A sustained reduction in IkappaB-beta may contribute to persistent NF-kappaB activation in human endothelial cells.

The responses of vascular endothelial cells (EC) to tumor necrosis factor-α (TNF), interleukin-1α (IL-1), and phorbol myristate acetate (PMA) were compared with respect to the kinetics of (i) NF-κB activation, (ii) IκB-α and IκB-β degradation, and (iii) NF-κB-dependent cell surface molecule expression. TNF rapidly (≤20 min) and persistently (>20 h) activates NF-κB; IL-1 rapidly activates NF-κB, but activity declines by 3 h and further by 20 h; PMA slowly and transiently activates NF-κB. Untreated EC contain the inhibitory proteins IκB-α and IκB-β. The onset of NF-κB activation correlates with degradation of IκB-α, but IκB-α reappears by 4 h without resequestration of NF-κB. TNF causes a rapid but partial (50%) reduction in IκB-β, which does not recover by 22 h; IL-1 and PMA cause slower and less sustained reductions in IκB-β. All three agonists induce de novo expression of E-selectin (CD62E) and vascular cell adhesion molecule-1 (CD106) and increase expression of intercellular adhesion molecule-1 (CD54) at 4 h. TNF induces sustained increases in vascular cell adhesion molecule-1 and intercellular adhesion molecule-1 and increases human leukocyte antigen class I molecules at 24 h. We conclude that TNF causes persistent activation of NF-κB in human EC and that this may result from sustained reductions in IκB-β levels.

Activation of the transcription factor NF-B is required for the full expression of many TNF-inducible 1 genes in vascular endothelial cells (EC), including E-selectin (CD62E, also known as endothelial-leukocyte adhesion molecule-1) (1), vascular cell adhesion molecule-1 (VCAM-1, CD106) (2), intercellular adhesion molecule-1 (ICAM-1, CD54) (3,4), and human leukocyte antigen (HLA) class I (5). Activated NF-B binds to 10-base pair recognition sequences (B sites, consensus GGGRHTYY-CC (6)) found in the 5Ј-flanking regions of these genes and promotes transcription through direct interaction with the transcription initiation complex (7). Cytokine-induced expression of these molecules on the lumenal surface of vascular endothelial cells is critical in recruiting leukocytes from the bloodstream into sites of inflammation (reviewed in Ref. 8).
NF-B is composed of members of the Rel family of ubiquitous transcriptional activators that includes p50 (also called NFKB1), p52 (NFKB2), p65 (RelA), c-Rel (Rel), and RelB (reviewed in Ref. 9). Within the Rel family, every member can form homo-and heterodimers except RelB, which is preferentially expressed in lymphoid cells and forms only heterodimers with p50 or p52 (10). Both p65 and c-Rel are transcriptional activators. The p50 homodimer can repress transcription from some promoters when expressed as a transgene in vivo (11), but it can activate transcription in vitro (12), and it is thought to mediate constitutive transcription of major histocompatibility complex class I heavy chain (13,14) and light chain genes (15). Similarly, p50 homodimers stimulate transcription of human immunoglobulin B-containing but not HIV B-containing promoters (12). Variations within the B consensus site may account for such differences, e.g., transcriptional activation mediated by the HIV B (GGGACTTTCC) is strongly induced in vitro by NF-B but the interferon-␤ B (GGGAAATTCC, also called PRDII) is weakly induced (16). Sequences flanking and within the B core can also contribute to differential responses, e.g., the binding of a high mobility group-like protein adjacent to an NF-B converts the NF-B from a transcriptional activator to a repressor (17), and the binding of the high mobility group I(Y) protein to the AT-rich center of the IFN-␤ B has been shown to facilitate the binding or activity of NF-B (18). The high mobility group I(Y) protein has been implicated in the transcriptional regulation of the endothelial-leukocyte adhesion molecule-1 (19) and VCAM-1 (20) genes by cytokines in EC.
Latent forms of NF-B are found in the cytoplasm where they are bound constitutively by members of a family of proteins called inhibitors of B (IB), which includes IB-␣ (also called MAD-3, 37 kDa), IB-␤ (46 kDa), IB-␥, and Bcl-3 (reviewed in Ref. 21). Both IB-␣ and IB-␤ bind to the strongly transactivating Rel proteins p65 and c-Rel and inhibit their binding to DNA (16,22). Neither IB-␣ nor IB-␤ bind to p50. The two additional characterized forms of IB are specific for p50-containing dimers. IB-␥ corresponds to the carboxyl-terminal portion of the p50 precursor protein p105 and is the product of alternative splicing (23); however, it has only been detected in murine pre-B cells (24). Bcl-3 is found in only low amounts in a restricted set of tissues, including certain B cell leukemias, and preferentially inhibits the binding of p50 to DNA (25). Bcl-3 is also reported to act as a transcriptional activator in association with p52 homodimers (26), although this Bcl-3-mediated transactivation may be an indirect effect of sequestering the transcriptional repressors p50 and p52 (27). Due to their low abundance and restricted tissue distribution, these IB isoforms are thought to play only minor roles in regulating NF-B in most cell types.
Activation of NF-B DNA binding potential and movement of NF-B to the nucleus results from the phosphorylation and subsequent proteolytic degradation of IB (28,29). IB-␣ is rapidly degraded in many cell types, including cultured human umbilical vein EC, following treatment with cytokines that activate NF-B (30,31). The synthesis of IB-␣ is itself induced by NF-B and has been proposed to mediate eventual resequestration of NF-B in the cytoplasm, forming a potential negative feedback loop (30). IB-␤, the second major IB, has recently been cloned and characterized in B lymphocytes, where it was shown to be degraded upon treatment with IL-1 or lipopolysaccharide, which cause a prolonged activation of NF-B in B cells, but not upon treatment with TNF or PMA, which cause only transient activation of NF-B in B cells (22). In contrast to IB-␣, agonist-induced degradation of IB-␤ occurs relatively slowly in these cells, appearing to decrease markedly only after 60 min of stimulation (at which time IB-␣ has returned to normal levels), and remains low during the treatment period. In addition, IB-␤ mRNA levels are not increased by activators of NF-B in B cells (22).
In the present study we have examined the effects of TNF, IL-1, and PMA on NF-B activation and IB degradation in human EC. Our data show that TNF causes the most persistent NF-B activation and the most persistent reduction in IB-␤ levels and has the greatest effects on NF-B regulated genes at later times.

EXPERIMENTAL PROCEDURES
Cells, Cytokine Treatments, and Harvest-Human umbilical vein EC were isolated by collagenase digestion, pooled between 2-4 donors, and cultured on collagen in 20% fetal calf serum/M199/glutamine/penicillin/ streptomycin (all Life Technologies, Inc.) supplemented with endothelial cell growth factor (Collaborative Research, Bedford, MA) as described previously (32). Confluent cultures of human umbilical vein EC (passage 3 or 4) were adjusted to 8 ml on 10-cm plates and treated for the times indicated in the figures with recombinant human TNF (100 units/ml, expressed in Escherichia coli; 2.5 ϫ 10 7 units/mg, a gift of W. Fiers, State University of Ghent, Ghent, Belgium or 5 ng/ml, R&D Systems, Minneapolis, MN), PMA (100 nM, Sigma), or recombinant human IL-1␣ (1 ng/ml, R&D Systems, Minneapolis, MN). Cells were harvested by rinsing twice with PBS (Mg ϩ2 -and Ca ϩ2 -free) and then incubating in 1 ml of 1 mM EDTA/PBS (10 min at room temperature). Cells were detached by pipetting and placed in microcentrifuge tubes on ice. Similar results were obtained when the cells were scrape-harvested into ice-cold PBS or harvested by trypsin treatment.
Cytoplasmic and Nuclear Extractions-Cytoplasmic and nuclear extracts were prepared by a modified mini-extraction protocol (33). Cells were pelleted (approximately 500 ϫ g, 4 min, 4°C) and resuspended in hypotonic buffer A (0.2 ml, 10 mM HEPES, pH 7.9, 10 mM KCl, 1.5 mM MgCl 2 ) supplemented with protease inhibitors (leupeptin and aprotinin each 1 g/ml, phenylmethylsulfonyl fluoride 0.5 mM) and incubated for 15 min on ice. Cells were lysed by the addition of 25 l 2.5% Nonidet P-40/buffer A and mixed by inversion, and the nuclei were pelleted (500 ϫ g, 4 min, 4°C). Supernatant was recovered (200 l of cytoplasmic extract), placed into microcentrifuge tubes, and frozen by immersion in liquid nitrogen. The nuclear pellet was resuspended in extraction buffer C (20 mM HEPES, pH 7.9, 0.45 M NaCl, 1 mM EDTA supplemented with the protease inhibitors) and incubated for 15 min at 4°C on a rocking platform, and then the tubes were centrifuged (10 min at 14,000 ϫ g, 4°C). Extracts were diluted 1:1 in buffer D (20 mM HEPES, pH 7.9, 0.1 M KCl, 0.2 mM EDTA, 20% glycerol (34)), yielding 1.5-2.0 g/l against a bovine serum albumin standard in a modified Bradford protein assay (Bio-Rad), frozen in liquid nitrogen, and stored at Ϫ20°C.
DNA Binding Assay (Electrophoretic Mobility Shift Assay)-Consensus sequence B DNA was purchased (5Ј-AGTTGAGGGGACTTTC-CCAGGC, Promega). The core B sequence that matches HIV and mouse Ig B is underlined (35). HLA class I B sequence oligos containing a symmetrical B sequence (5Ј-CGTTGGGGATTCCCCACTCC) and HLA interferon consensus sequence (CCACAGTTTCACTTCTG-CACCT (5)) oligos were synthesized (Critical Technologies Program, Yale Medical School) and annealed. Both (blunt-end) probe DNAs were labeled with [ 32 P]␥ATP with polynucleotide kinase (New England Biolabs, Beverly, MA) and separated from unincorporated nucleotides over a Sephadex G-25 (Pharmacia Biotech Inc.) spun column. To 1 l of double-stranded nonspecific DNA (1 g poly(dI)⅐poly(dC), Pharmacia) was added 4 l of nuclear extract (6 -8 g) and 5 l of diluted probe B (2 fmol, approximately 4,000 cpm). The reaction was mixed by gentle rocking and incubated at room temperature for 20 min, and then 5 l was loaded on a 3.5% polyacrylamide gel/0.25 ϫ TBE buffer (Tris, borate, EDTA) and separated at 220 V/18 cm for 2 h. Gels were dried under vacuum and exposed to a storage phosphor screen for quantification and documentation (PhosphorImager, Molecular Dynamics, Sunnyvale, CA). Competition experiments were performed as above except that 100-fold excess competitor DNA was added to the incubations in 1 l immediately prior to the addition of probe DNA in 4 l.
Antibody EMSA-To 1 l of nonspecific DNA (1 g poly(dI)⅐poly(dC)) was added 4 l (6 -8 g of protein) of nuclear extract, 2 l of water, and 1 l of affinity-purified rabbit polyclonal antisera specific for RelA (sc-109), p50 (sc-114), or cRel (sc-71; all antisera were from Santa Cruz Biotechnology, Santa Cruz, CA). After incubated on ice for 1 h, 2 l (2 fmol) of probe DNA was added, and the incubation was continued at room temperature for 20 min; then 5 l was loaded on a gel and analyzed as described above.
Western Blotting of IB-␣ and IB-␤-Cytoplasmic extracts that had been frozen in liquid nitrogen were thawed and ultracentrifuged at 300,000 ϫ g for 30 min at 4°C (90,000 rpm, TLA100 rotor, Beckman). The protein was quantified in a BCL protein assay with bovine serum albumin standard (Pierce). Extracts (25 g) were resolved by SDSpolyacrylamide gel electrophoresis, electroblotted onto a polyvinylidene difluoride membrane (Immobilon, Millipore), stained by antibodies against IB-␣ (Santa Cruz) or IB-␤ (rabbit antisera raised against the IB-␤ protein (22)) and a horseradish peroxidase-coupled secondary antibody (ECL, Amersham Corp.), and detected on film. Molecular weight standards were run in an adjacent lane (Rainbow markers, Amersham Corp.). Autoradiographs were quantitated by densitometry (Molecular Dynamics), and the level of significance was determined by a one-tailed Student's t test with unequal variance (36).

RESULTS
The time course of NF-B activation in EC after treatment with TNF, IL-1, or PMA was examined. Nuclear extracts prepared from EC were assayed for activated NF-B in an EMSA of DNA binding factors using a radiolabeled consensus sequence B probe containing the core B site, GGGACTTTCC, that matches the HIV-I and the murine Ig gene B sequences. Nuclear extracts from control cells form two complexes, a faster migrating lower complex (Fig. 1a, LC) and an upper complex (Fig. 1a, UC). The time courses of NF-B activation by the agonists TNF, IL-1␣, or PMA in endothelial cells were determined by testing nuclear extracts prepared from replicate cultures of EC treated for 20 min, 3 h, or 20 h. Different kinetics of NF-B activation were observed with each of the three agonists, TNF, IL-1, and PMA, varying in the rate of onset and persistence of NF-B activation (Fig. 1, a and b). TNF and IL-1␣ rapidly (Յ20 min) activate NF-B (Fig. 1a, lanes 3 and 9), which remains nearly constant for Ͼ20 h in the case of TNF and declines by 3 h in the case of IL-1␣. PMA activates NF-B more slowly and transiently, with little effect at 20 min but significant activation at the 3 h time point (Fig. 1a, lane 7) and declining to nearly control levels by 20 h. Similar results were obtained in seven independent experiments with TNF and PMA and five experiments with IL-1␣ using as probes either the Ig B or the HLA class I B. The amount of NF-B activated in two independent experiments was quantified, and the mean and standard deviations were determined (Fig. 1b).
To test the specificity of the DNA binding activity, competition experiments were performed with unlabeled B and an unrelated interferon consensus sequence DNA (Fig. 1c). Both the upper and lower complexes formed by TNF-activated NF-B binding to an HLA class I B probe DNA were completely competed by 100-fold excess unlabeled HLA class I B (Fig. 1c, lane 4) and Ig B (Fig. 1c, lane 5) but not by the unrelated interferon consensus sequence DNA (Fig. 1c, lane 6).
The doses of agonists used here were established as optimal in pilot experiments. Nevertheless, we considered the possibility that the decline in IL-1␣-activated NF-B at later times of treatment was the result of cytokine depletion. The addition of fresh IL-1␣ for 30 min to cultures that had been treated with IL-1␣ for 20 h (when activated NF-B has declined to approximately 65% of the maximum levels) resulted in the activation of only a small amount of additional NF-B (increasing to approximately 75% of maximal levels, data not shown). Thus, the decline of activated NF-B in IL-1-treated cells cannot be attributed to cytokine depletion.
The composition of TNF-activated NF-B could be expected to change over time because TNF treatment of EC has been shown to greatly increase the level of IB-␣ and p105 (p50 precursor) mRNAs and transiently induce cRel mRNA, whereas p65 mRNA levels are induced only slightly (31,30). Because different constituents of NF-B provide different transcriptional activities, such a change could contribute to the effects of persistent activators of NF-B. The composition of activated NF-B was investigated by antibody EMSA (supershifts) with nuclear extracts of cells treated with TNF for 30 min or 22 h (Fig. 2a) or treated with TNF, PMA, or IL-1 for 3 h (Fig. 2b). At early or late times of TNF treatment or after treatment with TNF, PMA, or IL-1 for 3 h, the upper complexes contain p65, the lower complexes contain p50, and fractions of both the upper and lower complexes contain cRel (Fig. 2). Therefore, no change was detected in the subunit composition of NF-B in EC during the first 24 h of TNF treatment, nor were differences detected in the subunit composition of NF-B activated by TNF, IL-1, or PMA at 3 h of treatment.
All known activators of NF-B (e.g., IL-1, lipopolysaccharide, TNF, and PMA) induce the degradation of IB-␣ in B cells, but only persistent activators (e.g., IL-1 and lipopolysaccharide) have been shown to induce the degradation of IB-␤. To compare the levels of IB-␣ and IB-␤ in agonist-treated EC, cytoplasmic extracts from treated cells were Western blotted with IB isoform-specific antisera. Consistent with previous reports (31,29), TNF and IL-1 rapidly induce nearly complete proteolysis of IB-␣ (Fig. 3a). In parallel with the slower activation of NF-B by PMA, proteolysis of IB-␣ induced by PMA is also somewhat delayed and reaches a maximum between 30 and 60 min (data not shown). IB-␤ is detected in the cytoplasm of untreated EC (Fig. 3b, lane 1), and all three agonists induce significant reductions in IB-␤ (Fig. 3, b and c). Similar results were obtained in five independent experiments. (The additional lower unmarked bands on the IB-␤ blot are not found consistently; these bands may correspond to degradation products, or they may result from cross-reactivity of the polyclonal antiserum.) To assist in evaluating the quantity of proteins in the extracts, one lane was intentionally loaded with one-half the normal protein load (Fig. 3, lane 2, in a and b, labeled half).
To obtain a measure of the extent of IB-␤ reduction, densitometry was performed on the autoradiograph shown in Fig. 3b and on autoradiographs from two additional independent ex- periments. All values were normalized by subtraction of background counts and division by the counts in the untreated lane. The time of treatment varied slightly among experiments, so for the purpose of data analysis, treatments of 20 -30 min, 3-4 h, and 20 -26 h were pooled, and the means and standard deviations were determined (Fig. 3c). TNF rapidly induces the reduction to approximately 25% of control IB-␤ levels, and levels of IB-␤ remain low throughout the period of treatment. IL-1 and PMA more slowly induce the reduction to approximately 40% of control IB-␤ levels, which return to control levels by 20 h of IL-1 treatment but remain low for up to 20 h after PMA treatment.
Newly synthesized IB-␣ has been reported to be localized in the nucleus of HeLa S3 cells treated briefly with TNF (37). Nuclear extracts from IL-1␣-treated EC were tested for the presence of IB isoforms. Although immunoreactive species were detected, only trace amounts of proteins of the same size as the cytoplasmic proteins were observed in both untreated or IL-1-treated nuclear extracts (data not shown), suggesting that nuclear IB is not a major cause of the reduced NF-B binding activity in at later times of IL-1 treatment.
NF-B is an important transcriptional regulator of the inducible endothelial cell surface molecules E-selectin, VCAM-1, ICAM-1, and HLA class I. To compare the activation of NF-B with the induction of these molecules, untreated EC and EC treated with TNF, PMA, or IL-1 were stained by indirect immunofluorescence and measured by fluorescence flow cytom-etry. All three agonists effectively induce expression of ICAM-1, VCAM-1, and especially E-selectin at 4 h, but TNF is much more effective than PMA or IL-1 in inducing ICAM-1 and VCAM-1 expression at later times (Fig. 4). TNF alone significantly induces expression of HLA class I molecules, which begin to increase only at later times of treatment. Therefore, the strength of induction of these persistent surface molecules correlates with the strength and persistence of NF-B activation.

DISCUSSION
The agonists TNF, IL-1, and PMA activate NF-B with different kinetics in EC. TNF activates NF-B rapidly and persistently, IL-1 activates NF-B rapidly but with less persistence, and PMA activates NF-B slowly and transiently. In contrast to these results obtained from endothelial cells, in B cells IL-1 activates NF-B for over 24 h, and TNF activates NF-B only transiently (22), whereas in a murine T cell lym-  (38). These contrasting results demonstrate a strong dependence on cell type for the behavior of agonists that activate NF-B.
No differences in subunit composition were detected in the NF-B activated by TNF at 30 min, 3 h, or 22 h of treatment or by PMA or IL-1 at 3 h (Fig. 2). In each case, the predominant upper complex contains p65, and the lower complex contains p50 and c-Rel. Similarly, no differences in Rel proteins activated by TNF, PMA, or IL-1 were noted in a mouse T cell line (38). Therefore, the increases in Rel and IB-␣ mRNA may not be directly reflected in increased protein expression. Such posttranscriptional control of IB-␣ expression has been observed in primary human monocytes (39). Alternatively, NF-B-induced transcription of the genes encoding IB-␣ and the p50 precursor (p105) might fail to produce changes in NF-B subunit composition because of compensating effects of IB-␣ induction and degradation together with the IB function of the p105 protein (IB-␥).
The persistence of NF-B activation by different agonists may differ depending upon the duration of the intracellular signals initiated by the agonist. For example, activation of NF-B in HeLa cells requires the continued presence of TNF; when TNF is washed away from the cultured cells, B binding activity rapidly disappears from the nucleus (data not shown). It seems likely, therefore, that TNF continues to induce the proteolysis of IB-␣ and release of NF-B, establishing a new balance through the concomitant NF-B-driven increased synthesis of IB-␣ and p105. In contrast, PMA activation of PKC in EC is followed by down-regulation of PKC and termination of the signal even in the continued presence of the PMA agonist (40).
The rapid depletion of IB-␤ in EC is the principal difference in the behavior of IB-␤ between EC and B cells in response to chronic activators of NF-B (22). IB-␤ mRNA is not increased by activators of NF-B (22), which may contribute to the failure of IB-␤ to return to resting levels in EC. TNF may also cause the persistent degradation of IB-␣, which is countered by concomitant NF-B-mediated increase in IB-␣ synthesis (31). PMA activation of NF-B is maximal around 3-4 h, which coincides with the maximum depletion of IB-␤ ( Figs. 1 and 3). Thus, NF-B released from IB-␤ may account for the majority of the PMA-activated NF-B. IL-1 activation of NF-B is rapid, which probably results from the rapid depletion of IB-␣, but not as strong as TNF activation, perhaps because IL-1 induces the degradation of IB-␤ more slowly. IL-1 activation of NF-B is also less sustained than TNF activation, which correlates with the less sustained reduction of IB-␤ induced by IL-1.
The results of these experiments support the hypothesis that persistent activators of NF-B reduce IB-␤ levels and further suggest that a rapid reduction in IB-␤ levels observed in EC may also contribute to initial NF-B activation. Correlations between NF-B and IB-␣ and -␤ will have to be sought in additional cell types and with additional agonists in order to test the generality of the model wherein both IB-␣ and IB-␤ contribute regulating nuclear translocation of NF-B. The results shown are from one of two independent experiments with similar results. EC were left untreated or treated with TNF, PMA, or IL-1 for the times indicated, harvested by trypsinization, incubated with monoclonal antibodies specific for E-selectin, ICAM-1, VCAM-1, or HLA class I, and stained with a fluoresceinated secondary antibody. Staining was quantified on a fluorescence-activated cell sorter flow cytometer, and the mean fluorescence channel number was determined from 2000 cells gated on forward and side scatter parameters.