E2F-1 Regulates Nuclear Factor-κB Activity and Cell Adhesion
Potential Antiinflammatory Activity of the Transcription Factor E2F-1
Background— The transcription factor E2F-1 promotes vascular smooth muscle cell apoptosis and is reported to inhibit apoptosis induced by tumor necrosis factor (TNF)-α in endothelial cells. Whether E2F-1 overexpression exerts potentially antiinflammatory effects in endothelial cells is not known.
Methods and Results— By immunoblotting and immunofluorescence, TNF-α treatment of human aortic endothelial cells (HAECs) with the control vector Ad.null was followed by rapid nuclear translocation of nuclear factor (NF)-κB p65, whereas nuclear translocation of p65 was markedly reduced in HAECs overexpressing E2F-1. Electrophoretic mobility shift assay and gel shift analysis of nuclear cell extracts confirmed that HAECs treated with a recombinant adenovirus encoding E2F-1 failed to associate with the binding domain of p65. Stimulation of the Ad.null-infected endothelial cells with TNF-α resulted in enhanced expression of endothelial intracellular adhesion molecule-1, vascular cellular adhesion molecule-1, and E-selectin and enhanced adhesion of monocytic U937 cells to the HAECs. Adhesion molecule expression and cell adhesion were reduced in E2F-1–transduced HAECs, associated with a marked decrease in phosphorylated IκB-α, required for nuclear translocation of NF-κB p65.
Conclusions— These findings suggest that E2F-1 stabilizes IκB and thereby may inhibit NF-κB–dependent processes involved in atherogenesis, including endothelial expression of E-selectin, vascular cellular adhesion molecule-1, and intracellular adhesion molecule-1 and cell adhesion to perturbed endothelial cells.
Received July 3, 2002; revision received September 4, 2002; accepted September 9, 2002.
The ubiquitous transcription factor nuclear factor (NF)-κB plays a pivotal role in the signaling cascade of proinflammatory events.1 Cytokines, including interleukin-1 and tumor necrosis factor-α (TNF-α), converge on a common pathway that leads to the release of the active NF-κB subunit p65 (RelA) from its cytoplasmic association with the regulatory p50 subunit (p105/p50, NF-κB1) and the NF-κB inhibitor IκB.2 In the stable trimeric complex of p65, p105/50, and IκB, IκB prevents nuclear translocation and DNA binding of p65 by masking its nuclear localization signal. In response to proinflammatory stimuli, IκB is phosphorylated and thereby targeted to poly-ubiquitination and rapid degradation by the multicatalytic 26S proteasome protease complex.3 Phosphorylation of IκB occurs through the activity of a family of IκB kinases, which are the upstream targets of TNF-α, interleukin-1, and many other cytokines.4
See p 2640
Among the E2F family of transcription factors, E2F-1 plays an ambiguous role in functioning in tissue-specific manner as oncogene at times and tumor suppressor at others.5–7⇓⇓ During the G0 and G1 cell-cycle phases, this S-phase transcription factor is kept in check by the retinoblastoma gene product (Rb) and becomes active only after its release from Rb in late G1.8,9⇓ In contrast with other members of the E2F family, E2F-1 can function to initiate both S-phase and apoptosis.7 Mice with inactivated E2F-1 gene have shown that the loss of proapoptotic E2F-1 functions is followed by T-cell hyperplasia and increased tumorigenesis.10,11⇓ Conversely, overexpression of E2F-1 in coronary vascular smooth muscle cells (VSMCs) induces apoptosis, which seems to be mediated by E2F-1–induced caspase-3 activity, is highly effective in suppressing VSMC growth, and may play a role in suppressing VSMC proliferation at the site of arterial balloon injury12 (also Chen Z, unpublished data, 1999)
Whereas VSMC apoptosis may, in principle, explain suppression of VSMC hyperplasia and migration at injured vessel sites, we hypothesized that E2F-1 may harbor additional activities of potential benefit for the vessel wall. Leukocyte adhesion to the activated endothelium is a key process in the initiation of atherosclerosis, and its suppression could result in the delay of atherogenesis.13,14⇓ We hypothesized that E2F-1 can inhibit adhesion molecule expression in cytokine-activated endothelial cells and pursued this hypothesis in cultured human aortic endothelial cells (HAECs). We demonstrate that E2F-1 inhibits in vitro vascular cellular adhesion molecule-1 (VCAM-1), intracellular adhesion molecule-1 (ICAM-1), and E-selectin expression and adhesion of monocytic cells to TNF-α–stimulated HAECs.
Materials and Antibodies
Recombinant human tumor necrosis factor (TNF-α), human soluble (s)E-selectin, and sVCAM-1 immunoassay kits were from R&D Systems. Antibodies against NF-κB p65, IκB, E2F-1 (KH95), sICAM-1, actin, and phospho-IκB were from Santa Cruz Biotechnology and were monoclonal except for the last 2 antibodies mentioned. Cy 3-conjugated AffiniPure Donkey anti-mouse lgG and Cy 2-conjugated AffiniPure Donkey anti-rabbit lgG were from Jackson ImmunoResearch Laboratories.
Primary human aortic endothelial cells (HAECs) were obtained from Cascade Biologics. HAECs (passage 2 to 4) were cultured in medium-200 enriched with growth supplement containing (in final concentrations) 2% FBS, hydrocortisone 1 mg/mL, human epidermal growth factor 10 ng/mL, basic fibroblast growth factor 3 ng/mL, and heparin 10 mg/mL. Cells from the human myelomonocytic leukemia cell line U93715 were from American Type Culture Collection and were propagated in RPMI-1640/10% FCS.
Preparation of Recombinant Adenoviral Vectors
Construction of the recombinant adenovirus encoding human E2F-1 has been described.16 In brief, the E2F-1 cDNA, cloned from ML-1 cells, was inserted into the shuttle plasmid pXCJL-1 containing the human cytomegalovirus promoter and the bovine growth hormone polyadenylation signal. Recombinants were generated by cotransfection into 293 cells of pXCJL-1-E2F-1 and pJM17, and viral stocks of Ad.E2F-1 and the empty control virus, Ad-null, were propagated in 293 cells and purified as previously described.12 The concentration of purified infectious viral particles was determined by plaque assay in 293 cells and expressed as plaque-forming units. Viral preparations were stored at −80°C and used if endotoxin-free (<0.125 EU/mL). Infections were performed in serum-free media for 6 hours at 37°C, after which serum-containing media was added and the cells were grown in plastic tissue culture dishes.
Preparation of Whole Cell Extracts and Immunoblotting
Cells were collected at 200g for 5 minutes and washed with PBS. The cell pellet was resuspended in lysis buffer (62.5 mmol/L Tris HCl, pH 6.8, 10% [vol/vol] glycerin, 2% [wt/vol] sodium dodecyl sulfate SDS], 1 mmol/L phenylmethylsulfonyl fluoride [PMSF], 1 μg/mL pepstatin A, and 1 μg/mL leupeptin and 5 μg/mL aprotinin). Cell homogenates were centrifuged at 15 000g and 4°C for 15 minutes. The protein content was determined with a Micro-BCA Protein assay (Pierce). An equal amount of protein (50 μg) was loaded onto SDS-polyacrylamide gels (10% to 20% gradient gels). Proteins were separated at 120 V for 1.5 hours and then blotted to nitrocellulose membranes (0.2 μm; Schleicher & Schuell) in transfer buffer (25 mmol/L Tris, 192 mmol/L glycine, 20% methanol [vol/vol]) at 18 V overnight at 4°C. The blots were blocked with 5% nonfat dry milk in PBST (PBS and 0.1% Tween-20) at room temperature for 1 hour. Membranes were incubated with antibodies recognizing p65 (1:1000), E2F1 (KH95 (1:1000), actin (1:1000), IκB-α,1:1000 phospho-IκB-α (1:1000), and ICAM-1.1:1000. Membranes were washed with PBST 6 times for 30 minutes and incubated with anti-mouse or anti-rabbit peroxidase-conjugated secondary antibodies for 1 hour. Blots were washed and developed using the ECL chemiluminescence detection reagent (Amersham Pharmacia). Membranes were stripped in standard buffer (2% SDS, 62.5 mmol/L Tris-HCl, 100 mmol/L 2-mercaptoethanol, pH 6.8) at 60°C for 20 minutes, washed twice in PBST for 30 minutes, and reprobed. Membranes were developed for 3 minutes in Supersignal West Pico substrate (Pierce).
Double Immunofluorescent Staining
HAECs grown on cover slips (12 mm diameter) were fixed with 4% paraformaldehyde. After 3 washes with PBS, the cells were treated with 100% methanol for 10 minutes at −20°C and permeabilized with 0.1% Triton X-100 for 10 minutes at room temperature, then washed 3 times with PBS. Nonspecific antibody binding was blocked for 10 minutes with PBS containing 10% FCS. The cells were incubated for 1 hour with the primary antibodies (mouse monoclonal anti–NF-κB p65 antibody and the rabbit polyclonal anti-E2F1 KH95 antibody). After 4 washing steps with PBS, the cells were incubated for 1 hour with Cy 3–conjugated AffiniPure donkey anti-mouse lgG (1:500) and Cy 2–conjugated AffiniPure donkey anti-rabbit lgG (1:300) and for the last 5 minutes with 1 μg/mL Hoechst 33258. The coverslips were then washed in PBS and deionized water and mounted on glass slides with Evathanol (Sigma). Cells were examined under a fluorescent microscope.
Nuclear Protein Extraction and Gel Shift Assay
Confluent HAECs were either left untreated or infected with Ad.null (MOI 100) or Ad.E2F1 (MOI 100) for 48 hours and then exposed to either medium alone or TNF-α (100 ng/mL) in the medium for an additional 5 hours. Nuclear protein extracts were prepared as described by Schreiber et al.17 An NF-κB oligonucleotide containing the NF-κB consensus sequence 5′-AGTTGAGGGGACTTTCCCAGGC-3′ was labeled with [γ-32P]ATP using T4 polynucleotide kinase for 20 minutes at room temperature in the presence of 50 μg poly(dI-dC) and 10 mmol/L Tris-HCl buffer, pH 7.5, containing 50 mmol/L NaCl, 0.5 mmol/L EDTA, 0.5 mmol/L DTT, 4% (wt/vol) glycerol, and 1 mmol/L MgCl2. The nuclear extracts (10 μg protein) were incubated with the radiolabeled NF-κB oligonucleotide and subjected to electrophoresis through 5% (wt/vol) polyacrylamide gels, which were subsequently dried, autoradiographed, and analyzed with the Fujix Bioimage Analyzer BAS2000 (Fuji Photo Film).
VCAM-1 and E-Selectin ELISA
ELISAs for E-selectin and VCAM-1 were performed according to the manufacturer’s instruction (R&D Systems) with lysates of the HAECs that had been treated with TNF-α (100 ng/mL) for 16 hours.
Monocyte Adhesion Assays
Confluent HAEC monolayers were grown on coverslips in 6-well tissue culture plates after infection with Ad.E2F1 and Ad.null (both at MOI 100). TNF-α (100 ng/mL) was then added to induce expression adhesion molecules. U937 cells15 were labeled with Claim-AM (Molecular Probe). After the HAECs were stimulated and washed, 2×105 labeled U937 cells per well were added to the HAECs and allowed to interact for 60 minutes at 37°C. After 3 washes with complete medium, cultures were fixed with 4% paraformaldehyde, and the attached monocytes were counted on an inverted microscope.
E2F-1 Overexpression Reduces Nuclear Translocation of NF-κB p65
After infection of the cells with Ad.E2F-1, immunoblots of nuclear extracts of HAECs revealed a striking reduction of NF-κB p65 (Figure 1). In addition, a slightly lesser p65 signal in Ad.E2F-1 compared with Ad.null-treated cells was observed, despite equal loading of lanes, as illustrated by similar actin signals. However, immunoblotting offered little information on the relationship between E2F-1 overexpression and failure of p65 to translocate at the individual cell level. Therefore, and in order to strengthen the relationship between E2F-1 overexpression and lack of nuclear translocation of p65, we performed immunofluorescence analysis of HAECs treated with recombinant Ad.null and Ad.E2F-1 (MOI 100) before exposure to TNF-α (100 ng/mL). As was apparent from this analysis, cells strongly overexpressing E2F-1 showed reduced, although not completely inhibited, translocation of p65. In contrast, nuclear translocation of p65 was readily apparent in cells with only weak expression of E2F-1 (Figure 2).
E2F-1 Reduces the Expression of ICAM-1, VCAM-1, and E-selectin in TNF-α–Stimulated HAECs
VCAM-1, ICAM-1, and E-selectin play a significant role in leukocyte adhesion to cytokine- or cholesterol-activated endothelial cells.13,14,18⇓⇓ To examine whether E2F-1 can influence the cytokine-induced expression of these adhesion molecules, we stimulated HAECs, pretreated with Ad.E2F-1 or Ad.null, with TNF (100 ng/mL). TNF-α induced robust expression of VCAM-1, E-selectin, and ICAM-1, as visualized, respectively, by ELISA and immunoblots (Figure 3). Compared with treatment with culture medium alone (mock), Ad.null (MOI 100) did not exert a detectable influence on the expression of the adhesion molecules, whereas pretreatment of the HAECs with Ad.E2F-1 (MOI 100) markedly inhibited TNF-α–induced expression of VCAM-1, ICAM-1, and E-selectin (Figure 3).
E2F-1 Inhibits Binding of U937 Cells to TNF-α–Stimulated HAECs
To determine whether the reduction in surface expression of adhesion molecules was associated with a functional consequence for the HAECs, we studied the effect of E2F-1 on the binding of U937 cells to TNF-α–stimulated HAECs. In the absence of TNF-α, confluent HAECs showed minimal binding to U937 cells (data not shown), but adhesion was substantially increased when the HAECs were treated with TNF-α with or without pretreatment of the cells with Ad.null. In contrast, pretreatment of the HAECs with Ad.E2F-1 markedly reduced the adhesion of U937 cells to the TNF-α–stimulated cells (Figures 4 and 5⇓)
E2F-1 Inhibits Cytokine-Induced Activation of NF-κB
Transcriptional regulation involving activation of NF-κB has been implicated in the cytokine-induced expression of VCAM-1, ICAM-1, and E selectin.19–21⇓⇓ To examine whether E2F-1 inhibits NF-κB activation, we performed gel shift assays with the use of a 32P-labeled oligonucleotide bearing a NF-κB consensus sequence. HAECs were uninfected (mock) or preinfected with Ad.null or Ad.E2F-1 and subsequently stimulated with TNF-α at 37°C. Gel shift assays showed that treatment of the mock and Ad.null-infected cells with TNF-α resulted in the appearance of shifted bands. In contrast, preinfection of the cells with Ad.E2F-1 markedly reduced the densities of the NF-κB shifted bands of lysates from cells stimulated with TNF-α (Figure 6).
E2F-1 Interferes With the NF-κB Signaling Pathway by Stabilizing Phosphorylated IκB-α
In endothelial cells, the TNF-α–induced expression of E-selectin, ICAM-1, and VCAM-1 is initiated by interruption of the inactive p50/p65/IκB heterotrimer. Because phosphorylation of IκB-α serine residues targets this NF-κB inhibitor to rapid proteolytic degradation, we examined whether E2F-1 influences cellular levels of IκB. Immunoblots of cell lysates showed that, whereas Ad.null-treated HAECs had markedly reduced IκB-α levels after NF-κB activation with TNF-α, IκB-α levels were unchanged in E2F-1–transduced cells exposed to TNF-α. Furthermore, immunoblots detecting phosphorylated IκB-α demonstrated that the treatment of the HAECs with Ad-E2F-1 markedly reduced the cellular level of phosphorylated IκB-α, suggesting that E2F-1 inhibits the degradation of IκB-α by inhibiting its phosphorylation (Figure 7).
We recently reported that the transcription factor E2F-1 induces vascular smooth muscle cell apoptosis12 in vitro and observed in a preliminary report that E2-1 overexpression suppressed neointima formation in balloon-injured arteries (Chen Z, unpublished data, 1999). Other reports found that this transcription factor can prevent TNF-α–induced endothelial cell apoptosis in vitro,22 closely mimicking the effects of a soluble TNF-α receptor, a specific TNF-α antagonist.23 To additionally study these divergent responses to E2F-1, we investigated the ability of E2F-1 to influence NF-κB activity in primary HAECs and found that the tumor suppressor E2F-1 inhibits expression of the cell adhesion molecules VCAM-1, ICAM-1, and E-selectin in the HAECs pretreated with sublethal concentrations of TNF-α and that this inhibition correlates with suppression of NF-κB activity and hypophosphorylation of IκB-α, the levels of which are stabilized in TNF-α–exposed endothelial cells overexpressing E2F-1. In addition, we demonstrated that diminished expression of the adhesion molecules in the E2F-1–transduced endothelial cells markedly reduces adhesion of a monocytic cell line (U937) to the transduced endothelial cells. Thus, E2F-1 harbors an unexpected ability to regulate NF-κB–dependent cell adhesion to cytokine-stimulated cells, and this ability may help elucidate additionally its potential vasoprotective effect in vivo.
We stimulated endothelial cells with TNF-α for several reasons. First, TNF-α has been implicated as an important cytokine, triggering endothelial apoptosis in vitro and in vivo.23 TNF-α, which originally was reported to induce liver cell apoptosis,24 was found by Losordo et al22 to induce apoptosis of HUVECs with the greatest effect on HUVECs in G1. Interestingly, TNF-α itself promoted G1 arrest of the endothelial cells. G1 arrest and apoptosis were associated with a decrease in the promoter activity and expression of E2F-1 and with a reduction in cyclin A and cyclin A–dependent kinase activity. Both cyclin A and its associated kinase activity could be restored in HUVECs by overexpression of E2F-1, thereby rescuing the mitotic activity of the endothelial cells despite continued presence of TNF-α.22 To additionally investigate the relationship between TNF-α and E2F-1, we used sublethal TNF-α concentrations, which upregulated expression of the endothelial adhesion molecules VCAM-1, ICAM-1, and E-selectin.
A second reason for our focus on TNF-α is that inhibition of TNF-α by soluble TNF-α receptor has been shown to accelerate reendothelialization of balloon-injured rat carotid arteries accompanied by reduced neointima formation at the site of injury.23 Those findings lend credence to the hypothesis that TNF-α, expressed by VSMCs and endothelial cells at the injured vascular site,25 inhibits endothelial regrowth22 and promotes intimal hyperplasia.26 Whether E2F-1 restores endothelial regrowth in vivo, mimicking its effects on cultured endothelial cells, is not yet clear. However, irrespective of its effect on VSMC and endothelial cell proliferation, we show that E2F-1 can inhibit the pleiotropic NF-κB transcription factor in endothelial cells, thereby exerting effects that could add to its potentially beneficial influence on the vasculature stressed in vivo by hypercholesterolemia or mechanical injury.13,14,27,28⇓⇓⇓
In addition to the well-established relationship between NF-κB activity and adhesion molecule expression, both downregulated by overexpressed E2F-1, the possibility that induction of endothelial cell apoptosis by E2F-1 may contribute to the decreased adhesion molecule expression deserves consideration. Tanaka et al29 recently reported that serum deprivation of E2F-1–overexpressing NIH3T3 and Saos-2 cancer cells sensitized these cells to apoptosis due to accumulation of reactive oxygen species (ROS). Accumulation of ROS was associated with reduced induction of the ROS scavenger MnSOD, caused, in turn, by E2F-1 inhibition of NF-κB activity. We have not measured ROS in our experiments and are therefore unable to exclude a potential role of ROS-mediated apoptosis. However, as Tanaka et al noted, in the presence of serum, E2F-1 overexpressing cells grew faster than mock transfected cells, stressing the importance of serum deprivation for the ROS-dependent apoptosis. Except for the 6-hour infection period, we kept the HAECs in our experiments in culture medium supplemented with serum and growth factors, reducing the likelihood of growth factor deprivation–induced ROS accumulation. Our experiments additionally suggest that adhesion molecule expression seemed to be decreased at a time when the HAECs were still viable and did not show nuclear features of apoptosis on Hoechst staining (Figure 4). In addition, as mentioned above, experiments by Losordo et al showed that overexpression of E2F-1 in endothelial cells was able to rescue these cells from TNF-α–induced apoptosis. Nonetheless, the influence of longer-term E2F-1 expression on endothelial cell viability in vitro and in vivo requires additional study.
Our finding that E2F-1 suppresses expression of adhesion molecules suggests that overexpression of E2F-1 may harbor, in principle, the potential to delay the initiation (and progression) of atherosclerotic lesions, whereas other functions, such as VSMC apoptosis12 and endothelial growth stimulation,22 may play an additional role in the prevention of the fibroproliferative development of early lesions. Whether the expression of other NF-κB–dependent responses, including tissue factor upregulation, is inhibited after E2F-1 gene transfer needs to be seen.
In conclusion, E2F-1, in addition to its ability to promote S-phase entry of cytokine-stimulated endothelial cells and apoptosis of VSMCs, may exert potentially antiinflammatory activity on endothelial cells, which may have the potential to inhibit an early step in vascular lesion formation.
This work was supported by a grant from the MacDonald Foundation, a Scientist Development Grant of American Heart Association, and a Texas State Grant. We thank Dr Ta-Jen Liu (M.D. Anderson Cancer Center, Houston, Tex) for the gift of Ad.E2F-1.
Guest Editor for this article is Valentin Fuster, MD, PhD, Mount Sinai Medical Center, New York, NY.
- ↵Pierce AM, Gimenez-Conti IB, Schneider-Broussard R, et al. Increased E2F-1 activity induces skin tumors in mice heterozygous and nullizygous for p53. Proc Natl Acad Sci U S A. 1998; 95: 8858–8863.
- ↵DeGregori J, Leone G, Alexander M, et al. Distinct roles for E2F proteins in cell growth control and apoptosis. Proc Natl Acad Sci U S A. 1997; 94: 7245–7250.
- ↵Shelat HS, Liu TJ, Hickson-Bick DL, et al. Growth suppression of human coronary vascular smooth muscle cells by gene transfer of the transcription factor E2F-1. Circulation. 2001; 103: 407–414.
- ↵Rosato RR, Wang Z, Gopalkrishnan RV, et al. Evidence of a functional role for the cyclin-dependent kinase-inhibitor p21WAF1/CIP1/MDA6 in promoting differentiation and preventing mitochondrial dysfunction and apoptosis induced by sodium butyrate in human myelomonocytic leukemia cells (U937). Int J Oncol. 2001; 19: 181–191.
- ↵Hunt KK, Deng J, Liu T-J, et al. Adenovirus-mediated overexpression of the transcription factor E2F-1 induces apoptosis in human breast and ovarian carcinoma cell lines and does not require p53. Cancer Res. 1997; 57: 4722–4726.
- ↵Schreiber E, Matthias P, Muller MM, et al. Rapid detection of octamer binding proteins with “mini-extracts,” prepared from a small number of cells. Nucleic Acids Res. 1989; 17: 6419.
- ↵Scalia R, Appel JZ III, Lefer AM. Leukocyte-endothelium interaction during the early stages of hypercholesterolemia in the rabbit: role of P-selectin, ICAM-1, and VCAM-1. Arterioscler Thromb Vasc Biol. 1998; 18: 1093–1100.
- ↵Schindler U, Baichwal VR. Three NF-kB binding site in the human E-selectin gene required for maximal tumor necrosis factor α-induced expression. Mol Cell Biol. 1994; 14: 5820–5831.
- ↵Iademarco MF, McQuillan JJ, Rosen GD, et al. Characterization of the promoter for vascular cell adhesion molecule-1 (VCAM-1). J Biol Chem. 1992; 267: 16323–16329.
- ↵Ledebur HC, Park T. Transcriptional regulation of the intercellular adhesion molecule-1 gene by inflammatory cytokines in human endothelial cells. J Biol Chem. 1995; 270: 933–943.
- ↵Spyridopoulos I, Principe N, Krasinski KL, et al. Restoration of E2F expression rescues vascular endothelial cells from tumor necrosis factor-α–induced apoptosis. Circulation. 1998; 98: 2883–2890.
- ↵Krasinski KL, Spyridopoulos I, Kearney M, et al. In vivo blockade of tumor necrosis factor-α accelerates functional endothelial recovery after balloon angioplasty. Circulation. 2001; 104: 1754–1756.
- ↵Tanaka H, Sukhova G, Schwartz D, et al. Proliferating arterial smooth muscle cells after balloon injury express TNF-α but not interleukin-1 or basic fibroblast growth factor. Arterioscler Thromb Vasc Biol. 1996; 16: 12–18.
- ↵Rectenwald JE, Moldawer LL, Huber TS, et al. Direct evidence for cytokine involvement in neointimal hyperplasia. Circulation. 2000; 102: 1697–1702.
- ↵Li H, Cybulsky MI, Gimbrone MA Jr, et al. An atherogenic diet rapidly induces VCAM-1, a cytokine-regulatable mononuclear leukocyte adhesion molecule, in rabbit aortic endothelium. Arterioscler Thromb. 1993; 13: 197–204.