ABSTRACT
The E-selectin cell adhesion protein plays a critical role in mediating
adherence of leukocytes to endothelium at sites of inflammation. Cytokine-induced E-selectin expression on the surface of endothelial cells is
transient; mRNA expression peaks at 3-4 h after induction and returns to basal levels within 24 h. The
mechanism for this transcriptional down-modulation is not known. Promoter binding factors responsible for induced
gene expression include NF-
[kappa]
B, which binds at three sites within the E-selectin promoter, and HMG-I(Y), which binds to the A/T-rich core found at the centre of these binding sites.
Distamycin is an antibiotic that also binds A/T-rich DNA and inhibits HMG-I(Y) DNA binding. To study the role of HMG-I(Y) in E-selectin expression, we have examined the effect of
distamycin on the cytokine-induced E-selectin expression cycle. We found that distamycin prolonged E-selectin expression, both by sustaining mRNA transcription and
by extending the transcript's half-life. The distamycin effect on transcription was mediated through one of
the three NF-
[kappa]
B-HMG-I(Y) binding sites (NF-
[kappa]
BII) within the promoter. This suggests that the NF-
[kappa]
B-HMG-I(Y) complex interacting at the NF-
[kappa]
BII site plays a role not only in cytokine induction of E-selectin expression, but also in its down-modulation.
E-Selectin (also known as CD62E and ELAM-1) is a membrane glycoprotein exclusively expressed on the surface
of endothelial cells during inflammation
in vivo
or after cytokine, lipopolysaccharide or phorbol ester induction
in vitro
(
1
-
5
). E-Selectin plays a critical role at inflammation sites in mediating the
adhesion to endothelial cells and extravasation of neutrophils and a subset of
T lymphocytes. Both up- and down-modulation of this selectin is tightly controlled. Several
transcription factors and their cognate DNA binding elements in the E-selectin promoter have been identified that are essential for activation
of the E-selectin promoter. NF-[kappa]B binding to the -94/-85 region in the E-selectin promoter is essential but not
sufficient for cytokine induction of the gene (
6
-
9
). Induction of the E-selectin promoter also requires NF-[kappa]B complexes to bind at positions -126 (
10
,
11
) and -116 (
10
,
12
). Moreover, E-selectin promoter activation requires members of the ATF family of
transcription factors that bind the NF-ELAM1 site at position -157 (
13
,
14
) and as yet unidentified protein(s) binding the NF-ELAM-2 site at position -104 (
13
). Tissue specificity of the E-selectin gene is associated with CpG methylation of the proximal promoter
(
15
).
Binding of NF-[kappa]B and ATF factors to their promoter sites is enhanced by a member
of the high mobility group family of proteins, HMG-I(Y) (
11
-
14
). High mobility group proteins are non-histone chromosomal proteins that have been shown to bind the minor groove
of A/T-rich regions of double-stranded DNA
in
vitro
(
16
). HMG-I(Y) also plays an important role in virus-mediated induction of the [beta]-IFN gene through its interaction with NF-[kappa]B (
17
,
18
).
Several compounds (e.g. distamycin, berenil, bis-benzamidine and netropsin) have been described that compete with HMG-I(Y) for binding in the minor groove of A/T-rich DNA regions (
19
-
21
). As HMG-I(Y) binds to the E-selectin promoter and plays a role in its activation, we decided to
analyse the effect of these specific competitors of HMG-I(Y) binding on activation of the E-selectin promoter by cytokine treatment of endothelial cells. In
contrast to our expectation of reduced or blocked E-selectin expression, we found that distamycin significantly prolonged E-selectin expression induced by both IL-1[beta] and TNF-[alpha]. We have studied the mechanism of this
distamycin effect to learn more about the role of HMG-I(Y) in the regulation of E-selectin transcription. We show that the NF-[kappa]BII element in the E-selectin promoter is responsible for the
distamycin effect and is likely to play a role in the physiological down-modulation of E-selectin gene expression, possibly through DNA-bending.
Human endothelial cells (HUVECs) were isolated from human umbilical cords by collagenase treatment and grown as previously described (
22
). All experiments were performed using HUVECs at passages 3-5.
HUVECs were grown to confluence in 5 cm Petri dishes, pre-treated for different times with distamycin (10 [mu]M) and induced for 4 h with 10 U/ml IL-1 in complete Clonetics medium containing 2% foetal calf serum.
Cells were removed with the help of a rubber policeman in phosphate-buffered saline (PBS), 1 mM EDTA pre-warmed at 37oC. Cells were collected by spinning for 5 min at 1000 r.p.m. at
4oC and re-suspended in PBS containing 10 mg/ml BSA and 0.1 mg/ml NaN
3
. Cells (4 * 10
5
) were incubated for 30 min at 4oC in the presence of the first monoclonal antibody, washed and incubated
for a further 15 min at 4oC in the presence of a second antibody (anti-mouse-IgG conjugated to FITC). The cells were washed in PBS, BSA and
submitted to FACS analysis. The MFI values from three different experiments are
expressed as a percentage of the maximal value obtained with 10 U/ml IL-1.
Total RNA was isolated from confluent HUVECs using the guanidinium
isothiocyanate/CsCl procedure (
23
). Total RNA (20 [mu]g/lane) was fractionated on a 1% agarose gel containing 2.2 M formaldehyde,
blotted onto nylon membrane filters (GeneScreen Plus; NEN) and hybridized to
nick-translated E-selectin-1, ICAM-1, VCAM-1 or human [beta]-actin
32
P-labelled cDNA probes (
24
). Filters were washed twice for 30 min at 37oC in 2* SSC, 0.1% SDS and for 15 min at 55oC in 0.2* SSC, 0.1% SDS and then exposed for different times to
Kodak XAR films at -80oC. Results were visualized by autoradiography and quantified using
an Ambis (San Diego, CA) radioactivity scanner.
HUVECs which had been left untreated or pre-treated for 24 h with distamycin (10 [mu]M) were incubated with human IL-1[beta] (10 U/ml) in the presence or absence of distamycin (10 [mu]M) and nuclei were isolated after 1, 6 or 24 h.
Nuclear run-on assays were performed as previously described (
22
) starting with 3 * 10
7
cells/treatment. For each assay, nuclei were re-suspended in 300 [mu]l transcription buffer containing 250 [mu]Ci [[alpha]-
32
P]UTP and incubated for 20 min at 30oC. The radiolabelled RNA was purified using the guanidinium
isothiocyanate/CsCl procedure (
23
) and equal amounts of c.p.m. were used for hybridization to nitrocellulose
filters that contained alkali-denatured target cDNA. The filters were prepared by slot blotting of 7.5 [mu]g target DNA followed by baking for 2 h
in vacuo
at 80oC. The DNAs used were plasmids pCDM8 and pELAM-1 (plasmid pCDM8 containing E-selectin cDNA;
22
). Results were visualized by autoradiography and quantified by densitometric
scanning of the autoradiograms.
The plasmid constructs were made by inserting different E-selectin promoter regions into the
Eco
RI site of a pCAT-promoter plasmid (Promega) containing the enhancerless SV40 promoter. The
plasmid used (as shown in Fig.
7
) contained; (i) the -383 promoter region with the NF-[kappa]BI and NF-[kappa]BII wild-type elements; (ii) the NF-[kappa]BI wild-type element and a
mutated NF-[kappa]BII element; (iii) a mutated NF-[kappa]BI and a NF-[kappa]BII wild-type element, as described earlier
(
14
).
Transfection of plasmid DNA into HUVECs was carried out using the DEAE-Dextran method as described (
25
). HUVECs were pre-treated for 24 h with distamycin, transfected and treated with distamycin
alone or in combination with IL-1[beta] (40 U/ml). After 24 h treatment, the cells were collected by
scraping and cellular extracts were used for chloramphenicol acetyl transferase
(CAT) assays essentially as described (
26
). CAT enzymatic activity was quantified using an Ambis (San Diego, CA)
radioactivity scanner.
For DNase I footprinting a PCR-generated DNA fragment that spanned part of the E-selectin promoter (-162 to -71) and contained the three NF-[kappa]B sites was used: GTA ACA CAG AGT TTC TGA
CAT CAT TGT AAT TTT AAG CAT CGT G
The
in vitro
DNase I assay protocol was performed by incubating this fragment for 10 min at room temperature with 1.5 U/ml DNase I,
after pre-incubation for 1 h on ice in the absence or presence of distamycin at
various concentrations. Digested products were analysed on 8% polyacrylamide
(19:1 acrylamide/bis-acrylamide), 50% urea, 0.5* TBE sequencing gels.
Subcloning of the NF-[kappa]BI element into pBEND2 was described previously (
28
). In order to insert the NF-[kappa]BII element, oligonucleotides CTA GCC CGG GAA TAT CCA CGA TG and
TCG ACA TCG TGG ATA TTC CCG GG were phosphorylated, annealed and cloned into
the
Xba
I-
Sal
I sites of pBEND2 using standard procedures (
29
). PBEND2 (
30
) was a kind gift of Dr S.Adhya. The resulting construct was sequenced to verify
that it contained a single insert. The polylinkers plus insert from both
constructs were PCR amplified using oligonucleotides PCR-BEND2a and PCR-BEND2b (
28
). The resulting fragments were treated with phenol, cut separately with the
different restriction enzymes, treated with phosphatase, purified from a
MetaPhor agarose (FMC BioProducts) gel, end-labelled using [[gamma]-
32
P]ATP and purified again on a 12% non-denaturing polyacrylamide gel. The total length of each of the probes was
139 bp. The probes were incubated for 20 min on ice in 25 [mu]l BCA buffer (
13
,
27
,
31
) and 3 [mu]g poly(dI[middot]dC) (Pharmacia) with or without 0.1 mM distamycin. Migration was at
room temperature in a 6.5% acrylamide, 0.5* TBE gel (1.5 mm thick, 16 * 18 cm; Hoefer Scientific). The gel was run at 10 mA for 1.5 times
the time taken for the bromophenol blue marker to run off the gel.
E-Selectin expression on the surface of endothelial cells peaks 4-5 h after cytokine induction and both protein and mRNA return to
their basal levels within 24 h. This drop in E-selectin mRNA transcription occurs even in the continuous presence of
cytokine (Fig.
1
A;
1
,
6
) and even though high levels of NF-[kappa]B persist up to 48 h post-induction (
32
). To study the role of HMG-I(Y) in this cytokine-induced, transient E-selectin expression, we treated HUVECs with IL-1 in conjunction with distamycin. Since distamycin is
capable of inhibiting HMG-I(Y) binding to DNA (
19
), we anticipated an inhibition of E-selectin expression. In contrast, we observed that when E-selectin expression was induced by IL-1 in the presence of distamycin, the high level of E-selectin protein measured by flow cytometry (FACS) was
significantly prolonged (Fig.
1
A). At 24 h post-IL-1 induction, distamycin maintained E-selectin expression at ~70-80% of the maximal level observed at 4 h post-induction. At 96 h, the level of E-selectin expression in IL-1 plus distamycin-treated cells had only
fallen to ~50% of the maximum level.
We wished to evaluate whether an enhanced transcription initiation rate or
increased mRNA stability are part of the mechanism by which distamycin prolongs
E-selectin expression. To examine mRNA stability we measured the E-selectin mRNA half-life in the presence and absence of distamycin following IL-1 treatment. The transcript's half-life was determined by monitoring the level of E-selectin mRNA at different times following
treatment of cells with actinomycin D, an inhibitor of ongoing transcription. A
stabilization of E-selectin mRNA was seen when cells were induced by cytokine in the presence
of distamycin (Fig.
5
A). To determine a distamycin effect on the rate of E-selectin gene transcription, we performed transcriptional run-on analysis. In nuclei purified from IL-1-induced endothelial cells pre-treated with distamycin, ongoing E-selectin transcription was assayed at
different times following IL-1 induction. Transcription of the E-selectin gene was maintained longer following distamycin treatment
(Fig.
5
B and C, lanes 4, 6 and 8 versus 3, 5 and 7). We conclude that the prolonged E-selectin expression caused by distamycin results both from mRNA
stabilization and prolonged E-selectin transcription.
To analyse the region(s) in the E-selectin promoter to which distamycin binds, we performed
in vitro
DNase I footprinting. We used a fragment containing the region of the E-selectin promoter previously defined as being sufficient to mediate
cytokine induction (positions -162 to -71). In the presence of distamycin, the region corresponding to
the NF-[kappa]BII site (-117 to -108) was protected (Fig.
6
). In contrast, neither the NF-[kappa]BI (-84 to -860) nor the NF-[kappa]BIII (-117 to -118) site was strongly
protected in the presence of distamycin. These results suggest that the effect
of distamycin on E-selectin gene expression is specifically mediated through the NF-[kappa]BII site.
We also attempted to narrow down the E-selectin promoter element(s) involved in distamycin's up-regulation in a functional assay. E-Selectin promoter-CAT reporter constructs in which specific NF-[kappa]B sites had been mutated were tested in
transient transfection assays. As found previously (
6
), the 383 bp proximal E-selectin promoter sequence was sufficient to mediate IL-1 induction of a fused reporter gene (Fig.
7
A). Using the same construct, we also found that induction was greatly enhanced
when cells had been treated with IL-1 in the presence of distamycin (Fig.
7
A). In this assay system, distamycin alone was only able to weakly activate the
E-selectin promoter. Thus, the distamycin effect on endothelial cell
expression of E-selectin is mediated by a restricted region of the E-selectin promoter. When the NF-[kappa]BI site was mutated, IL-1 induced very weak expression of the E-selectin promoter and distamycin
significantly enhanced this IL-1 induction (Fig.
7
B). This result suggests that the distamycin effect is not mediated by the NF-[kappa]BI site. Finally, a mutation in the NF-[kappa]BII site resulted in a near complete loss of IL-1 induction (Fig.
7
C) and in a complete loss of any distamycin effect. These data indicate that the
NF-[kappa]BII site plays a critical role in E-selectin mRNA transcription; in contrast to the NF-[kappa]BI site, its loss cannot be functionally
compensated for by addition of distamycin. These findings are consistent with
the role we propose for NF-[kappa]BII as the site of distamycin action.
A possible mode of action for distamycin could be the displacement of a
transcriptional repressor or to increase the binding of a
trans
-activator to the E-selectin promoter. However, band shift experiments with
radiolabelled oligonucleotides carrying NF-[kappa]BI or NF-[kappa]BII sites and nuclear extracts prepared from cells
treated with interleukin or interleukin plus distamycin failed to show a
difference in protein binding (data not shown). We therefore hypothesize that
distamycin has a subtler effect on the E-selectin promoter. Many HMG-like proteins are known to bend the DNA to which they bind (
37
,
38
), whereas binding of distamycin results in a `straightening out' or `unbending'
of DNA (
39
-
41
). We have previously shown that several transcription factors known to activate
the E-selectin promoter also induce DNA bending when bound to their E-selectin promoter sites. Moreover, we have shown that intrinsically
bent DNA sequences can functionally replace factor binding to some of the
crucial elements that activate the E-selectin promoter (
28
). A possibility for distamycin's mode of action is that it alters the geometry
of the E-selectin promoter. We therefore examined how distamycin affects bending of
the E-selectin NF-[kappa]BI and NF-[kappa]BII sites by use of a technique called phasing
analysis (
30
). For this assay, the NF-[kappa]B-containing DNA elements were cloned into a plasmid between
two direct repeats carrying multiple restriction sites (Fig.
8
A). The elements were excised from the vector using one of the set of
restriction enzymes that cut in the repeats. The fragments were radioactively
labelled, incubated with distamycin and migrated in a polyacrylamide gel. As
all fragments have near identical molecular weights and electrostatic charge,
altered migration patterns are indicative of differences in conformation. The
results in Figure
8
B show migration of the NF-[kappa]BI and NF-[kappa]BII DNA fragments incubated in the presence or absence
of distamycin. As judged by its migration, insertion of the NF-[kappa]BII element resulted in a high degree of distortion of the DNA, but
addition of distamycin caused all fragments to run with the same mobility (Fig.
8
B, right panel). In contrast, the NF-[kappa]BI element caused considerably less distortion (Fig.
8
B, left panels), both in the absence and presence of distamycin.
We have recently shown (
11
) that HMG-I(Y) is involved in the regulation of E-selectin gene expression by enhancing NF-[kappa]B binding to three regions in the proximal promoter.
From the known properties of distamycin to bind the minor groove of A/T-rich DNA regions and to block HMG-I(Y) binding to DNA, we hypothesized that this compound would
inhibit the induction of E-selectin expression. Contrary to our expectation, distamycin significantly
extended the duration of maximal E-selectin expression. We show here that distamycin treatment both prolongs
the duration of E-selectin transcription and stabilizes E-selectin mRNA. The stabilization of E-selectin mRNA by distamycin is puzzling, since distamycin is
not known to bind RNA. A possibility is that an E-selectin mRNA degradation pathway exists that itself depends on
transcription and translation. More evidence for such a pathway comes from our
earlier finding (
22
) that the protein synthesis inhibitor cycloheximide also increases the E-selectin mRNA half-life. It would be of interest to investigate distamycin's effect on
other unstable mRNAs.
Our run-on results support distamycin extending E-selectin maximal expression by acting through the protein complex
binding at the NF-[kappa]BII element on the promoter. The three findings consistent with
this mechanism are: (i) DNase I protection footprinting carried out in the
presence of distamycin shows that only the NF-[kappa]BII (-126) site is protected from DNase digestion; (ii) mutation
of NF-[kappa]B sites shows that NF-[kappa]BI is dispensable for a functional distamycin
response, whereas deletion of the NF-[kappa]BII site results in no response at all; (iii) distamycin reduces
bending on the NF-[kappa]BII site, but not the NF-[kappa]BI site.
Our results suggest that distamycin treatment mediates its effect predominantly
(or only) through one site (NF-[kappa]BII) of the E-selectin promoter. This observation suggests that the complex
that binds site II has a distinct role from those binding the other two sites,
possibly because of a different protein composition (
11
,
32
). Furthermore, the sequence of the NF-[kappa]BII site, unlike the NF-[kappa]BI site, differs from the canonical NF-[kappa]B binding sequence consensus (
42
,
43
). The fact that the NF-[kappa]BII element, unlike the NF-[kappa]BI site, is completely conserved between mouse and man
(
44
) is another indication of a special role for this element. It was found
recently that binding of different NF-[kappa]B complexes is determined by the precise sequence of the NF-[kappa]B binding site within the promoter (
45
). The sequence of the NF-[kappa]BII site differs from the consensus sequence at the seventh
nucleotide, where an adenine instead of a pyrimidine residue (consensus
sequence) is found (
45
). This sequence difference may allow a different NF-[kappa]B complex to bind the site, explaining the preferential binding of
distamycin and the subsequent effect on E-selectin expression.
HMG-I(Y) has been described as a specific co-activator of NF-[kappa]B (
45
). Our results may be explained by a dual role of the complex that binds the NF-[kappa]BII site of the E-selectin promoter. A number of examples exist where NF-[kappa]B binding sites play a role in the repression of
gene expression (
46
-
49
). In one example Dorsal, a
Drosophila
rel-family protein, not only activates but also represses transcription (
50
). Recently, a
Drosophila
HMG1-like protein (DSP-1) has been identified that converts Dorsal and NF-[kappa]B from activators into transcriptional repressors (
51
). Thus, HMG-like proteins can determine whether NF-[kappa]B functions as an activator or repressor of transcription,
perhaps by local DNA (un)bending. A mammalian homologue of DSP-1 (as yet unidentified) may be involved in the down-regulation of E-selectin expression by binding at or close to the NF-[kappa]BII site. Distamycin may prolong E-selectin expression by inhibiting the
binding of such a repressor.
An alternative, structure-based hypothesis may explain the effect of distamycin on E-selectin expression. HMG-I(Y) has been thought of as an architectural component of the
NF-[kappa]B-promoter complex. The DNA binding domain of HMG-I(Y) has a predicted planar crescent shape, similar to
distamycin (
52
). Thus distamycin may replace HMG-I(Y) structurally and functionally in the site II complex when it is
normally lost in down-regulation of the promoter, perhaps after phosphorylation. The continuous
presence of distamycin would allow transcription; removal would stop
transcription. These predictions are consistent with our observations that
removal of distamycin results in rapid down-regulation of E-selectin expression. Our finding that other DNA binding compounds
not well known to bend DNA did not affect E-selectin transcription (Fig.
4
) suggests that distamycin's DNA bending capacity plays a role in this
transcriptional on/off switch. It was recently found that distamycin also binds
NF-[kappa]B elements in the HIV promoter (
53
), but in this case transcriptional effects were not explored.
This study provides evidence for the importance of the NF-[kappa]BII site of the promoter in down-regulation of E-selectin expression and suggests HMG-I(Y) DNA bending as a modulator of NF-[kappa]B function. Identifying the
components of the NF-[kappa]BII binding complex and how they change with time following
cytokine induction of the promoter will be important. Clarifying the precise
means by which the E-selectin promoter is down-regulated will also add significant insight into the many other
promoters whose regulation involves NF-[kappa]B-like complexes and why and how these evoke different
transcriptional responses.
REFERENCES
Return