure of HUVECs to TNF-α and iron, either as TBI or NTBI, accelerated hydroxyl radical production compared with cells treated with iron or TNF-α alone (Figure 6). These data support the hypothesis that intracellular iron is required for generation of the potent hydroxyl radical in TNF-α–mediated cell injury, although in the present study we did not determine the levels of intracellular free iron but observed only the increase in iron uptake. These results are consistent with previous findings that TNF-α–induced and IL-1β–induced cytotoxicity was enhanced by iron accumulation in alveolar epithelial cells.²⁸ Involvement of intracellular oxidants and iron in TNF-α–mediated cytotoxicity also was supported by previous observations that antioxidants as well as iron chelators inhibited TNF-α-induced cytolysis.^29,30 In themselves, O₂⁻ and H₂O₂ are poorly reactive oxidants easily limited by multiple enzymatic antioxidant defense systems. However, in the presence of iron, these oxidants can be activated to form hydroxyl radical via the iron-catalyzed Haber-Weiss reaction or Fenton reaction. Hydroxyl radical is cytotoxic because of its ability to initiate lipid peroxidation, damage membranes, oxidize sulfhydryl compounds, and inactivate enzymes and transporters.^11,31

In endothelial cells, TNF-α-induced gene expression of adhesion molecules including vascular cell adhesion molecule (VCAM)-1, intercellular adhesion molecula-1, and E-selectin has been found to be mediated by intracellular ROS generation.^4,32 In addition, monocyte chemoattractant protein-1 production apparently was required for iron-mediated generation of hydroxyl radical via the Haber-Weiss reaction.⁴ These reports suggested that abundant intracellular iron might serve as a common early signal activating gene expression for these adhesion molecules and chemokines. Therefore, TNF-α–induced intracellular iron accumulation as observed in the present study could be involved in promoting expression of several atherogenic genes that tend to propagate endothelial damage and promote its progression to atherosclerosis.

Elevated body iron stores have been reported to be associated with increased risk of myocardial infarction in a large cohort study.³³ Transferrin and iron deposits have been shown to be prominent in human atherosclerotic lesions;^34,35 this epidemiologic and pathologic evidence would appear to support participation of iron in development of vascular disease. Another clinical study suggested that the redox-activity of iron might contribute to endothelial dysfunction in patients with atherosclerosis, considering reports of beneficial effects of iron chelation with deferoxamine on endothelial function in patients with coronary artery disease.³⁶ Recently several lines of evidence suggested that cellular iron signaling and iron-mediated oxidative damage are relevant to cardiovascular disease.³⁷ Although these observations had suggested that iron might accelerate endothelial dysfunction and atherosclerosis, direct evidence had never been reported. We presently demonstrated that intracellular iron sequestration was induced by TNF-α, in association with accelerated production of hydroxyl radical in endothelial cells. However, further investigation should be necessary for verifying the clinical relevance of cytokine-induced iron sequestration to the prevalence of atherosclerosis.

In summary, TNF-α induced upregulation of iron import proteins and downregulation of iron export protein in HUVEC via activation of IRP1. This modification of iron transport proteins could cause accumulation of iron in endothelium, increasing oxidative stress that might contribute importantly to the pathophysiologic processes of atherosclerosis and cardiovascular disease.

Footnotes