Introduction
Cellular stress plays a crucial role in the development of atherosclerosis. Oxidative stress, one of the main factors in the process of atherogenesis, is caused by reactive oxygen species that induce inflammation, promote leukocyte recruitment, and inhibit endothelial cell (EC) proliferation [1,2,3,4]. Furthermore, oxidative stress induced in the mitochondria causes the accumulation of mitochondrial DNA damage and the progressive dysfunction of the respiratory chain, which results in apoptosis and favors plaque rupture [5]. By providing antioxidant activity and promoting ATP synthesis, respectively, thioredoxins and ATP synthase protect cells from oxidative damage and increase ATP levels to maintain normal mitochondrial functions [6,7,8]. The family of heterogeneous nuclear ribonucleoproteins (hnRNPs) is also involved in the response to oxidative stress: by specifically binding RNA motifs, hnRNPs stabilize mRNA transcripts such as endothelial nitric oxide synthase and vascular endothelial growth factor (VEGF) and regulate protein expression under oxidative stress [9,10].
Endoplasmic reticulum (ER) stress also contributes to atherogenesis. In advanced atherosclerosis, the accumulation of free cholesterol induces ER stress that causes macrophage apoptosis, stimulates inflammation in vulnerable plaque, and promotes plaque destabilization [11,12]. In addition, the accumulation of misfolded/damaged proteins that aggregate in the lumen of the ER causes ER stress and dysfunction. Glucose-related proteins (GRPs), located in the lumen of the ER, are responsible for properly folding proteins and degrading misfolded/ubiquitinated proteins. Under normal conditions, GRPs are induced when ER stability is threatened to prevent ER stress-induced cell death [13,14,15].
Human plasma LDL can be chromatographically divided by charge into 5 subfractions, L1–L5. L5 is the most negatively charged LDL subfraction found in the circulation of individuals with risk factors for coronary artery disease, including hypercholesterolemia [15,16,17,18,19]. L5 has been shown to have biologic effects similar to those of oxidized LDL (ox-LDL), which is oxidized in vitro. L5 and ox-LDL are more electronegative than unoxidized LDL, and electronegative LDL contains higher proportions of conjugated dienes and thiobarbituric acid [20]. Copper-oxidized LDL distinctively contains products of modified apolipoprotein (apo) B100, whereas L5 contains primarily apoB100, as well as apoCIII, apoAI, and apoE [15,16].
We previously showed that L5 is potentially atherogenic, evident by its ability to induce inflammation, EC dysfunction, leukocyte recruitment, monocyte differentiation, smooth muscle cell proliferation, and foam cell formation [15,16,17,21]. We also found that atheroma-derived LDL induced oxidative stress and ER stress in human umbilical vein ECs (HUVECs) when compared to native LDL [22].
The results of the present study further indicate a role for L5 in inducing cellular stress that can lead to atherogenesis. We show evidence that L5 promotes inflammation, mitochondrial dysfunction, and reduced cell viability, whereas L1 does not. Furthermore, we compared the mRNA and protein expression profiles of 7 stress-related proteins in HUVECs treated with either L5 or L1. Given that these proteins are known to function in RNA metabolism, repair of the ER, and the mitochondria, we discuss how L5 may orchestrate downstream signaling that leads to atherogenesis.
Subjects and Methods
Subjects
Our study conforms to the principles outlined in the Declaration of Helsinki. We isolated LDL from the plasma of patients with homozygotic familial hypercholesterolemia (2 women, 33 and 41 years old, and 1 boy, 6 years old) [16,21]. All donors gave informed consent.
Isolation of LDL Subfractions
We separated LDL from the plasma of patients into subfractions (L1–L5) by using anion exchange chromatography [15,16,22].
Cell Culture
HUVECs were grown in supplemented M199 medium (Biochrom AG) in a humidified atmosphere containing 5% CO2. Cells were propagated by trypsin digestion at the confluent stage and used for experiments between passages 3 and 5 [22].
Cell Treatment
HUVECs grown to 90% confluence were maintained in medium containing 5% fetal calf serum and were treated with phosphate-buffered saline (PBS) or 50 µg/ml of L1 or L5 for 24 h as previously described [22].
Detection of Monocyte Chemotactic Protein-1
Monocyte chemotactic protein-1 (MCP-1), an indicator of inflammation, was detected either by Western blot (described below) or by using an enzyme-linked immunosorbent assay (ELISA) kit (BioSource). MCP-1 secreted into the medium was measured and estimated spectrophotometrically at 450 nm.
Western Blotting
Preparation of whole-cell lysates and Western blotting was performed as described previously [22]. Antibodies specific for GAPDH, Grp58, Grp78, Grp94, ORP150, ATP synthase, hnRNP C1/C2, peroxiredoxin 3 (Prdx3), inositol-requiring enzyme 1 (IRE-1), and phosphorylated protein kinase RNA-like ER kinase (p-PERK; Santa Cruz Biotechnology) were used to detect the corresponding proteins. Signals were visualized and quantified by using Quantity One software (Amersham Biosciences).
Cell Viability
As previously described [22], cell viability was measured by using the MTT [3-(4,5-dimethyl-2-yl)-2,5-diphenyltetrazolium bromide] method. HUVECs were grown in 24-well plates at a density of 5 × 104 cells/well. Cells grown to 90% confluence were treated with PBS, L1, or L5 (50 µg/ml each) for 24 h, washed, and incubated with MTT (0.5 mg/ml) for 8 h. DMSO (1 ml) was then added to solubilize the formazan salt, which was measured spectrophotometrically at 562 nm. The data, calculated relative to L1, represent the mean from 5 independent experiments. p < 0.05 was considered significant.
JC-1 Staining and Flow Cytometry
Changes in the mitochondrial membrane potential (DCm) after different treatments were studied by staining with the cationic dye JC-1 (5,5′,6,6′-tetrachloro-1,1′,3,3′-tetraethylbenzimidazol-carbocyanine iodide). JC-1 accumulates in the mitochondria of healthy cells and fluoresces red (560 nm). When the DCm collapses, JC-1 uptake is limited to the cytoplasm where it fluoresces green (530 nm). Hence, a collapse in the DCm is indicated by a reduction in the red/green fluorescence-intensity ratio. HUVECs were pretreated with PBS or 50 µg/ml of L1 or L5 for 1 h in a cell culture incubator and then incubated with 10 mg/ml JC-1 (Sigma-Aldrich Biotechnology) for 20 min at room temperature. Cells were washed twice with PBS and observed by fluorescence microscopy (Delta Vision, Applied Precision).
In a parallel experiment, cells treated in the same manner as above were digested with trypsin, washed twice with PBS, and then analyzed on an LSR flow cytometer (Becton Dickinson) to detect green fluorescence at excitation/emission wavelengths of 485/530 nm and red fluorescence at excitation/emission wavelengths of 485/590 nm.
Two-Dimensional Electrophoresis and Image Analysis
Two-dimensional (2-D) electrophoresis was performed as described previously [22]. Although over 50 proteins were originally identified to have altered expression in response to L5 [22], only 7 proteins related to cellular stress are examined in this study (table 1): ER proteins Grp94, Grp78, Grp58, and ORP150; RNA-related protein hnRNP C1/C2; and mitochondrial proteins ATP synthase and Prdx3. Protein spots were detected automatically and confirmed manually; then, the volume of each spot was normalized as a fraction of the volume of all the spots on the gel and expressed relative to the same value for the L1 group. Each intergroup comparison of samples was carried out on 4 replicate gels.
In-Gel Digestion and Mass Spectrometry
Protein spots were destained, dehydrated, and digested with trypsin as previously described [22]. Peptides were eluted in 0.8 ml of matrix solution (α-cyano-4-hydroxy cinnamic acid; 8 mg/ml in 70% v/v acetonitril/1% formic acid) directly onto a target plate and subjected to analysis by mass spectrometry. A QStar hybrid quadrupole time-of-flight mass spectrometer (QTOF; Applied Biosystems) equipped with a matrix-assisted laser desorption ionization (MALDI) source and a nitrogen laser (337 nm) was used to acquire MALDI-MS and -MS/MS spectra. The MS/MS spectra were used to automatically search the NCBInr protein database by using MASCOT software (http://www.matrixscience.com/). Parameters used for identification were described previously [22].
Reverse Transcription Polymerase Chain Reaction
RNA was isolated from ECs by using RNAzol (TEL-TEST). One microgram of total RNA was used for reverse transcription with PowerScript reverse transcriptase (Clontech), oligo-dT primers, and random primers. One-twentieth of the cDNA was then used as template for PCR. Amplification of cDNA was performed in 50 µl of PCR Master Mix (Becton Dickinson) containing 10 µM forward primers and reverse primers and approximately 50 ng of cDNA. PCR cycling parameters were as follows: 1 cycle at 94°C for 1.5 min, and 30 cycles at 94°C for 30 s, 64°C for 30 s, and 72°C for 1 min. The primer sequences for the genes examined were
(1) Grp58: forward primer 5′-CGAATGTTGAGTCTCTGGTGAA-3′ and reverse primer 5′-CAAAGTAATCCTGCAGGAACCT-3′, which amplified a 513-bp cDNA fragment;
(2) Grp78: forward primer 5′-CTCGAATTCCAAAGATTCAGCAACT-3′ and reverse primer 5′-CTCCACAGTTTCAATACCAAGTG-3′, 299 bp;
(3) Grp94: forward primer 5′-CAGGAAGATGGCCAGTCAACT-3′ and reverse primer 5′-GATGGTCTCTGCCATATTGGTT-3′, 497 bp;
(4) ORP150: forward primer 5′-ACACTCCGAGACCTGGAGAA-3′ and reverse primer 5′-CAACACAGGCTTCTCTGTGG-3′, 480 bp [23];
(5)Prdx3: forward primer 5′-CGCACTCTTAGACTTAACT-3′ and reverse primer 5′-CATAATTGGTTCCTTGCCTTCTA-3′, 406 bp;
(6) ATP synthase: forward primer 5′-GGTCGGTGAAGGATCCCAAAA-3′ and reverse primer 5′-CAAGTCCTCAATGGTCATCTGA-3′, 428 bp;
(7) hnRNP C1/C2: forward primer 5′-GCAGGTGTGAAACGATCTGCA-3′ and reverse primer 5′-CCTTGATCAACTCCAGCTGGT-3′, 532 bp, and
(8) GAPDH: forward primer 5′-ACAACTCTCTCAAGATTGTCAGCAA-3′ and reverse primer 5′-ACTTTGTGAAGCTCATTTCCTGG-3′, 518 bp.
Statistical Analysis
The differences observed between LDL subfractions L1 and L5 were compared by analysis of variance using SAS software (version 9.0; SAS Institute), followed by a post hoc Dunnett test. All data are expressed as mean ± SEM, and the differences were considered significant at p < 0.05.
Results
Effect of L1 and L5 on MCP-1 Expression, Cell Viability, and Mitochondrial Dysfunction
The secretion of MCP-1 was used as an indicator of inflammation. The dose effect of L5 and L1 on the inflammatory response of HUVECs is shown in figure 1a (50 µg/ml, p < 0.05; 100 µg/ml, p < 0.01; n = 5). Because 50 µg/ml of L5 sufficiently induced inflammation, this concentration was used in subsequent experiments.
Fig. 1
a Concentration-dependent effect of L5 on MCP-1 levels. b, c Induction of MCP-1 expression examined by Western blotting of cell lysates (b) and ELISA of the supernatants of HUVECs (c). d, e Disruption of mitochondrial membrane integrity by L5 examined by immunofluorescence microscopy (d) and flow cytometry (e). * p < 0.05, ** p < 0.01; n = 5.
Cells treated with L5 showed a 3-fold increase in MCP-1 expression, whereas treatment with L1 (50 µg/ml) did not affect MCP-1 expression (fig. 1b). In parallel, ELISA analysis revealed that treating cells with L5 markedly increased the concentration of MCP-1 detected in cell lysates (>2,000 pg/105 cells; p < 0.05), whereas L1 exhibited no effect (L1, 125 ± 79 pg/105 cells vs. PBS, 157 ± 55 pg/105 cells; fig. 1c).
Cell viability was determined by the MTT assay. The viability of cells treated with L5 was calculated as the relative ratio to the number of cells surviving L1 treatment. In cells treated with L5, viability was 25% less than in cells treated with L1, suggesting that L5 decreased cell survivability compared to L1.
Mitochondrial function was assessed by the color and pattern of JC-1 immunofluorescence staining (fig. 1d). Red (indicative of normal mitochondrial function) and green cytoplasmic fluorescence staining were observed in HUVECs treated with either PBS or L1. However, cells treated with L5 showed a marked reduction in red fluorescence and a reciprocal increase in green fluorescence, suggesting loss of DCm. A similar reduction in red fluorescence was observed by flow cytometry in the presence of L5, but not L1 (fig. 1e).
Protein Expression Profiles of HUVECs Treated with L1 or L5
The protein expression profiles of HUVECs treated with PBS, L1, or L5 were studied by 2-D electrophoresis and mass spectrometry (fig. 2a). The relative amounts of proteins expressed were quantified with the use of ImageMaster software (fig. 2b) and calculated as the relative ratio to L1 (L1 treatment = 1). Compared to treatment with L1, treatment of HUVECs with L5 significantly upregulated hnRNP C1/C2 expression (1.49 ± 0.12 vs. 1.00 ± 0.14; p < 0.05) and downregulated the expression of Grp58, Grp78, Grp94, ATP synthase, and Prdx3 (0.77 ± 0.06 vs. 1.00 ± 0.06, 0.81 ± 0.08 vs. 1.00 ± 0.08, 0.70 ± 0.11 vs. 1.00 ± 0.11, 0.75 ± 0.10 vs. 1.00 ± 0.10, 0.65 ± 0.08 vs. 1.00 ± 0.09, respectively; p < 0.05, n = 4); there was also a nonsignificant reduction in ORP150 expression (0.89 ± 0.39 vs. 1.00 ± 0.39). These results are similar to those previously observed when L5 and LDL derived from normal volunteers were compared [22]. Figure 3a shows an example of MALDI-QTOF performed for spot 4, which identifies a ‘fingerprint’ spectrum of peptides. Figure 3b–f are MS/MS spectra of spot 4, which was identified as Grp58.
Fig. 2
Proteomic analysis of proteins expressed in HUVECs treated with PBS, L1, or L5. a 2-D electrophoresis of proteins in L5-treated cells (left); excerpts showing proteins of interest in cells treated with PBS, L1, or L5 (right). Numbers refer to identified proteins listed in table 1. Results are representative of 4 independent experiments. b Quantitative analysis of the identified proteins in cells treated with PBS, L1, or L5. * p < 0.05; n = 4.
Fig. 3
Example of spot identification by MALDI-QTOF and multiple MALDI-MS/MS sequencing. a Identification of peptide mass fingerprint spectra of protein spot 4 by MALDI-QTOF. Arrows represent sites of MALDI-MS/MS sequencing performed to confirm protein identification. b–f MS/MS spectra of representative amino acid sequences. Spot 4 was identified as Grp58.
Effect of L1 and L5 on Gene and Protein Expression in HUVECs
We used RT-PCR to examine the expression of these 7 genes in HUVECs treated with PBS, L1, or L5 (fig. 4a).Gene expression in cells treated with L5 was calculated as the relative ratio to gene expression in cells treated with L1 (L1 treatment = 1). Treatment of HUVECs with L5 significantly decreased gene expression of Grp58, Grp94, ORP150, Prdx3, and ATP synthase relative to L1 (0.70 ± 0.15 vs. 1.00 ± 0.14, 0.11 ± 0.28 vs. 1.00 ± 0.24, 0.66 ± 0.14 vs. 1.00 ± 0.16, 0.83 ± 0.07 vs. 1.00 ± 0.08, 0.76 ± 0.13 vs. 1.00 ± 0.11, respectively; p < 0.05). In addition, L5 significantly increased gene expression of hnRNP C1/C2 relative to L1 (1.34 ± 0.13 vs. 1.00 ± 0.15; p < 0.05). The decrement in Grp78 expression observed in cells treated with L5 was not statistically significant.
Fig. 4
RT-PCR and Western blot analysis of the indicated mRNA transcripts (a) and proteins (b) (identified in fig. 2) in HUVECs treated with PBS, L1, or L5. mRNA and protein levels are normalized to GAPDH and analyzed relative to L1. * p < 0.05, ** p < 0.01; n = 5. RT-PCR and Western blots are representative of 5 independent experiments.
Immunoblotting of protein expression corroborated the findings from the 2-D analysis, showing a consistent pattern of protein expression (fig. 4b) for cells treated with L1 or L5. The results of immunoblotting revealed that L5 significantly reduced the expression of Grp94, Grp58, ORP150, ATP synthase, and Prdx3 relative to L1 (0.62 ± 0.11 vs. 1.00 ± 0.13, 0.70 ± 0.08 vs. 1.00 ± 0.08, 0.67 ± 0.06 vs. 1.00 ± 0.08, 0.80 ± 0.07 vs. 1.00 ± 0.09, 0.72 ± 0.13 vs. 1.02 ± 0.11, respectively; p < 0.05 or p < 0.01), whereas L5 enhanced the expression of hnRNP C1/C2 relative to L1 (1.67 ± 0.26 vs. 1.07 ± 0.23; p < 0.01). A significant change was not observed for the protein expression of Grp78.
To confirm the ER stress response following L5 treatment in HUVECs, we examined the expression of markers of maladaptive unfolded protein response, p-PERK and IRE-1, by immunoblotting. Relative to L1, L5 significantly (p < 0.05) enhanced the protein expression of the ER stress sensors p-PERK and IRE-1 in HUVECs (fig. 5).
Fig. 5
Western blot analysis of the ER stress sensors p-PERK and IRE-1 in HUVECs treated with PBS, L1, or L5. p-ERK and IRE-1 protein levels are normalized to those of GAPDH and analyzed relative to L1. * p < 0.05; n = 3.
Discussion
Previously, we found that L5 purified from the plasma of hypercholesterolemic patients induced endothelial inflammation and dysfunction, leukocyte recruitment, monocyte differentiation, smooth muscle cell proliferation, and foam cell formation – all processes that contribute to atherogenesis [15,16,21]. We also reported that atheroma-derived LDL induced cellular stress in HUVECs [22].
In the present study, we showed that L5, but not L1, induced severe inflammation and mitochondrial dysfunction in HUVECs and decreased cell viability, possibly initializing the process of atherogenesis. The exposure of ECs to L5 induces oxidative and ER stress similar to that of atheroma-derived LDL, possibly through a mechanism involving RNA destabilization and dysfunction of the ER and the mitochondria in HUVECs.
Mitochondrial Dysfunction Induced by L5
Mitochondrial dysfunction induced by L5 is indicated by JC-1 immunofluorescence. The collapse of the DCm seen in L5-treated cells is an early characteristic feature of oxidative injury and apoptosis. Additionally, protein and mRNA levels of the mitochondrial proteins ATP synthase and Prdx3 were reduced in HUVECs treated with L5. ATP synthase, abundant in the mitochondria, serves as the key enzyme for ATP synthesis. In addition, it plays an important role in protecting fibroblast growth factor 2 (FGF2) from proteolytic cleavage, stabilizing its biologic function in cell proliferation [24,25]. ATP synthase has also been identified on the cell surface of several cell types, including human tumor cells, hepatocytes, keratinocytes, adipocytes, and ECs [26,27]. On hepatocyte membranes, ATP synthase is a high-affinity receptor for apoAI, and it protects against atherogenesis by mediating high-density lipoprotein endocytosis and enhancing the delivery of cholesterol to the liver [27,28]. Prdx3, a member of the antioxidant family of thioredoxin peroxidases in the mitochondria, protects cells from oxidative damage and is essential for maintaining normal mitochondrial function [6,7]. Prdx3 also protects cells from the apoptosis-inducing effects of high levels of H2O2 [6,29]. Moreover, thioredoxins have the ability to improve vascular EC function and prevent the development of atherosclerosis by reducing oxidative stress and increasing NO bioavailability [30]. Given the protective, anti-atherogenic functions of these proteins, the downregulation of ATP synthase and Prdx3 suggests a mechanism by which L5 may increase oxidative stress and mitochondrial dysfunction observed in HUVECs.
RNA Destabilization Induced by L5
The expression of hnRNP C1/C2 was increased in HUVECS treated with L5 when compared to those treated with L1. The hnRNP C proteins, which belong to a large family of RNA motif-binding RNPs, are among the most abundant nuclear proteins involved in mRNA biogenesis, DNA repair [31,32], the cell cycle, apoptosis [33,34], and the maintenance of cellular homeostasis [35]. In normal proximal internal carotid arteries, hnRNP C proteins are expressed predominantly in the endothelium, with significantly lower expression in medial smooth muscle. However, in pre-atherosclerotic intimal hyperplasia, hnRNP C proteins are upregulated in the artery wall [36].
Furthermore, oxidative stress induces hnRNP phosphorylation, which in turn modulates RNA stabilization and stimulates expression of VEGF [10,15]. It has also been shown that angiotensin II enhances the binding of hnRNP to VEGF mRNA and increases the efficiency of VEGF mRNA translation in mice with type II diabetes [37]. In addition, VEGF has been reported to recruit macrophages to atherosclerotic plaque, resulting in the accumulation of inflammatory cells in the plaque [8]. Previously, we showed that L5 increased the expression of VEGF in ECs [21], which may occur as a result of hnRNP C1/C2 upregulation. Therefore, it is possible that inflammation of ECs may result from increased VEGF expression through hnRNP C1/C2 activation.
ER Stress Induced by L5
In addition to oxidative stress, L5, but not L1, induced ER stress by suppressing the expression of Grp58, Grp94, and ORP150. GRPs are responsible for properly folding proteins and degrading misfolded/ubiquitinated proteins and are induced during ER stress to prevent ER stress-induced cell death [13,14,15]. In cells exposed to L5, the reduced expression of GRPs may compromise several anti-atherogenic functions of GRPs. For example, GRPs assist with the correct folding and lipidation of apoB100 during its maturation [38,39], as well as the proteasomal degradation of apoB [40]. GRPs also modulate ER stress-induced insulin resistance [41,42] and other related mechanisms that may directly or indirectly affect the development of atherosclerosis [43,44,45]. Furthermore, GRPs control the quality of newly synthesized LDL receptors by retaining mutant LDL receptors in the ER to be degraded [46] and by binding to scavenger receptor A, which regulates the receptor-mediated uptake of modified LDL into the macrophage [47].
Additionally, the ER chaperone protein ORP150 protects ECs against ER stress-induced apoptosis [48,49]. ER stress induced by the accumulation of unfolded proteins in the ER is considered a survival pathway; however, prolonged or severe ER stress leads to programmed cell death via specific pathways originating from the ER [50]. In human mammary ECs, the expression of ORP150 protects against ER stress-induced apoptosis caused by ox-LDL, and the attenuated expression of ORP150 increases ox-LDL-induced cell death [48]. Our results show that L5 reduced the expression of ORP150, potentially promoting apoptosis through a mechanism that is similar to that of ox-LDL.
L5, but not L1, also enhanced the protein expression of the ER stress sensors p-PERK and IRE-1. The combination of the loss of DCm and the increased levels of ER stress-response sensors observed in L5-treated cells indicates that ER-initiated apoptotic signaling may play a role in the mechanism of L5-induced apoptosis.
In support of these findings, our previous studies have shown that genetically augmenting Akt, which is known to reduce ER stress, renders cells resistant to L5 [18]. In addition, we have shown that L5 induces EC apoptosis by disrupting a FGF2-PI3K-Akt autoregulation mechanism [18] and that homocysteine, an ER stress inducer, also leads to EC death by inhibiting FGF2 transcription [51]. Furthermore, our unpublished data suggest that homocysteine and L5 have synergistic effects when added simultaneously to EC cultures.
Our study highlights the effects of LDL charge in inducing cellular stress. Previously, it has been shown that the total protein concentration in L5 subfractions was 50% higher than that in L1 subfractions [37]. L5 showed distinctive biologic and physicochemical properties from L1. Furthermore, L5 contained greater concentrations of apoE, apoAI, apoCIII, and fragmented apoB100, which may contribute to the electronegativity of L5 [16,37]. Further studies are necessary to determine the composition and unique modifications of L5 that contribute to its negative charge and, potentially, its cytotoxic effects. In addition, studies are warranted that aim to identify the active components of L5, such as phospholipids and derivatives (e.g. platelet-activating factor-like lipids and lysolecithins [15,52]).
In summary, we showed that L5, but not L1, induced stress in HUVECs by causing inflammation and mitochondrial dysfunction (fig. 6). Furthermore, we confirmed the altered expression of proteins induced by L5 but not L1. Based on the known functions of these proteins, we pinpoint potential key players in the activation of cellular stress that leads to inflammation and apoptosis and may contribute to the initiation of atherogenesis.
Fig. 6
A schematic illustrating how L5-induced cell stress may potentiate atherogenesis through the proteins identified in our study.
Acknowledgments
The authors thank Nicole Stancel, PhD, of the Texas Heart Institute at St. Luke’s Episcopal Hospital, for editorial assistance in the preparation of this paper.
This work was funded in part by research grants NSC-94-2314-B-002-001 (Lee), NSC-98-2314-B-002-121-MY3 (Hsu), and NSC-100-2314-B-039-040-MY3 (C.-H. Chen) from the National Science Council of Taiwan; a research grant from the Philip Morris External Research Program (C.-H. Chen); research grant 1-04-RA-13 from the American Diabetes Association (C.-H. Chen); research grant HL-63364 (Yang) from the National Institutes of Health, and research grant DOH101-TD-B-111-004 (C.-H. Chen) from the Taiwan Department of Health Clinical Trial and Research Center of Excellence. D. Tang is supported by training grant HL07812-10 from the National Institutes of Health.

Get Permission