(Received for publication, March 4, 1997, and in revised form, May 23, 1997)
From the Unité de Cancérologie
Expérimentale, U119 INSERM, 27 boulevard Leï Roure,
13009 Marseille, France
Stimulation of highly purified primary T
lymphocytes through CD2 and CD28 adhesion molecules induces a long-term
proliferation, dependent on persistent autocrine secretion of
interleukin 2 (IL-2), high and prolonged expression of inducible
CD25/IL-2 receptor chain (IL-2R
), and secretion of growth
factors such as the granulocyte-macrophage colony-stimulating factor
(GM-CSF). CD28 costimulation appears to activate cytokine gene
expression through conserved
B-related CD28 response (CD28RE) or
cytokine 1 (CK-1) elements in addition to canonical NF-
B-binding
sites. In this report, we assess: 1) the evolution of the expression,
over an 8-day time period, of the Rel/NF-
B family of proteins in
costimulated versus TcR/CD3-stimulated primary T cells; 2)
the impact of changes on the in vitro occupancy of GM-CSF
B and CK-1, as well as IL-2R
B sites; and 3) the differential
regulation of newly synthesized p65 and c-Rel by I
B proteins. We
show that CD2+CD28 stimulation specifically induces, at maximal T cell
proliferation phase, sustained nuclear overexpression of
NFKB2 p52 and c-Rel subunits which might rely on
long-lasting processing of p100 precursor for p52 and increased
neosynthesis of c-Rel. This up-regulation correlates with sustained
occupancy of GM-CSF
B and CK-1 elements by both proteins.
Conversely, these subunits do not appear to bind to the IL-2R
B
site. Costimulation, but not TcR/CD3 stimulation, appears supported by
sustained down-regulation of both I
B
and -
regulators.
Furthermore, contrary to p65, c-Rel appears to display little affinity
for p105, p100 and I
B
regulators.
Specific activation of T lymphocytes is initiated by the
engagement of the clonotypic T cell receptor
(TcR)1 by antigenic peptides
presented by major histocompatibility complex molecules at the surface
of antigen presenting cells. Monoclonal antibodies (mAbs) to TcR or to
the receptor-associated CD3 molecule can mimic effects of antigen
recognition in the presence of monocytes (1). However, proliferation of
highly purified T cells requires a second signal, provided by the
engagement of the T cell accessory molecule CD28 with its cognate
ligand B7 expressed on antigen presenting cells (reviewed in Refs.
2-4). In addition to costimulating the antigen-dependent
pathway, CD28 can act in concert with the CD2 adhesion molecule, which
is present on nearly all T lymphocytes, to up-regulate T cell
responsiveness. As for the CD28 pathway, T cell activation via CD2 can
be mimicked by the in vitro use of mAbs (5, 6). The CD2
pathway, although not physically linked to the CD3·TcR complex, is
metabolically related to it, in that modulation of the receptor complex
abrogates the effects of anti-CD2 antibodies (7, 8). In combination
with CD2, or with CD3 mAb, CD28 mAb induces a high-level, long-lasting, IL-2-dependent and monocyte-independent T cell stimulation
which is sustained by persistent high-level expression of the IL-2
high-affinity receptor (9, 10). T cell proliferation is also associated with the prolonged secretion of high levels of various cytokines known
to be normally synthesized by accessory cells, namely IL-1, CSF-1,
and GM-CSF (11). All these up-regulations correlate with increased
levels of corresponding mRNA pools, resulting from increased mRNA stability and transcriptional activity (9, 12, 13).
CD28 costimulation is thought to activate lymphokine gene transcription
through unique motifs, initially characterized as the CD28 response
element (CD28RE) in the IL-2 promoter, and found conserved within
several lymphokine genes (14, 15). The CD28RE motifs, also termed
cytokine 1 (CK-1), are distinct from, but related to, B elements in
that they bind to, and can be activated by, proteins of the Rel/NF-
B
family (16-19). Remarkably, the IL-2 promoter, as well as several
other lymphokine or growth factor promoters including those of GM-CSF
and CSF-1, contains a consensus
B site in the vicinity of the
CD28RE. Although the enhancer role of this
B motif has been
controversial, it seems that, in normal T cells, the
B and CD28RE
sites are not a redundant pair since mutation of the
B site cannot
be compensated for by a functional CD28RE (20). CD28 costimulation also
activates transcription from the human immunodeficiency virus type 1 long-terminal repeat (21, 22) and the IL-2R
gene promoter (10, 23,
24) through their respective
B elements. The predominant role of
Rel/NF-
B factors in the regulation of the expression of genes
crucial for immune functions has been further demonstrated by studies
of knock-out mice lacking different functional Rel/NF-
B genes
(reviewed in Ref. 25). Interestingly, mice deficient for the c-Rel
protooncogene exhibited defects in production of IL-2, IL-3, GM-CSF,
tumor necrosis factor-
, and interferon-
cytokines, but displayed
unaltered expression of IL-2R
, as well as a number of other cell
surface receptors (26, 27).
In vertebrates, Rel/NF-B proteins are homo- and heterodimers
encoded by a small multigene family including NFKB1
(p50/p105), NFKB2 (p52/p100), RelA (p65),
RelB, and c-Rel genes (reviewed in Refs. 25 and
28-30). Rel-related proteins share a conserved 300-amino acid
amino-terminal domain (Rel homology domain) that encompasses sequences
required for their dimerization, nuclear targeting, and binding to DNA
and I
B regulators. In addition, c-Rel, p65, and RelB all contain
carboxyl-terminal transcriptional transactivation domains. The p50 and
p52 proteins are derived from the NH2-terminal half of p105
and p100 precursors, respectively, by proteolytic cleavage. In most
types of resting cells, the majority of Rel complexes are sequestered
in the cytoplasm as inactive complexes associated with the I
B family
of ankyrin motif-rich inhibitory proteins (reviewed in Refs. 29, 31,
and 32). The ankyrin domains appear to interact with Rel subunit
dimerization domains, thus hindering the nuclear localizing sequence.
I
B proteins include I
B
, -
, -
, -
, and -R proteins.
p105 and p100 proteins, the COOH-terminal half of which also contains
repeats of ankyrin domains, behave as Rel inhibitors as well. The
molecular processes involved in the stimulation-induced release of
Rel/NF-
B proteins have best been defined for I
B
. Activators of
NF-
B induce rapid phosphorylation of I
B
followed by its
ubiquitination and degradation by the proteasome complex. The
Rel/NF-
B/I
B system forms an interregulated network. Thus,
following activation by inducers such as tumor necrosis factor-
,
PMA, or IL-1, resynthesis of I
B
, p105, and p100 proteins is
up-regulated by the NF-
B factor, itself, via
B motifs present in
the promoter of the I
B genes (reviewed in Ref. 31). This ensures the
transient activation of the factor and the replenishment of the
cytoplasmic pools. CD28 costimulation appears to be distinct from the
above stimuli in that it causes a sustained down-regulation of I
B
(33, 34). It has been assumed that this continued down-regulation leads
to CD28 enhanced nuclear translocation of c-Rel (35), which, in turn,
causes sustained up-regulation of IL-2 gene expression. However, there has been no search for evidence of a physical association between I
B
and c-Rel to validate this hypothesis.
In this study, using a combination of immunoblotting, metabolic
labeling/coimmunoprecipitation, and EMSA analyses with a large panel of
anti-Rel/NF-B subunit antibodies, we have defined the major Rel
complexes operating in resting human primary T lymphocytes, as well as
in cells activated via CD2+CD28. We focused on the modifications
specifically induced by costimulation, in comparison to those observed
after CD3-mediated stimulation, and examined the consequences of these
modifications on the in vitro binding activity of the
corresponding complexes to the IL-2R
and GM-CSF
B or CK-1
sites.
T cell purification from
peripheral blood and activation were performed as described previously
(10). Primary T cells were maintained in RPMI, 10% fetal calf serum.
Stimulations were performed with the following mAbs, used alone or in
combination, at saturating concentrations. Anti-CD2 mAb 39C1.5 (rat
IgG2a) and 6F10.3 (mouse IgG1) were used as
purified mAbs at 10 µg/ml each. Anti-CD28 248 (mouse IgM) and
anti-CD3 289 (mouse IgG2a) were obtained from Dr. A. Moretta (Cancer
Institute, Genova, Italy) and were used as ascites fluid (1/400
dilution) or as purified mAb (10 µg/ml), respectively. CD3 mAb was
used coated onto Petri tissue culture dishes (CD3c). T cell activation
was controlled by proliferation assays and CD25/IL-2R
expression.
Cytosolic and nuclear extracts were prepared as described previously (10). Briefly cells were harvested and washed in cold Tris-buffered saline, then resuspended in 0.4 ml (per 107 cells) of buffer A (10 mM Hepes, pH 7.8, 10 mM KCl, 2 mM MgCl2, 1 mM dithiothreitol, 0.1 mM EDTA, 0.1 mM phenylmethylsulfonyl fluoride) supplemented with leupeptin (10 µg/ml; Sigma), and incubated on ice for 15 min. 25 µl of 10% Nonidet P-40 solution (Sigma) were next added, and cells were mixed vigorously for 15 s and then centrifuged (13,000 rpm, 15 s). Supernatants corresponding to the cytosolic fraction were used directly for immunoblotting. Pelleted nuclei were resuspended in 50 µl of buffer C (50 mM Hepes, pH 7.8, 50 mM KCl, 300 mM NaCl, 0.1 mM EDTA, 1 mM dithiothreitol, 0.1 mM phenylmethylsulfonyl, 10% (v/v) glycerol). After mixing for 20 min, the nuclei were centrifuged for 5 min at 13,000 rpm, and the supernatants containing the nuclear proteins were transferred to clean tubes. They were used, as such, for Western and EMSA analyses. Protein concentration was measured with a commercial kit (Bio-Rad) by the method of Bradford (36).
AntiseraPolyclonal rabbit antisera were raised against
Rel/NF-B peptides. They include: an anti-p52/100
NH2-terminal peptide (amino acids 4 to 16), that reacted in
immunoblotting, immunoprecipitation, and EMSA; an anti-p50 nuclear
localizing sequence peptide (amino acids 351 to 364), that reacted in
immunoprecipitation and EMSA; an anti-p65 COOH-terminal peptide (amino
acids 531 to 543), that reacted in immunoblotting, immunoprecipitation,
and EMSA; and an anti-c-Rel peptide directed against both an internal
(amino acids 151 to 165) and a COOH-terminal (amino acids 606 to 619) amino acid stretch, that reacted in immunoblotting and
immunoprecipitation. Antisera against p50/105 NH2-terminal
peptide were generous gifts from Michael Karin (San Diego School of
Medicine, La Jolla, CA (37) and John Hiscott, Lady Davis institute,
Montreal, Canada). Anti-c-Rel peptide 1135 was a kind gift from Nancy
Rice (National Cancer Institute, Frederick, MD) (38). Polyclonal rabbit
antisera p65H and I
BH were raised against recombinant p65 and I
B
(6X His, Novagen, R & D System) and were immunoprecipitating. Purified anti-I
B
COOH terminus (sc-371), anti-Sp1 (sc-59X) and -Sp3
(sc-644) antisera were purchased from Santa Cruz Biotechnology.
Cells were washed twice,
prepulsed for 45 min in methionine (Met) and cysteine (Cys)-free RPMI,
10% dialyzed fetal calf serum, and pulsed for 15 min in the above
medium supplemented with a 1:1 mixture of [35S]Met and
-Cys (2 mCi/ml/5.107 cells). Labeling was stopped by a
10-fold dilution of the samples in complete RPMI medium supplemented
with 4 mM Met and Cys. Labeled cells were lysed (whole cell
lysate) in modified RIPA lysis buffer R (25 mM Tris, pH
7.4, 50 mM NaCl, 0.5% deoxycholate, 0.5% Nonidet P-40,
0.1% SDS, and 2 mM EDTA) supplemented with protease (10 µg/ml pepstatin, antipain, and aprotinin and 0.5 mM
phenylmethylsulfonyl fluoride) and phosphatase inhibitors (100 µM orthovanadate and molybdate). Cell extracts were first
treated with protein A-Sepharose (50 µl sedimented beads) and normal
rabbit serum (1/500 dilution), for 1 h at 4 °C. Precleared
supernatants were treated with immunoprecipitating antibodies (1/500
dilution) overnight at 4 °C, followed by protein A-Sepharose (50 µl, 1 h at 4 °C). Immunoprecipitates were then washed four
times in complete extraction buffer. For reprecipitations, washed
immunoprecipitates were boiled in dissociating buffer R (buffer R
lacking SDS, supplemented with 1% SDS and 0.5% mercaptoethanol) for 3 min. Samples were next diluted 10 times in complete buffer R
supplemented with proteases and phosphatase inhibitors, and rapidly
treated with protein A-Sepharose (50 µl, 30 min at 4 °C) to get
rid of the first immunoprecipitating antibody. Supernatants were then
treated with the second antibody (5 h at 4 °C), followed by protein
A-Sepharose (50 µl, 1 h at 4 °C). Proteins bound to washed
protein A-Sepharose beads were eluted and reduced by warming the
samples at 95 °C for 3 min in sample buffer containing 5% mercaptoethanol. Samples were next cooled to room temperature before
alkylation with 5 mM iodoacetamide and fractionation by SDS-polyacrylamide gel electrophoresis on a 7.5 to 12% acrylamide gradient gel. 14C-Methylated molecular weight markers
(Amersham) were run in parallel. After electrophoresis, gels were
treated with an intensifying solution (Amplify, Amersham) for
fluorography. For quantifications, dried gels were analyzed with a
Molecular ImagerTM PhosphorImaging System (Bio-Rad).
Cell extracts were boiled in reducing buffer for 3 min before loading onto a 10% SDS-polyacrylamide gel. An equal amount of proteins (15 µg) were loaded per well. Following electrophoresis, the gels were transferred onto an Immobilon-P membrane (Millipore) at 50 V for 18 h using a transblot apparatus (Bio-Rad). Residual binding sites were blocked by incubation for 2 h at room temperature in phosphate-buffered saline supplemented with 5% nonfat dry milk. After incubation with antiserum, filters were washed six times with phosphate-buffered saline, 0.1% Tween for 5 min at room temperature. Then peroxidase-labeled anti-rabbit serum from donkey preformed complex was added (1/10,000 dilution) and incubated at room temperature for 1 h. After washing as described above, filters were reacted 1 min with chemiluminescent substrate luminol (ECL kit; Amersham) and revealed by brief exposure (15 to 60 s) using autoradiography films (Kodak X-Omat). For quantification, films were scanned by densitometry using a BioImage whole band analyzer (Millipore Co.).
OligonucleotidesThe sequences of the oligonucleotides were
as follows (B or
B-like (CK-1) motifs are in boldface; the
mutated (m) nucleotides used to disrupt the binding sites are
underlined): IL-2R
B, CAACGGCAGGGGAATCTCCCTCTCCTT;
IL-2R
Bm, CAACGGCAGCTCAATCTCCCTCTCCTT; GM-CSF
B/GC, GGGAGGCGGGGGAACTACCTGAGT; GM-CSF
B/GCm, GGGAGGCGGCTCAACTACCTGAGT; GM-CSF CK-1,
AACTGTGGAATCTCCTGGCCC; GM-CSF CK-1m,
AACTGCTCAATCTCCTGGCCC; SV40 Sp1/GC-box,
CGATGGGCGGAGTTAGGGACGGGA.
Oligonucleotides were
endlabeled using [-32P]ATP and polynucleotide
kinase, annealed, and purified on a chroma-spin column (CLONTECH) for isolation of double-stranded probes.
Reactions with equal amounts of nuclear extracts (1 µg/reaction) were
performed in a 20-µl final volume in buffer C containing 50,000 cpm
probe, 0.25 µg of polydeoxyinosinic-deoxycytidylic acid, and 20 µg
of bovine serum albumin, for 30 min at room temperature. Unlabeled double-stranded competitor, specific antiserum, or recombinant protein
was preincubated with cell extracts 20 min prior to addition of the
probe. Binding reaction mixtures were loaded on a 5% nondenaturing polyacrylamide gel in 0.25 × TBE (1 × TBE, 89 mM Tris, 89 mM boric acid, 1 mM
EDTA). After electrophoresis, dried gels were exposed at
80 °C or
room temperature, with or without intensifying screen, according to the
intensity of the radioactive signals.
We previously observed that the stimulation of
primary T lymphocytes with CD2+CD28 mAbs induces a high level of
proliferation and IL-2 secretion as compared with stimulation by the
association of CD28 plus coated-CD3 (CD3c) mAbs (10). Both
proliferative responses lasted for more than 14 days. The two combined
stimuli induced high level expression of IL-2R chain, but CD2+CD28
activation induced the most sustained phenotype. In sharp contrast, CD3
alone induced a much weaker and shorter proliferative response.
Costimulation of primary T lymphocytes by CD2+CD28 mAbs was chosen,
therefore, as a model for activation and long-term maintenance of
proliferation. As an initial insight into the physiological role of
NF-
B, we demonstrated that the increased expression of IL-2R
transcription is associated with the long-term persistence in the
nucleus of two inducible IL-2R
B motif binding complexes, namely
NF-
B p50/p65 as well as an ill-defined protein-DNA complex, and of constitutive (p50)2 homodimers. NF-
B up-regulation was
correlated with an increase in the level of the nuclear p50 subunit
(23). Using an enlarged set of serological tools, we have now extended our analysis to all four human Rel proteins and have compared their
nuclear levels in the course of CD2+CD28 versus CD3
stimulation.
Immunoblots using polyclonal antipeptides directed against p52/100N,
p50/105N, p65C, and c-Rel are shown in Fig.
1A. In addition to
constitutive nuclear expression of the p50 subunit, a low level of p65
and c-Rel subunit(s) was detected in nuclei from unstimulated T cells.
This low level of detection, which as previously been observed in some
cases (39), and not in others (17, 35), might reflect the presence of
non-truly resting cells in some purified T cell populations. Activation
of T lymphocytes through either CD2+CD28 or CD3 led to an increase in
the level of all of the subunits peaking at 5 to 16 h for p65 and
96 h for p52, p50, and c-Rel. However, overexpression was more
sustained after costimulation. This is particularly striking for p52
and c-Rel, the amounts of which were still strongly increased at day 8. Moreover, maximal induction of these two subunits was higher than that
obtained after CD3 stimulation alone. To assess whether the general
decrease of Rel nuclear proteins in long-term CD3-stimulated
lymphocytes reflected some mechanism of down-regulation or an overall
decrease in the rate of the cellular metabolism, each corresponding
cytosolic pool was examined. As shown in Fig. 1B, sustained
overexpression of p100 and p105 precursors was observed, while the
level of p65 subunit remained constant. These profiles favor a model by
which nuclear down-expression is post-translationally regulated. It has
been proposed that CD28 costimulation induces sustained down-regulation of IB
(33) and I
B
(34, 40), leading to enhanced
translocation of c-Rel. Our data support this hypothesis (Fig.
1C, lanes 1-7). Conversely, the transient
character of nuclear c-Rel in CD3-stimulated cells could rely on the
replenishment of I
Bs cytosolic pools, as suggested by the unchanged
level of both inhibitors compared with costimulation by CD2+CD28 (Fig.
1C, lanes 8-13). Examination of cytosolic profiles in
CD3-stimulated cells (Fig. 1B) revealed, however, that the
premature decay of nuclear c-Rel results from decreased neosynthesis
rather than increased cytosolic retention. Finally, comparison of
cytosolic and nuclear profiles from both CD3- and CD2+CD28-stimulated
cells indicated an increase with a similar time course of c-Rel and p52
suggesting that their up-regulation is neosynthesis-mediated (this
figure and data not shown). This possibility is reinforced by our
finding of a correlated increase in protein neosynthesis, as assessed
by pulse-radiolabeling (data not shown).
Increased nuclear expression of p50, p65, and c-Rel proteins after
costimulation with anti-CD28 mAb has been reported by others in a
similar model system (17, 35). However, in the previous studies, the
time period analyzed did not exceed 21 h and, therefore, precluded
a comparison between Rel protein up-regulation and the onset of active
proliferation of the cells, which occurs later. The increased detection
of nuclear p52, p50, and c-Rel at a time when cells started dividing,
led us perform this comparison. The rate of proliferation of T
lymphocytes was measured by [3H]thymidine incorporation
and compared with the amount of Rel/NF-B subunits quantified by
densitometry scanning of the gels. Fig. 1D shows a striking
correlation between maximal T cell proliferation and maximal p52, p50,
and c-Rel nuclear expression.
Given the time course of subunit
modifications detected in nuclei from T lymphocytes stimulated by
anti-CD2+CD28 mAbs, we analyzed Rel protein interactions in resting or
proliferating T cells. Steady-state labeling of the cells, followed by
extraction of cytosolic and nuclear fractions and exhaustive
immunoprecipitations with various antisera directed at Rel/IB
subunits revealed many changes upon activation (data not shown). Hence,
in resting cells, the vast majority of Rel proteins detected
corresponded to p65 subunits retained in the cytosol through
interactions with p105, p100, and I
B proteins. Some nuclear Rel
proteins became detectable following activation, likely corresponding
to p50 and c-Rel homodimers and p50/p65 heterodimers. However, even in
activated T cells, a major proportion of the factors appeared
sequestered in the cytosol.
To validate our first conclusions on the composition of Rel/NF-B
complexes in CD2+CD28-activated T lymphocytes and evaluate the
contribution of I
B proteins, we performed a double
immunoprecipitation analysis. Activated T cells, at day 6, were
pulse-labeled with [35S]methionine and -cysteine for 15 min to detect newly formed interactions. A whole cell lysate was
prepared using a modified RIPA solution of lower stringency. It was
immunoprecipitated by either pooled anti-p65 (C and H) or anti-c-Rel
antibodies (IP1). Washed precipitates were dissociated from protein
A-Sepharose beads by boiling in 1% SDS and reprecipitated by all of
the other anti-subunits reagents (IP2). Results in Fig.
2 indicate that some newly synthesized p65 subunits indeed associate with either p100 or p105 as well as, to a
large extent, with I
B
proteins (Fig. 2, lanes 3, 4, and 6, respectively). Similarly to p65, subpopulations of
newly synthesized c-Rel proteins appeared to interact with either of the three I
B proteins (Fig. 2, lanes 9, 10, and
12), although little association was detected, in particular
with p105 and I
B
, compared with p65. The slightly higher relative
labeling of c-Rel, due to greater (1.4-fold) methionine and cysteine
content compared with p65, cannot alone account for the major
differences in codetection. Rather, the present data suggest that newly
synthesized p65 proteins are predominantly sequestrated by p105 and
I
B
inhibitors, whereas c-Rel subunits display a much lower
affinity for these proteins. Similar conclusions could be drawn from
the IP1 profiles of resting T lymphocytes (result not shown), although
double immunoprecipitations could not be carried out because of
limiting c-Rel detection. Results in Fig. 2 also provide evidence for a
subpopulation of p65/c-Rel heterodimers, as shown by the anti-c-Rel or
anti-p65 IP2 profiles (Fig. 2, lanes 5 and 11).
Since p65 protein was reproducibly detected in the anti-c-Rel IP2
precipitate, the two subunits might display a high affinity for each
other and thus, either incompletely dissociate in SDS or reassociate in
the course of the IP2.
Costimulation through CD2+CD28 Induces Sustained in Vitro Occupancy of GM-CSF
The functional consequences of the major changes in the
composition and regulation of Rel complexes were analyzed at the level of Rel protein DNA binding to GM-CSF B and CK-1, as well as IL-2R
B sites. Results were compared with those obtained after single stimulation of the cells with coated anti-CD3 antibodies.
B-binding proteins present in nuclear extracts were detected by EMSA with relevant synthetic oligonucleotide probes. Note that the GM-CSF
B
probe encompasses a consensus decameric
B site overlapping GC-rich
and CK-2 (CLE-2) elements and, therefore, is a potential target for
Rel/NF-
B, Sp1, and CREB·activating transcription factor proteins
(41). However, our EMSA conditions, set-up to detect Rel/NF-
B-DNA
complexes, were not optimized for CREB·activating transcription
factor complex formation (42).
As illustrated in Fig. 3, clear
differences were observed in the occupancy of the GM-CSF sites by
nuclear proteins from CD2+CD28- versus CD3-stimulated cells.
EMSAs performed with the B/GC motif and nuclear extracts from
unstimulated T lymphocytes gave three major retarded bands (A,
C, and D). Both stimuli led to the early appearance (30 min) of a fourth DNA-protein complex (B). However, this
induced binding activity was more sustained after costimulation. No
strong occupancy of the CK-1 site was detected in resting cells. A
major retarded band (E), together with a weaker band
(F), was detected 5 h after activation by both stimuli.
However, whereas this DNA binding activity appeared transient in the
case of CD3 stimulation, a second wave was observed (from day 4 to 6)
after CD2+CD28 costimulation. Conversely, as observed earlier (10), both stimuli induced a comparable long-lasting binding to the IL-2R
B site. Hence, a similar pattern of constitutive (A and D) and inducible (B) retarded bands was
obtained.
We performed competition experiments with sets of specific, wild-type
and mutated, or nonspecific, unlabeled oligonucleotides to determine
the DNA binding specificity of the complexes binding to each of the
probes. As shown in Fig. 4, upper
panel, a 100-fold excess of unlabeled oligonucleotides
corresponding to wild-type GM-CSF B/GC (lane 2) or
IL-2R
B (lane 4) totally abrogated bands B and D,
whereas the intensities of bands A and C were only partially
diminished. A similar excess of GM-CSF
B/GC oligonucleotide mutated
in the
B motif had no effect on the four bands (Fig. 4, lane
3). An Sp1/GC-box oligonucleotide totally abrogated bands A and C
(Fig. 4, lane 5). Altogether, these data indicate that bands
B and D correspond to
B-binding complexes and bands A and C to
GC-box binding ones. They suggest, in addition, that the efficiency of
binding of the latter complexes might in part be influenced by nearby
B binding complexes. Results in the middle panel show
that an excess of unlabeled oligonucleotides corresponding to wild-type
GM-CSF CK-1 (lane 2) or IL-2R
(lane 4) motifs
totally abrogated bands E and F, whereas CK-1 site mutated in the
B-like sequence (lane 2) or Sp1 site (lane 5)
had no effect. Bands A and B correspond, therefore, to complexes
specific for the
B-like sequence. Finally, results in the lower
panel show that an excess of wild-type IL-2R
(lane 2) or
GM-CSF
B/GC (lane 4) oligonucleotides totally suppressed
bands B and D and partially diminished band A, whereas IL-2R
oligonucleotide mutated in the
B consensus had no effect (lane
3). Of note is the fact that, although the IL-2R
motif competed
for protein binding to GM-CSF CK-1 probe, GM-CSF CK-1 motif did not
compete for the binding to IL-2R
probe (result not shown). This
asymmetry is likely due to the weak affinity of Rel proteins for CK-1
sites (16). Unexpectedly, although the IL-2R
B site does not
contain a prototypical GC-box, our Sp1 consensus oligonucleotide
totally abrogated band A (lane 5), but left bands B and C
intact. Altogether, these inhibition patterns suggest that complexes B
and D contain
B-specific proteins and complex A, proteins of the
GC-box binding family. We found, in addition, that recombinant Sp1
proteins comigrated with band A and, therefore, directly recognize the
IL-2R
motif (lane 6).
Costimulation through CD2+CD28 Induces the Specific in Vitro Occupancy of GM-CSF
DNA-protein complexes were next characterized by
including different anti-NF-B or -Sp protein antisera in binding
reaction mixtures. Fig. 5 reports the
analysis of the specific complexes formed with a GM-CSF
B probe and
nuclear extracts from unstimulated T lymphocytes, or cells in the early
(30 min) or late (day 6) activation phase following either CD2+CD28 or
CD3 triggering. Constitutive complexes D and A supershifted in the
presence of anti-p50 (Fig. 5, lane 2, all panels) and
anti-Sp1 (Fig. 5, lane 10, all panels) antisera,
respectively, but were unaffected by all of the other anti-Rel/NF-
B
reagents (Fig. 5, lanes 4, 7, and 8, panels
1-5). Therefore, they likely contain p50 homodimers and Sp1
protein, respectively. Constitutive complex C contains Sp3 proteins
since it was abrogated in the presence of anti-Sp3 antiserum (data not
shown). From 30 min on to day 6 following activation by both stimuli,
inducible band B was diminished, or abrogated and supershifted, with
anti-p50 or anti-p65 antisera, respectively (Fig. 5, lanes 2 and 4, panels 2-5) and, thus, contained p50/p65
heterodimers. In addition, band B was greatly diminished when
anti-c-Rel was added to nuclear extracts of costimulated cells, at day
6 of activation (Fig. 5, lane 8, panel 4), suggesting the
formation of c-Rel containing DNA-binding complexes in the proliferation phase. A parallel induction of p52 containing complexes is supported by the supershift of band C
upon addition of anti-p52 antibody (Fig. 5, lane 7, panel 4).
Proteins specifically binding to the GM-CSF CK-1 site are analyzed in
Fig. 6. Major complex E, induced after a
5-h stimulation of the cells with either anti-CD2+CD28 (panel
1) or -CD3c (panel 2) antibodies, was supershifted or
greatly diminished when anti-p52 (Fig. 6, lane 3) or
anti-c-Rel (Fig. 6, lane 4) antibodies were added to the
cell extracts. None of the other anti-Rel/NF-B reagents affected
complex E migration, suggesting that it mostly contains p52/c-Rel
heterodimers. Similar conclusions could be drawn on complex E formed in
the presence of nuclear extracts from costimulated cells at day 6 (Fig.
6, panel 3, lanes 3 and 4). Although minor complex F contains
B-binding proteins (Fig. 4), none of our antisera had a clear effect on its migration. More reagents, recognizing a
larger set of epitopes, might thus be needed to delineate its composition.
Proteins specifically binding to the IL-2R
B site were next
analyzed (Fig. 7). Constitutive complex
A, retarded similarly to Sp1 complex (Fig. 4), was supershifted when
nuclear extracts were mixed with anti-Sp1 antibody and indeed contained
Sp1 proteins (data not shown). Constitutive complex D was supershifted
by anti-p50 but not by other anti-Rel/NF-
B reagents (Fig. 7,
panels 1-5, compare lane 2 to lanes 4, 7, and 8), and, thus, likely contained p50 homodimers.
From 30 min to 6 days following activation by both types of stimuli,
inducible complex B was abrogated by both anti-p50 and anti-p65
antibodies (Fig. 7, panels 1-5, lanes 2 and 4,
respectively) and thus contained p50/p65 heterodimers. Unlike the
GM-CSF
B site, the IL-2R
site did not seem to bind significant
amounts of c-Rel or NFKB2 p52 proteins. This was unexpected in view of
earlier reports on c-Rel binding to the IL-2R
B site (43, 44) and
by the fact that this motif competed for the binding of p52/c-Rel
complexes to the GM-CSF CK-1 probe (Fig. 4, middle panel).
The latter discrepancy suggests that the IL-2R
probe can bind minor
undetected levels of p52/c-Rel proteins and displace, when in large
excess, their binding to the low affinity CK-1 motif. Table
I summarizes the composition of the
DNA-protein complexes.
|
Activators of Rel family factors
are known to trigger an early nuclear translocation of NF-B p50/p65
heterodimers from pre-existing cytosolic pools. The present EMSA
analyses confirm that stimulation of primary human T cells through
either TcR/CD3 or CD2+CD28 induces the rapid activation of this complex
(23, 35, 39). It appears likely, as previously suggested by others
(39), that these activation pathways only mobilize a minor pool of
factors undetected in our immunoblot analysis. Also, we had no evidence
for early acceleration of c-Rel translocation upon costimulation, as
reported for Jurkat T cells stimulated with PMA plus anti-CD28 (33).
These differences may relate to cell type or to the fact that anti-CD2
plus anti-CD28 costimuli are weaker c-Rel inducers than PMA plus
anti-CD28 (35). Strikingly, all significant modifications, except
RelA(p65) detection which peaked at 5 to 16 h, occurred in the
late phase of activation. Hence, p50, p52, and c-Rel nuclear levels
were maximal at day 4 after single or costimulation. These increases
correlated with a similar delayed increase of cytosolic p105, p100, and
c-Rel pools, suggesting that the up-regulation largely occurs via
neosynthesis. Conversely, as described previously (35), no significant
change in the level of cytosolic p65 was observed following either form of stimulation favoring a model of post-translational modification. These conclusions are reinforced by our finding of an increased rate of
biosynthetic labeling of p105, p100, and c-Rel compared with p65 (data
not shown). Increased neosynthesis of p50/105, p52/100, and c-Rel might
result from increased transcription of NFKB1,
NFKB2, and REL genes, which, unlike that encoding
p65, are potential targets for positive Rel/NF-
B autoregulation
(45-48). Late autoregulated transcription is probably indirectly
induced following either form of stimulation, as reported for the
TcR/CD3-mediated delayed increase in p50 and c-Rel mRNA levels
through autocrine secretion of the NF-
B inducer, tumor necrosis
factor-
(39, 49). Importantly, our data show that costimulation
induced a more sustained nuclear expression of all Rel proteins than
TcR/CD3-mediated stimulation. We conclude from comparison of cytosolic
and nuclear pools that long-term up-regulation of Rel protein nuclear
expression relies on several mechanisms. Sustained c-Rel nuclear
expression appears to be supported mostly by sustained neosynthesis and
might require cooperation of transcription factors that are
specifically induced, either directly or indirectly, via CD28
costimulation. Alternatively, sustained c-Rel neosynthesis may result
from CD28-induced stabilization of its mRNA, as observed for
interleukin mRNAs (9, 12). In contrast, sustained p50, p52, and p65
nuclear expression appears to rely on post-translational regulation
which is not maintained in CD3-stimulated cells.
IB
and I
B
are two major cytoplasmic
inhibitors which exhibit regulatory properties toward p65 and c-Rel.
I
B
degradation appears to be induced by most stimuli and
regulates the immediate activation of NF-
B. In contrast, I
B
degradation might require signals leading to persistent activation,
such as lipopolysaccharide and IL-1 (50). The CD28 costimulatory signal
has been shown to augment the activation of Rel factors by enhancing
the degradation of I
B
, as well as promoting a rapid degradation
of I
B
(34). We show here that, similar to I
B
(33), I
B
expression is also down-regulated for up to a week following CD28
costimulation. Two recent reports suggest that I
B
, rather than
I
B
, is the regulator of c-Rel nuclear translocation in T cells
(51, 52). Our data on the physical interactions of newly synthesized
p65 and c-Rel proteins with Rel/I
B partners fit with this
possibility. Thus, contrary to p65 which strongly interacted with
I
B
, c-Rel appeared to display little affinity for this inhibitor.
In addition, c-Rel, unlike p65, only weakly interacted with the p105 or
p100 precursors. Hence, similar to I
B
and -
, these inhibitors
seem to mediate Rel/NF-
B cytosolic retention in a subunit-, as well as a cell- and stimulus-specific manner (53, 54).
The CD28 pathway might specifically contribute to
Rel-mediated gene activation by, nonexclusively: 1) amplifying or
sustaining the binding of Rel/NF-B factors induced by TcR/CD3
stimulation; 2) modifying the composition of Rel/NF-
B dimers that
bind to sites targeted by TcR/CD3-derived signals; and 3) inducing the binding of new Rel/NF-
B factors to new
B (or
B-related) sites. The latter possibility was supported by the identification, in the
promoter of the IL-2 gene, of the
B-like CD28-responsive element
(CD28RE) (14-16) which involves CD28-up-regulated c-Rel protein in
addition to NF-
B subunits (17). The unique relevance of CD28RE to
the CD28 pathway has been questioned, however, by the finding that the
formation of the CD28-responsive complex was not strictly induced by
CD28 costimulation (35, 55) and, that the alteration of CD28RE did not
abrogate the increased IL-2 transcription induced by CD28 natural
ligands (20). Furthermore, recent independent data have brought to
light uncertainties on the composition of CD28RC by providing evidence
that NFAT, but not Rel/NF-
B proteins, bind to CD28RE upon
up-regulation of T cells via CD28 or via Tax protein of human T cell
leukemia virus type 1 (56, 57). The IL-2R
B site might be
differentially occupied by Rel proteins upon single stimulation through
TcR/CD3 or costimulation through CD28, as it was reported to mediate
the activation of the IL-2R
gene promoter via both NF-
B (10) and c-Rel (43, 44). However, one can question the implication of c-Rel,
since IL-2R
is normally expressed on T cells from c-Rel deficient
mice (27). To clarify some of these issues, we examined the
consequences of TcR/CD3- or CD2+CD28-induced Rel subunit modifications on the in vitro binding of these proteins to the
B and
B-like (CK-1) sites of the GM-CSF promoter (16, 41, 58), which resemble the
B/CD28RE sites of the IL-2 promoter. We also
reinvestigated Rel protein binding to the
B site of the IL-2R
promoter. Our data indicate that in resting primary T lymphocytes, the
GM-CSF
B site is essentially bound by p50 homodimers, similar to the IL-2R
B site (Refs. 10 and 23, and this study). This finding is
in line with the silencing role of p50 homodimers (24, 44, 59). Both
TcR/CD3 and CD2+CD28 stimulation triggered the binding of NF-
B
p50/p65 heterodimers to GM-CSF
B sites, from immediate (30 min) to
late (day 6) activation phases. Only CD2+CD28 costimulation enabled
binding of c-Rel and p52 proteins (likely as c-Rel/p52 and c-Rel/p65
heterodimers) at the late phase. CD2+CD28 long-term induction of Rel
binding to the GM-CSF
B site resembles the induction mediated by
HTLV-1 Tax protein. Hence, whereas stimulation of Jurkat T cells by
PMA-Ca2+ ionophore only induced binding of p65 to this
site, costimulation via Tax activated both p65 and c-Rel binding (41).
Our results are in contrast with those that demonstrate the absence of
c-Rel binding to the GM-CSF
B site after costimulation via CD28.
(60). This discrepancy might reside in the origin of the cell,
i.e. tumor cell line versus primary T cells, or,
most importantly, the differences in time points of analysis,
i.e. few hours versus several days
post-stimulation. Unlike GM-CSF
B sites and similar to the related
CD28RE site, the GM-CSF CK-1 site is unbound in resting T lymphocytes,
a finding relevant to the lack of affinity of p50 homodimers for
non-consensus
B-related sequences (17, 61). As expected for a
CD28RE-related site (19), the GM-CSF CK-1 site did not bind NF-
B
heterodimers from either TcR/CD3- or CD2+CD28-stimulated T cells.
Strikingly, both types of stimulation initiated an early wave of
binding (detected at 5 h time point) of c-Rel and p52 proteins
(likely as c-Rel/p52 heterodimers) which correlates in time with, and
might thus reflect, the transient autoregulated mRNA increases that
we observed (data not shown). Here again, EMSAs appear to enable the
detection of Rel nuclear changes that were not evident by
immunoblotting. Costimulation, but not TcR/CD3 stimulation, induced a
second wave of binding of c-Rel and p52 in late activation phase (day 6 time point). Altogether, our data suggest that costimulation of primary
T cells via CD28 specifically sustains the nuclear expression of both proteins and induce their long-term binding to both GM-CSF
B and
CK-I sites. The implication of c-Rel in GM-CSF gene up-regulation has
best been suggested by the lack of secretion of this growth factor in
Rel
/
T cells costimulated via CD28 (27). Little
information, however, is available on a possible up-regulatory role of
p52 protein. Although such a role should be substantiated by
transcriptional activation assays, our finding that, similar to c-Rel,
p52 nuclear overexpression parallels T cell proliferation, suggests
that both proteins are involved in the up-regulation of growth
regulating interleukin genes. In line with this possibility, p52
(NFKB2/lyt-10) has been implicated in proliferative diseases
(62) and is overproduced in Tax expressing transformed cells (63).
RelA(p65) has been reported as a strong transcriptional activator of
CD28RE within the IL-2 promoter (18) and was shown to bind to the
GM-CSF CK-1 site upon CD28-mediated costimulation (60). We had no
indication on the binding of p65 to GM-CSF CK-1 site in our activation
model, although a serological bias of detection cannot be ruled out. Thus, we could not elucidate the composition of the faster migrating inducible CK-1 complex, although competition assays indicated that it
contains
B-binding proteins. Unlike the case of GM-CSF
B and CK-1
sites, we did not detect any specific effect of CD2+CD28 costimulation
on Rel protein binding to the IL-2R
B site. As previously
observed (10), both TcR/CD3 and costimulation induced an immediate and
long-lasting binding of p50/p65 heterodimers, but neither significant
c-Rel, nor p52 binding was detected. This contrasts with the reported
activation by c-Rel of the IL-2R
B promoter in cotransfected
cells (43, 44), but fits with the finding that Rel
/
mice express normal amounts of surface of IL-2R
genes (27).
The GM-CSF B site overlaps a poly(GC)-rich element, which is a
potential target for GC-binding proteins. It was shown that within the
murine promoter, the GC-box binds predominantly and constitutively Sp1
factor and is required for full stimulation of transcription elicited
by PMA/A23187 treatment (58). In this report, we found a similar
constitutive binding of Sp1 factor to the human
B/GC probe and
detected, in addition, the binding of Sp3 at the resting and
nonproliferating phase (data no shown). Given the reported repressor
role of Sp3 (64), we hypothesize that a balance between Sp3 and Sp1
regulates the silencing or the activation of the GM-CSF promoter in
conjunction with Rel factors.2 Surprisingly, our
data also show constitutive and direct binding of Sp1 factor to the
IL-2R
B probe, although this probe does not contain a consensus
GC-box. Our preliminary observations suggest that, similar to the
binding of serum response factor to the adjacent SRE/CArG box (not
included in our probe) (24), the binding of Sp1 is impeded by the
binding of p50 homodimers (data not shown). It has also been reported
that Sp1 can repress IL-2R
gene transcription (44), whereas we
failed to detect any occupancy at the IL-2R
promoter Sp1-binding
site by genomic in vivo footprinting (24). Hence, the role
of Sp1 in the regulation of the IL-2R
gene is far from being
elucidated. We are currently testing whether the Sp1 factor interacts
physically and cooperates functionally with NF-
B and serum response
factors to promote the transcription of the IL-2R
gene.
We thank C. Mawas and W. Hempel for critical reading of the manuscript.