ABSTRACT
NeP1 binds to the F1 silencer element of the chicken lysozyme gene and, in the
presence of TR, v-ERBA or RAR, synergistically represses transcriptional activity. This repression involves a silencing mechanism acting independently of the relative promoter position. Here we show that NeP1
alone can induce a significant directed bend on DNA. The chicken homologue of
human NeP1, CTCF, shows identical binding and bending properties. In contrast,
the isolated DNA binding domain of CTCF efficiently binds DNA, but fails to
confer bending. Similarly, the TR-RXR hetero- or homodimer, binding adjacent to NeP1 at the F2 sequence, do not
show significant DNA bending. The binding of the T3 ligand to TR changes
neither the magnitude nor the direction of the NeP1 induced bend. However, when
all factors are bound simultaneously as a quaternary complex, the TR-RXR heterodimer changes the location of the bend center, the flexure
angle and the bending direction.
The -2.4 kb silencer of the chicken lysozyme gene is inactive in mature,
lysozyme expressing macrophages, and is active in all other cell types tested.
This activity correlates with the presence of a DNase I-hypersensitive site in the chromatin (
1
). The silencer DNA consists of two protein binding sites that are both required
for full functional activity (
2
). One site is bound by NeP1, whereas the second site is bound by the thyroid
hormone receptor (TR). These silencer modules are termed F1 and F2
respectively, and can repress gene activity independently from each other (
2
,
3
). The repression is increased synergistically when both modules are bound by
their respective factors. NeP1 binds as a monomer to F1 (
3
) and TR binds as a homodimer or a heterodimer with the retinoid-x-receptor (RXR) to F2 (
4
,
5
). Synergistic repression is converted to synergistic induction in the presence
of thyroid hormone (T3) (
2
).
Binding of NeP1 to DNA is characterized by an
in vitro
footprint region of ~50 bp interrupted by a DNase I-hypersensitive site (
2
,
3
). Therefore, we wondered whether such a long stretch of DNA might be bent by
NeP1. Bending may be required for possible nucleosome binding and/or for the
assembly of other interacting partners (
6
,
7
).
Here we analyzed bending effects of NeP1 binding to F1 in the absence or the
presence of TR-TR homodimers or TR-RXR heterodimers. NeP1 shows a significant bending activity. In
the presence of TR or TR-RXR the bending angle is reduced. In addition, as compared with the NeP1-DNA complex, the position and orientation of the bend is changed
in the quaternary NeP1-TR-RXR complex.
NeP1 was purified from HeLa cell nuclear extract and prepared as described (
8
,
9
). The nuclear proteins were applied onto a Q-Sepharose column to enrich NeP1. The fractions eluting from 350 to 500 mM
NaCl were further fractionated with a heparin-Sepharose column. For both columns, HS-buffer (25 mM HEPES-KOH, pH 7.6, 5 mM MgCl
2
, 1 mM EGTA, 1 mM DTT, 10% glycerol) was used. The resulting fractions with
detectable NeP1 DNA binding activity in EMSA were eluted with 700 mM NaCl using a step gradient. After dialysis using binding buffer [10 mM HEPES, 10 mM KCl, 2.5 mM MgCl
2
, 10% (w/v) glycerol, pH 7.6], the protein fractions were applied onto a F1 DNA-affinity column. Fractions eluting with 300 mM NaCl were used in all EMSA
experiments as described below.
Human TR[alpha]1 was expressed in
Escherichia coli
. The pET-hTR[alpha]1 vector (a kind gift from L. J. DeGroot) was transformed into the
bacterial strain BL21(DE3)pLYS S (
10
). The culture was grown at 37oC to an O.D. of 0.4, IPTG was added to 0.5 mM final concentration and the
culture was incubated for 5 h at 20oC with gentle agitation. Bacteria were harvested by centrifugation and the
pellet was resuspended in 20 ml of lysis buffer (20 mM Tris, pH 8, 100 mM NaCl, 0.5 mM MgCl
2
, 1 mM PMSF, 1% aprotinin). The suspension was sonicated on ice in order to obtain a clear lysate, 2 ml of DEN (0.1 M DTT, 0.5 M EDTA, 5% NP-40) was added and bacterial debris was removed by centrifugation at 60 000
g
. Soluble proteins were precipitated from the supernatant by addition of 0.33
g/ml ammonium sulfate and recovered by centrifugation at 60 000
g
. The pellet was resuspended in 1 ml HS buffer and the solution was dialyzed
against HS buffer (see above). The fraction was loaded onto a heparin-Sepharose column and proteins were eluted using a linear KCl gradient. The fractions were
tested for the presence of hTR[alpha] by a gel retardation assay using a DR-4 probe and by SDS-PAGE followed by Coomassie staining or western blotting.
Eluates at 0.4-0.7 M KCl contained a >95% pure preparation of bacterially expressed,
soluble hTR[alpha]1.
Human RXR[alpha] was prepared by
in vitro
transcription and translation of pSG-hRXR[alpha] (
11
) using reticulocyte lysate in combination with the TNT-Kit (Promega). For the EMSA experiments, 2 [mu]l of a translation reaction were used per lane.
Chicken full length CTCF and the CTCF DNA binding domain were expressed in COS-1 cells using the plasmids pSG5-CTCF, a kind gift from E. M. Klenova (
12
), and pAB[Delta]-CTCF ZnFg (to be published elsewhere). COS 1-cells (2-3 * 10
6
) were transfected with 25 [mu]g DNA using the standards protocols. After cultivation for 48 h the cells
were collected, resuspended in 200 [mu]l binding buffer (20 mM HEPES, pH 7.8, 400 mM KCl, 20% glycerol, 2 mM DDT) and frozen in cold methanol (-80oC). After thawing on ice and sedimenting the cell debris (11 000
r.p.m., 4oC, 10 min) the supernatant was used for EMSA (see below).
Oligonucleotides containing F1 and F2 modules were cloned into the
Hin
dIII/
Acc
I site of pBluescript II SK+ (Stratagene). The F2 oligonucleotide was orientated
in sense relative to F1 and contains the natural 8 bp distance to the F1
module. The resulting pSK+ F1/F2s was partially digested with
Bss
HII and ligated with the purified
Bss
HII fragment of pSK+ F1/F2s to generate pSK+ (F1/F2)
2
with a dimerized polylinker region containing the F1/F2 modules. This plasmid
was used for circular permutation analysis by digestion with the indicated
restriction enzymes (see Fig.
2
A). The resulting fragments of 281 bp were end-labelled with [[alpha]-
32
P]dNTPs using Klenow enzyme. The radioactive fragments were cut out of a 5%
polyacrylamide gel and eluted in TE buffer (10 mM Tris, 1 mM EDTA, pH 7.6) under gentle agitation for 15 h at room temperature.
Plasmids pRN169-174, used to generate phasing vectors, were provided from Rainer
Niedenthal (
13
). Each plasmid contains three phased A-tracts, inducing an intrinsic bend, followed by a spacer region of 9, 11,
13, 15, 17 or 19 bp downstream of an
Eco
RV cleavage site. Intrinsic bend containing DNA fragments were generated by
cutting pRN169-174 with
Eco
RV/
Bam
HI and ligating the DNA fragment into the
Eco
RV/
Bam
HI site of pSK+ F1/F2s. The resulting plasmids, pSK+P1-P6, were cleaved with
Bss
HII to generate the set of 296-306 bp long DNA fragments for the phasing analysis.
DNA-protein binding reactions for the electrophoretic mobility shift analysis
(EMSA) were carried out in 40 [mu]l 1* binding buffer (described above) supplemented with 1-4 [mu]g salmon sperm DNA and 0.5-1.0 [mu]g poly(dI-dC) depending on the incubated protein
amounts. After preincubation for 15 min on ice, 15-40 fmol of each radioactive probe was added and incubated for 20 min at
room temperature. DNA-protein complexes were analyzed on nondenaturating polyacrylamide gels
[5% (w/v) acrylamide; 0.125% (w/v) bisacrylamide] in TBE buffer (90 mM Tris, 90
mM borate, 2 mM EDTA, pH 8.3). Electrophoresis was performed at 4oC with a field strength of 7 V/cm for 16 h.
The mobilities of the complexes in the circular permutation and phasing analysis
were corrected for any variation in probe mobilities by dividing the complex
mobilities with the free probe mobilities. The resulting data were expressed
relative to the complex with the highest mobility (
14
-
17
). In the case of circular permutation analysis, the data were plotted as a
function of the distance from the middle of the F1 element to the nearest end
of the fragment. The best fit to a cosine function was determined through
PlotIt (Scientific Programming Enterprises, Haslett, MI, USA). Bend centers and
standard errors were calculated using the resulting equations for the cosine
functions. The DNA flexure angle (f) was determined by the equation y
min
/y
max
= cos(f/2) where y refers to the minimal (min) or maximal (max) relative
mobility (
16
,
18
).
For phasing analysis the relative mobilities were plotted as a function of the
distance between intrinsic and NeP1 induced bend centers. The best fit curve
was assembled by PlotIt and used for calculating bending direction and bend
angle (
14
,
15
,
17
). The bend angle [alpha] was estimated by the equation
[alpha] = 2[inv.tan.(0.5A
PH
/tan 27)]
introduced by Kreppola and Kurran (
16
,
18
) where A
PH
is the phasing amplitude.
In order to analyze the
in vitro
DNA conformation within the silencer protein complex, an
in vivo
-like composition of silencer factors bound to the DNA had to be
established. Therefore, we studied the DNA binding of all three proteins NeP1,
TR and RXR, involved in synergistic transcriptional repression of the chicken
lysozyme silencer (
2
-
4
,
19
-
21
).
First we tested the DNA binding of individual fractions one by one with the
F1/F2-containing probe to analyze the resolution and specificity of all
resulting bands (Fig.
1
). The TR is detectable as a monomer and homodimer complex bound to the F2
element (lane 2) as has been shown previously (
21
). To demonstrate specificity of the TR complexes, a high affinity TR-binding oligonucleotide (
4
,
22
) containing a repeated binding site spaced by 4 nt (DR4) was used for
competition. Both complexes are sensitive to competition with DR4 (Fig.
1
, lane 3), in contrast with the competition with the non-TR binding F1 sequence (lane 4). In addition, TR homodimers are
specifically identified by their reduced DNA binding affinity in the presence
of the T3 ligand (
23
), as can be seen in lane 5. The combination of TR and RXR leads to one
additional band that can be competed with DR4 (lanes 6 and 7). The slower
migration of the TR-RXR complex is due to the higher molecular weight of RXR (
22
,
23
). Again, addition of T3 only reduces the homodimeric TR complex (lane 9).
Purifed NeP1 generates a major retarded complex in addition to a complex of
higher mobility (
3
,
19
) (Fig.
1
, lane 10). Both can be competed with the specific binding site (F1) (lane 11).
In the presence of TR, NeP1 forms two additional slower migrating complexes,
which can be identified as NeP1-TR monomer and NeP1-TR-TR homodimer by their sensitivity to DR4 competition (lane 14) or F1 competition (lane 13). Again, the TR
homodimer within the NeP1-TR-TR complex is abolished by T3 (lane 19). Further addition of RXR
yields the slowest migrating complex (NeP1-TR-RXR in lane 15), which is sensitive to F1 or DR 4 competition
(lanes 16 and 17), but not to competition with a glucocorticoid receptor
binding site (GRE, lane 18) nor to T3 incubation (lane 19). Since RXR is
in vitro
translated and TR expressed in
E.coli
, a mixture of unlabelled, unprimed reticulocyte lysate and of a protein extract from non-expressing
E.coli
cells has been tested and shown to be free of any shifting activity (lane 1). Therefore, this system allows us to demonstrate the influence on the DNA
structure, by each factor alone or in combination.
To identify changes on DNA conformation caused by any of the protein complexes,
we analyzed their mobility by circular permutation analysis (
14
,
15
,
17
). The DNA constructs were cut with different restriction enzymes as indicated
(Fig.
2
A). This results in DNA fragments of identical length with the F1/F2 sequence
placed in various positions along the length of the fragment. A bend in a DNA
fragment will be detectable, because its migration in a native polyacrylamide
gel is determined by the three-dimensional distance of both ends. The permutated DNA fragments were
incubated with NeP1, TR and RXR (Fig.
2
B). This results in a pattern of seven complexes as identified above. All of the
four NeP1-containing complexes show a migration specificity dependent on the
permutated probe used.
All complexes lacking NeP1 do not show any bending, except for the
Sal
I fragment which exhibits a higher mobility for the complexes lacking NeP1. This is probably due to the proteins binding to the very tip of the DNA of this particular probe (see Fig.
2
A) resulting in a `head-on' migration through the gel with the protein complex pulled behind. This
non-bending of TR, TR-TR or TR-RXR on the F2 element is obvious from the calculated plots
of this experiment (see below, Fig.
2
C). Thus, TR monomer, homodimer or RXR heterodimer do not bend DNA on the F1/F2
silencer.
The relative mobilities of each complex were plotted as the best fit to a cosine
function (Fig.
2
C). Using circular permutation analysis it is not possible to discriminate
between a directed bend and a higher flexibility of the DNA caused by the bound
protein. Therefore, we will use the term DNA flexure angle for the result of
this assay. The DNA flexure angle was calculated to be 100o in the case of NeP1 bound to DNA. When TR was bound either as a monomer, a
homodimer or a heterodimer with RXR in conjunction with NeP1, the DNA flexure
angle was diminished to 92-90o. The addition of T3 did not change the DNA flexure angle (data not
shown). The bend centers can be calculated for each of the bending complexes as
the points of maximal mobility determined from the minima of the curves shown
in Figure
2
C. The bending center is located upstream (distal from F2) of the F1 center,
when NeP1 is the only protein bound. The analysis of six independent experiments determined a position of -16.1 bp (+-0.02) upstream of the F1 center. In the presence of the TR or RXR
together with NeP1, the location of the bend center is shifted towards the F2
element. The TR monomer complex with NeP1 shows a bend center at -11.8 (+-0.02) bp and the NeP1/TR-TR at -4.5 (+-0.03) bp upstream of the F1 center. When
NeP1 bends DNA in the presence of TR-RXR heterodimer the bend center is located at +2.8 (+-0.02) bp downstream of the F1 center. Exactly the same position
has been determined previously as a site of DNase I hypersensitivity by
in vitro
footprinting (
2
).
To investigate whether the NeP1 induced DNA flexure angle is at least in part
caused by a directed bend, a phasing analysis was carried out. A tract of six
adenine residues is known to bend DNA with an angle of 18-21o (
14
). Accordingly, three phased A-tracts bend DNA with an angle of 54o. The resulting bend is always directed to the minor groove. DNA
constructs were generated by inserting this triple A-tract region upstream of the F1/F2 element. This defined A-tract region was phased around the helical axis by inserting 2, 4,
6, 8, 10 and 12 bp as a spacer between the intrinsic A-tract bend and the F1/F2 sequences. If analyzed on a native polyacrylamide
gel, the resulting complexes will migrate faster when both bends are oriented opposite to each other or will be retarded when the induced and intrinsic bends are in phase with each other.
EMSA analysis with these six different probes and combinations of the three different silencer proteins were carried out. The free probes show different mobilities due to different three-dimensional end-to-end distances. The pattern of retarded complexes (Fig.
3
A) was similar to that with the F1/F2 probe. The mobilities of the DNA-protein complexes were plotted as a function of the distance between both the intrinsic and NeP1-induced bend centers (Fig.
3
B). When NeP1 alone is bound to the set of six different constructs, the distance
between both bend centers covers a range from 68 to 78 bp, with a maximal
mobility at 76 bp ([bcong] 7.25 helical turns distance). Therefore, the center of the NeP1-induced bend is oriented towards a direction between major and
minor groove.
In order to analyse the functional properties of NeP1 in detail we isolated and
microsequenced NeP1 (Burcin, Lottspeich, Arnold, Runge and Renkawitz, in
preparation). The sequencing results and all of the tested binding properties
demonstrated that human NeP1 is identical to chicken CTCF. Therefore, we used
the chicken CTCF c-DNA clone (
12
) to express the chicken protein in COS-cells. In order to compare the bending properties of CTCF with NeP1, we
selected three of the permutated DNA fragments (Fig.
2
A), showing the largest difference in complex retardation for NeP1 (see Fig.
2
B). These probes were incubated with extracts from untransfected COS-cells, from COS-cells expressing CTCF and for comparison with purified NeP1 (Fig.
4
A). All of the three protein sources (the endogenous COS-protein, CTCF and NeP1) generate the same probe specific retardation. Since the DNA binding domain of CTCF is quite complex (11 zinc fingers), we wondered whether the DNA binding domain by itself would be
sufficient for DNA bending.
The chicken lysozyme gene is regulated by several regulatory elements. One is
the -2.4 kb silencer, which consists of the two modules F1 and F2. Here we
analyzed whether the proteins binding to these elements may effect the DNA
conformation. The silencer protein NeP1 generates a DNase I footprint of ~50 bp (
3
) and induces DNA flexibility with a flexure angle of 100o. This induced bend is not located in the center of the F1 sequence, but
rather it is found ~16 bp outside of the center, distal from the F2 element. Since NeP1 binds
as a monomeric protein to a sequence showing no sequence repetition, such as
palindromic or direct repeats, a possible position for the bend center cannot
be predicted from the sequence.
We are grateful to L. Schäfer-Pfeiffer for excellent technical assistance, to M. Hollenhorst for help with the mathematical calculations, to
R. Niedenthal for providing the pre-bent vectors, and to L. J. DeGroot for pET-hTR[alpha]1 and E. M. Klenova for pS65-CTCF. Also we would like to thank A. Baniahmad and M.
Short for critically reading the manuscript. This work contains parts of the
Ph.D. thesis of R. Arnold. The work was supported by grants from the
Sonderforschungsbereich 272 and from the Fond der Chemischen Industrie.
REFERENCES
Return