Induction of Plasminogen Activator Inhibitor-1 in Endothelial Cells by Basic Fibroblast Growth Factor and Its Modulation by Fibric Acid
Plasminogen activator inhibitor-1 (PAI-1) inhibits fibrinolysis and proteolysis. Basic fibroblast growth factor (bFGF) stimulates angiogenesis, which requires regional proteolysis. Because modulation of vasculopathy requires tight control of proteolysis, effects of bFGF on PAI-1 expression in endothelial cells (ECs) were characterized. bFGF increased PAI-1 mRNA and accumulation of PAI-1 protein in conditioned media in human umbilical vein ECs. The bFGF-mediated increase in PAI-1 mRNA was attenuated by inhibition of extracellular signal-regulated kinase kinase in human ECV304 cells. The rate of decrease in PAI-1 mRNA after actinomycin D treatment was not affected by bFGF. Transient transfection assays of the human PAI-1 promoter-luciferase construct demonstrated that bFGF-induced PAI-1 transcription was dependent on the elements within the −313 to −260 bp relative to the transcription start site. This region contains an E26 transformation specific 1 (Ets-1)-like site. Electrophoretic mobility shift assay showed that bFGF increased nuclear translocation or DNA binding of the Ets-1-like transcription factor to the PAI-1 promoter. Nucleotide substitution to disrupt the Ets-1-like site reduced bFGF-stimulated promoter activity. Fenofibric acid, an agonist ligand for the peroxisome proliferator-activated receptor-α, inhibited basal and bFGF-stimulated PAI-1 expression. By inducing PAI-1 expression from ECs, bFGF may control proteolysis and fibrinolysis in vessel walls.
Alterations in the function of the fibrinolytic system have been implicated in the pathogenesis of thromboembolic phenomena and vasculopathy.1 Fibrinolytic system activity is regulated at transcriptional, translational, and posttranslational levels and is influenced by specific growth factors.2 Plasminogen activator inhibitor (PAI)-1, the primary physiological inhibitor of plasminogen activators, is a pivotal inhibitor of fibrinolysis2 and of proteolytic processes3 associated with vascular remodeling4 and fibrosis.5
Endothelial cells (ECs) elaborate fibrinolytic system proteins, including tissue-type plasminogen activator (tPA) and urokinase-type plasminogen activator.2,6⇓ Basic fibroblast growth factor (bFGF), a growth factor elaborated by vascular cells, modulates the expression of plasminogen activators in ECs and smooth muscle cells.7– 10⇓⇓⇓ Excessive plasmin generation and proteolysis induced by bFGF-mediated overproduction of plasminogen activators may interfere with cell adhesion or atherosclerotic plaque development. However, the influence, if any, of bFGF on the synthesis of PAI-1 in ECs has not been elucidated. The potential influence may well contribute to physiological or pathophysiological consequences of the elaboration of bFGF. The local synthesis of PAI-1 and the balance between plasminogen activation and inhibition determine the generation of plasmin, which, in turn, influences regional fibrinolytic balance, thus modulating the atherothrombotic11 and angiogenic12 properties of ECs.
We have previously reported that a lipid-lowering fibric acid derivative, which activates peroxisome proliferator-activated receptor-α (PPARα), reduces PAI-1 expression in human hepatoma cells in vitro.13 In the present study, we characterized the influence of bFGF and fibric acid on the elaboration of PAI-1 in vitro by cultured ECs and sought to delineate the potential mechanisms that may underlie increased PAI-1 expression in vivo.
Human umbilical vein ECs (HUVECs) were obtained from Clonetics. A human endothelium-derived cell line, ECV304 cells (ECV cells), were obtained from American Type Culture Collection. Penicillin-streptomycin solution, medium 199, DMEM, trypsin, genistein, and actinomycin D were purchased from Sigma Chemical Co. PD98059 was from Research Biochemical International, and GF109203X was from Tocris. Recombinant human bFGF was from Genzyme, and fenofibric acid was from Kaken. Mouse anti-human PAI-1 IgG was from American Diagnostica. Rabbit anti-human E26 transformation specific 1 (Ets-1)/Ets-2 IgG, which exhibits broad cross-reactivity with members of the Ets family proteins, was from Santa Cruz Biotechnology. All other chemicals were of the highest available commercial grade.
Cell Culture Procedures
HUVECs between the second and third passages were grown to confluence under 5% CO2 at 37°C with EC basal medium (EBM-2, Clonetics). ECV cells were grown to confluence in medium 199 containing 10% calf serum (Hyclone), 100 U/mL penicillin, and 100 μg/mL streptomycin. Confluent ECs were incubated in serum-free DMEM for 24 hours and then incubated with fresh serum-free DMEM containing bFGF. Media were collected and stored at − 80°C. In some experiments, fenofibric acid was prepared as previously described,13 and ECV cells were preincubated with fenofibric acid for 24 hours in serum-free DMEM before exposure to bFGF. When inhibitors were used, ECV cells were pretreated with the inhibitor for 1 hour before the addition of bFGF. Cell viability was determined by trypan blue exclusion and MTT assay (Sigma).
Assays for PAI-1 Activity, PAI-1 Antigen, tPA Antigen, and Total Protein
PAI-1 activity was measured as previously described.14 Human PAI-1 antigen in the media was measured by an ELISA specific for human PAI-1 (Biopool). Human tPA antigen in the media was measured by ELISA (American Diagnostica). Total protein in the conditioned media was assayed with the BCA protein assay (Pierce).
Assays for PAI-1 Antigen by Western Blotting
Equivalent amounts of protein from conditioned medium were diluted 1:1 with sample buffer (0.25 mol/L Tris-HCl [pH 6.8], 40% glycerol, 4% SDS, 20% β-mercaptoethanol, and 0.01% bromophenol blue) and loaded on a 10% SDS-polyacrylamide gel. PAI-1 antigen expression in the conditioned media was assayed by Western blotting as previously described15 with antibodies specific for human PAI-1 and quantified by the method of Huang and Amero.16
Isolation of Total RNA and Northern Blotting
Total RNA was isolated by an acid guanidinium thiocyanate-phenol-chloroform method. Northern blot analysis was performed as described previously.13 Gel-purified polymerase chain reaction (PCR) fragments for PAI-1 (1112 to 1612) and β -actin, radiolabeled with [α-32P]dCTP by random priming, were used as probes. The radioactivity of hybridized bands was detected by autoradiography, which was then subjected to densitometry.
Promoter-Luciferase Vector and Expression Plasmid
The human PAI-1 promoter 5′ flanking region17 from −747 to 15 (762 bp) was amplified by PCR from human genomic DNA isolated from umbilical vein with the forward primer 5′-TCCAACCTCAGCCAGACAAG-3′ (−747 to −727) and the reverse primer 5′-TCTCCTCGTGTCGACACAAA-3′ (−5 to 15), respectively. The PCR product was gel-purified and subcloned into the multiple cloning sites of the pT7Blue T-Vector (Novagen). The T-vector was digested with EcoRI and HindIII and subcloned into the multiple cloning sites of the promoterless Renilla luciferase reporter gene vector pRL-null (Promega). Basal expression of luciferase activity of the PAI-1 promoter vector was detected by the PAI Full (−747/15) luciferase vector. The 5′ deletion mutants were generated by PCR with the 5′ primers complementary to the PAI-1 gene sequence. Deletion mutants of PAI-1 promoter vectors were constructed as follows: PAI 1F (−553/15) luciferase, PAI 2F (−313/15) luciferase, PAI 3F (− 260/15) luciferase, and PAI 4F (−205/15) luciferase. Point mutation of PAI-1 promoter vectors was also constructed. Basal expression of luciferase activity of PAI-1 promoter vector was detected by the PAI Full luciferase vector, which has a normal Ets-1-like site (5′-GGACATCCGGGAG-3′). The mutation vector has 2 point mutations in the Ets-1-like site as underlined (5′-GTTCATCCGGGAG-3′).
DNA Transfection and Luciferase Assay
ECV cells were cultured to ≈80% confluence. The cells were cotransfected with each PAI-1 promoter Renilla luciferase fusion DNA reporter construct (20 μg of pRL vector) and a Firefly luciferase pGL vector to control for transfection efficiency (20 μg) introduced by electroporation (240 V, 960 microfarad). These cells were cultured in DMEM supplemented with 10% calf serum for 20 hours, stimulated with bFGF (10 ng/mL) in DMEM containing 10% calf serum for 24 hours, and harvested. Cell lysate luciferase activity was measured with the use of a Dual-Luciferase Reporter Assay System (Promega) and luminometer (Turner Design). Normalized luciferase activity was calculated as the ratio of luciferase activity to control vector activity. Results for each reporter construct were expressed as percent induction compared with results in transfected unstimulated cells.
Electrophoretic Mobility Shift Assay
For electrophoretic mobility shift assay (EMSA), ECV cells were stimulated for 2 hours with bFGF before nuclear extracts were prepared. The oligonucleotide (ACGCTAGGACATCCGGGAGCATG) spanning the Ets-1-like site (as underlined above) in the human PAI-1 promoter was end-labeled with [α-32P]dCTP by the Klenow fragment of DNA polymerase I and purified. Nuclear extracts (0.3 μg) were incubated with the labeled oligonucleotide. DNA-protein complexes were separated from free probe by 5% nondenaturing polyacrylamide gel, and autoradiography was performed. Specificity was determined by the addition of an excess of unlabeled (cold) oligonucleotide to the nuclear extracts before formation of DNA-protein complexes.
Data are mean±SEM. Differences were assessed by ANOVA with the Bonferroni least significant post hoc tests for comparisons within multiple groups. Significance was defined as P< 0.05.
Effects of bFGF on PAI-1 Expression in ECs
A 2.1-fold increase in PAI-1 mRNA expression induced by bFGF was seen with HUVECs, as assessed by Northern blotting (see online Figure IA, which can be accessed at http://atvb.ahajournals.org). The baseline concentration of PAI-1 in the conditioned media of HUVECs was 181.7±32.7 ng/mL (n=4, ELISA). bFGF increased PAI-1 protein accumulation in the media in a concentration-dependent fashion (online Figure IB). Peak effects were seen with 10 ng/mL (1.8±0.3-fold over control). Baseline activity of PAI-1 in the conditioned media was 12.0±0.6 arbitrary units per milliliter at 4 hour (n=4). bFGF increased PAI-1 activity at a concentration of 0.1 ng/mL (1.7±0.4-fold over control) and 1 ng/mL (2.3±1.1-fold over control). At 10 and 100 ng/mL, PAI-1 activity somewhat diminished (1.4±0.4-fold over control and 1.5±0.8-fold over control, respectively). The baseline concentration of tPA in the conditioned media of HUVECs was 22.1±2.2 ng/mL (n=4, ELISA). bFGF increased tPA protein accumulation in the media (24.9±2.7 at 1 ng/mL and 24.0±2.8 at 10 ng/mL, P<0.05).
An increase in PAI-1 mRNA expression was seen in ECV cells as assessed by Northern blotting (see online Figure IC). The expression of PAI-1 mRNA by bFGF was increased in a concentration-dependent fashion, with a 1.4-fold increase over control at 10 ng/mL and a 2.0-fold increase at 100 ng/mL. The baseline concentration of PAI-1 in the conditioned media of ECV cells was 254±28 ng/mL (n=5, ELISA). bFGF increased PAI-1 protein accumulation in the media in a concentration-dependent fashion (see online Figure ID). Peak effects were seen with 10 ng/mL bFGF (1.9±0.2-fold over control). The response was diminished somewhat with concentrations of 100 ng/mL (1.4±0.2-fold over control). The baseline concentration of tPA in the conditioned media of ECV cells was 0.71±0.05 ng/mL (n=4, ELISA). bFGF increased tPA protein accumulation in the media (0.88±0.05 at 1 ng/mL and 0.93±0.02 at 10 ng/mL, P<0.01). Total protein content in the conditioned media was not altered by bFGF in HUVECs and EVC cells (results not shown).
Effects of PD 98059 and Actinomycin D on PAI-1 mRNA
To investigate the intracellular mechanisms involved in the induction of PAI-1 mRNA, ECV cells were treated for 1 hour with each of the various inhibitors of intracellular signaling pathway before the addition of bFGF (Figure 1A). Inhibition of extracellular signal-regulated kinase (ERK) kinase with PD98059 blocked basal and bFGF-stimulated PAI-1 mRNA levels. In contrast, GF109203X, an inhibitor of the protein kinase C pathway, and genistein, an inhibitor of tyrosine kinase, had no significant effect. β-Actin mRNA levels did not change in any experimental conditions. To determine whether bFGF influences PAI-1 mRNA half-life, ECV cells were stimulated with bFGF for 2 hours, medium was changed, actinomycin D (5 μg/mL) was added with or without bFGF, and cells were incubated for up to 6 hours (Figure 1B). The rate of PAI-1 mRNA decrease was not changed by bFGF, suggesting that the PAI-1 mRNA increase by bFGF was due to an increase of transcription from the PAI-1 promoter.
Determination of DNA Regions Critical for Basal and bFGF-Inducible PAI-1 Transcriptional Activity
To identify the 5′ flanking region of the PAI-1 gene responsible for the effects of bFGF, transient transfections with several PAI-1 promoter-luciferase reporter constructs were performed in ECV cells (see online Figure IIA, which can be accessed at http://atvb.ahajournals.org). Higher basal promoter activity was observed in 1F and 2F, suggesting that the region between −747 and −553 bp may contain a repressor element (see online Figure IIB). bFGF increased promoter-driven luciferase activity by 67±16% (see online Figure IIC, n=18). Relative to the largest promoter fragment tested, bFGF effect was reduced with deletion of the region at −747 to −553 bp and −553 to −313 bp. Deletion of the region at −313 to −260 bp completely abolished the bFGF effect. Deletion of the region at −260 to −205 bp resulted in no further reduction. These data indicate that the major sequence determinant of responsiveness resides between −313 and −260 bp. To determine whether a protein that can bind to the Ets-1-like site was responsible for the effect of bFGF, EMSA was performed by using oligonucleotides corresponding to the PAI-1 Ets-1-like site. The oligonucleotide bound, besides a constitutive protein complex, a bFGF-inducible nuclear protein complex, suggesting that bFGF induced nuclear translocation and DNA binding of the Ets-1-like transcription factor to the PAI-1 promoter (Figure 2A). The addition of a 10- to 100-fold excess of cold probe inhibited DNA-protein complex formation in a dose-dependent manner. The addition of the Ets-1/Ets-2 antibody inhibited DNA-protein complex formation. To further characterize the responsible region, a mutant construct containing a 2-nucleotide substitution in the Ets-1-like site was generated. Compared with the bFGF-induced increased promoter activity in the wild-type, substitution in the Ets-1-like site reduced bFGF stimulation (Figure 2B). These results suggest that the Ets-1-like site is critical for the effect of bFGF.
Effects of Fenofibric Acid
Fenofibric acid diminished PAI-1 mRNA expression in ECV cells in a dose-dependent manner, as assessed by Northern blotting (Figure 3A). Fenofibric acid markedly diminished but did not totally abolish the accumulation of PAI-1 protein elaboration from ECV cells exposed to bFGF (Figure 3B). At a concentration of 100 μmol/L, it suppressed baseline PAI-1 accumulation by 75±17% (n=3) and suppressed the increase induced by bFGF by 65±15% (n=3). When fenofibric acid was present in the media of ECV cells transfected with the PAI-1 promoter-luciferase reporter construct, basal PAI-1 promoter activity and bFGF-stimulated PAI-1 promoter activity were diminished (n=4, Figure 3C). Thus, fenofibric acid inhibited PAI-1 expression at least partly at the level of transcription. Fenofibric acid did not inhibit bFGF-induced nuclear translocation or DNA binding of Ets-1-like transcription factor to the PAI-1 promoter (Figure 2A). Fenofibric acid decreased the baseline concentration of tPA in the media by 84±08% and suppressed the increased induced by bFGF by 82±10% (n=3). Cell viability and total protein in the conditioned media were unaffected by fenofibric acid at the concentrations used (results not shown).
In the present study, bFGF was shown to influence the production of PAI-1 in ECs. A paradoxical decrease of PAI-1 activity with high concentrations of bFGF in the media of HUVECs may reflect coinduction of plasminogen activators by bFGF.7,8⇓ Increased PAI-1 protein correlated with increased mRNA and activity of the PAI-1 promoter. The induction of PAI-1 was inhibited by the ERK kinase inhibitor, suggesting that an ERK kinase-dependent pathway is involved in the regulation of endothelial PAI-1 gene expression by bFGF. The results in the present study extend our previous observation that the ERK kinase pathway plays a critical role in mediating the response of PAI-1 to angiotensin II in smooth muscle cells18; furthermore, these results offer a novel insight into the mechanism of PAI-1 gene response in 2 distinct vascular cell components. The concentrations of the inhibitors used were relevant in ECV cells.19– 21⇓⇓ Search for the putative regulatory sequence in the promoter revealed a sterol regulatory element binding protein-like site (−569 to −558 bp), an activator protein (AP)-1-like site (−334 to −324 bp), an Ets-1-like site (−296 to −284 bp), and a CCAAT/enhancer-binding protein-like site (−224 to −210 bp). Deletion analysis of the PAI-1 promoter suggested that a region between −313 and −260 bp was primarily mediating the bFGF response, showing that the Ets-1-like site may be involved in the bFGF-inducible PAI-1 promoter activity. EMSA and mutational analysis of the Ets-1-like site further suggested that this Ets-1-like sequence is responsible for the bFGF-induced PAI-1 expression. Although an antibody specific for Ets-1 did not inhibit DNA-protein complex formation (results not shown), the addition of Ets-1/Ets-2 antibody raised against the carboxy-terminal domain (highly conserved among the Ets gene family and essential for DNA binding) inhibited DNA-protein complex formation. These results suggest that other members of the Ets gene family may be at least partly responsible for mediating bFGF effects.
PAI-1 inhibits fibrinolysis and proteolysis,2,3⇓ which are critical in vascular remodeling,4 fibrosis,5 and atherosclerosis.11 Control of proteolysis is important for angiogenesis. PAI-1 may diversely modulate angiogenesis. Low PAI-1 levels may result in excessive proteolysis, failure of cell adhesion, prevention of coordinated EC sprouting,22,23⇓ and inhibition of angiogenesis. Tumor cells fail to invade in PAI-1 knockout mice because of a lack of vascularization.24 PAI-1 promoted angiogenesis by the inhibition of proteolysis,25 suggesting that an inhibition of excessive proteolysis by PAI-1 may contribute to neovascularization. Alternatively, inhibition of proteolysis can also inhibit angiogenesis because of the inhibition of fibrinolysis in the provisional matrix,26 which might affect EC migration. Ets-1 modulates the angiogenic properties of ECs,27 and growth factors that promote angiogenesis may also contribute to atherosclerotic plaque development.28 Thus, proteolysis mediated by plasmin requires tight control during angiogenic processes, and induction of PAI-1 by bFGF as shown in the present study may contribute to this fine control of neovascularization.
Fenofibric acid diminished the increased accumulation of PAI-1 induced by the exposure of cells to bFGF. Because basal accumulation of PAI-1 was affected also by fenofibric acid, constitutive synthesis was modified as well. Fenofibric acid binds and activates PPARα, and activation of PPARα negatively influences the AP-1-dependent pathway.29 In the present study, deletion of the promoter region containing an AP-1-like site inhibited the effect of bFGF, and EMSA suggested that fenofibric acid did not inhibit nuclear translocation or DNA binding of the Ets-1-like transcription factor to the PAI-1 promoter. Thus, the inhibition of PAI-1 synthesis by fenofibric acid may involve the AP-1 activation pathway.
The potential consequences of fenofibric acid action are of particular clinical interest. Use of fibric acid derivatives decreases the incidence of cardiac events in patients with coronary artery disease and low LDL levels although only a modest increase in HDL is induced.30 Although increased local fibrinolysis may induce degradation of the extracellular matrix and destabilization of advanced plaques, reduced PAI-1 expression by ECs may shift the local balance to increased fibrinolytic capacity in blood that could limit the extent of acute thrombosis after plaque rupture. A shift in the local fibrinolytic balance toward increased fibrinolysis may decrease thrombotic risk, thereby contributing to the clinical benefit seen with fibrates independent of their salutary effects on lipids.
This work was supported in part by grants-in-aid for scientific research from the Ministry of Education, Science, Sport, and Culture of Japan and from the Ministry of Welfare and Labor and by a grant-in-aid from the Mitsui Life Social Welfare Foundation (Tokyo). The authors thank Kaori Abumiya, Mizuho Kasai, and Akiko Aita for technical assistance and Lori Dales for secretarial support.
Received September 25, 2001; revision accepted February 17, 2002.
- ↵Collen D, Lijnen HR. Basic and clinical aspects of fibrinolysis and thrombolysis. Blood. 1991; 78: 3114–3124.
- ↵Drew AF, Tucker HL, Kombrinck KW, Simon DI, Bugge TH, Degen JL. Plasminogen is a critical determinant of vascular remodeling in mice. Circ Res. 2000; 87: 133–139.
- ↵Pepper MS, Belin D, Montesano R, Orci L, Vassalli JD. Transforming growth factor-beta 1 modulates basic fibroblast growth factor-induced proteolytic and angiogenic properties of endothelial cells in vitro. J Cell Biol. 1990; 111: 743–755.
- ↵Yamamoto C, Kaji T, Furuya M, Sakamoto M, Kozuka H, Koizumi F. Basic fibroblast growth factor suppresses tissue plasminogen activator release from cultured human umbilical vein endothelial cells but enhances that from cultured human aortic endothelial cells. Thromb Res. 1994; 73: 255–263.
- ↵Bochaton-Piallat ML, Gabbiani G, Pepper MS. Plasminogen activator expression in rat arterial smooth muscle cells depends on their phenotype and is modulated by cytokines. Circ Res. 1998; 82: 1086–1093.
- ↵Wang W, Chen HJ, Schwartz A, Cannon PJ, Rabbani LE. T cell lymphokines modulate bFGF-induced smooth muscle cell fibrinolysis and migration. Am J Physiol. 1997; 272: C392–C398.
- ↵Eitzman DT, Westrick RJ, Xu Z, Tyson J, Ginsburg D. Plasminogen activator inhibitor-1 deficiency protects against atherosclerosis progression in the mouse carotid artery. Blood. 2000; 96: 4212–4215.
- ↵Mandriota SJ, Pepper MS. Vascular endothelial growth factor-induced in vitro angiogenesis and plasminogen activator expression are dependent on endogenous basic fibroblast growth factor. J Cell Sci. 1997; 110: 2293–2302.
- ↵Nordt TK, Kornas K, Peter K, Fujii S, Sobel BE, Kubler W, Bode C. Attenuation by gemfibrozil of expression of plasminogen activator inhibitor type 1 induced by insulin and its precursors. Circulation. 1997; 95: 677–683.
- ↵Schneider DJ, Sobel BE. Augmentation of synthesis of plasminogen activator inhibitor type 1 by insulin and insulin-like growth factor type I: implications for vascular disease in hyperinsulinemic states. Proc Natl Acad Sci U S A. 1991; 88: 9959–9963.
- ↵Sawa H, Sobel BE, Fujii S. Potentiation by hypercholesterolemia of the induction of aortic intramural synthesis of plasminogen activator inhibitor type 1 by endothelial injury. Circ Res. 1993; 73: 671–680.
- ↵Takeda K, Ichiki T, Tokunou T, Iino N, Fujii S, Kitabatake A, Shimokawa H, Takeshita A. Critical role of Rho kinase and MEK/ERK pathways for angiotensin II-induced plasminogen activator inhibitor-1 gene expression. Arterioscler Thromb Vasc Biol. 2001; 21: 868–873.
- ↵Liu J, Gao B, Mirshahi F, Sanyal AJ, Khanolkar AD, Makriyannis A, Kunos G. Functional CB1 cannabinoid receptors in human vascular endothelial cells. Biochem J. 2000; 346: 835–840.
- ↵Bajou K, Masson V, Gerard RD, Schmitt PM, Albert V, Praus M, Lund LR, Frandsen TL, Brunner N, Dano K, Fusenig NE, Weidle U, Carmeliet G, Loskutoff D, Collen D, Carmeliet P, Foidart JM, Noel A. The plasminogen activator inhibitor PAI-1 controls in vivo tumor vascularization by interaction with proteases, not vitronectin. implications for antiangiogenic strategies. J Cell Biol. 2001; 152: 777–784.
- ↵Stefansson S, Petitclerc E, Wong MK, McMahon GA, Brooks PC, Lawrence DA. Inhibition of angiogenesis in vivo by plasminogen activator inhibitor-1. J Biol Chem. 2001; 276: 8135–8141.
- ↵Delerive P, Martin-Nizard F, Chinetti G, Trottein F, Fruchart JC, Najib J, Duriez P, Staels B. Peroxisome proliferator-activated receptor activators inhibit thrombin-induced endothelin-1 production in human vascular endothelial cells by inhibiting the activator protein-1 signaling pathway. Circ Res. 1999; 85: 394–402.
- ↵Rubins HB, Robins SJ, Collins D, Fye CL, Anderson JW, Elam MB, Faas FH, Linares E, Schaefer EJ, Schectman G, Wilt TJ, Wittes J. Gemfibrozil for the secondary prevention of coronary heart disease in men with low levels of high-density lipoprotein cholesterol: Veterans Affairs High-Density Lipoprotein Cholesterol Intervention Trial Study Group. N Engl J Med. 1999; 341: 410–418.