(Received for publication, December 6, 1996, and in revised form, February 14, 1997)
From the Biology Department, Beckman Research
Institute of the City of Hope, Duarte, California 91010 and the
§ Section of Comparative Biology, Medical Research Council
Clinical Sciences Centre, Royal Postgraduate Medical School,
Hammersmith Hospital, DuCane Road,
London W12 ONN, United Kingdom
The Xist (X inactive specific transcript) gene plays an essential role in X chromosome inactivation. To elucidate the mechanisms controlling Xist expression and X inactivation, we examined in vivo DNA-protein interactions in the Xist promoter region in a female mouse cell line (BMSL2), which has distinguishable Xist alleles. In vivo footprinting was accomplished by treatment of cells with dimethyl sulfate or ultraviolet light, followed by ligation-mediated polymerase chain reaction of purified DNA. The expressed allele on the inactive X chromosome and the silent allele on the active X chromosome were separated by the use of a restriction fragment length polymorphism prior to ligation-mediated polymerase chain reaction. The chromatin structure of the Xist promoter was found to be consistent with the activity state of the Xist gene. The silent allele (on the active X chromosome) showed no footprints, while the expressed allele (on the inactive X chromosome) showed footprints at a consensus sequence for a CCAAT box, two weak Sp1 sites, and a weak TATA box.
Early in embryogenesis, X chromosome inactivation
(XCI)1 causes the heterochromatinization
and genetic silencing of one of the two X chromosomes in female
mammalian cells (see Refs. 1 and 2). The details of the establishment
and maintenance of XCI are not yet known, but recent work has indicated
that an X-linked gene, Xist (X inactive specific
transcript), plays a critical role at least in the establishment stage
(reviewed in Refs. 3 and 4). The Xist gene (XIST
in humans) is expressed only from the inactive X chromosome (Xi) and
thus only in female somatic cells. Xist codes for a large
RNA that is preferentially localized in the nucleus and has no
conserved open reading frame. The gene maps to the X inactivation
center, which includes the X-controlling element, a cis-acting locus
that controls the likelihood of inactivation of a given X chromosome.
Recent evidence suggests that the X-controlling element locus (5), as
well as other potential key elements of XCI (6, 7), are separate from
Xist. Although the X inactivation center region including
Xist is apparently not necessary for the maintenance of XCI,
at least in cultured cells (8), other recent evidence suggests a
possible role at several stages of XCI. The level of Xist
RNA increases early in embryogenesis just prior to the time of XCI,
which is indicative of a role in establishment (9). In addition,
deletion of one Xist allele by gene targeting in female
embryonic stem cells has provided strong evidence that Xist
is necessary in cis for XCI to occur both in vitro and
in vivo (6), although it apparently does not affect
mechanisms for counting or X chromosome selection (10). Furthermore,
cytological studies showing localization of Xist RNA to Xi
(10, 11) suggest some role in maintaining the structure of the inactive
X, despite the relatively low levels of Xist transcripts
(12). Lee et al. (13) have recently found that in ES cells
sequences contained within a 450-kb YAC including the Xist
gene are sufficient to induce inactivation when integrated into
autosomes in multiple copies. In addition to the Xist locus,
the YAC includes a GC-rich island 15 kb 3 to the Xist gene
that is hypermethylated on the active X chromosome, to an extent that
varies with different X-controlling element loci (7). Taken together,
these studies suggest that protein-DNA and protein-protein interactions
at the Xist promoter are likely to be important for
understanding the establishment of XCI. In addition to studies
indicating differential DNA methylation in the Xist 5
region (14), there has so far been one published report on DNA-protein
interactions at the promoter (15). In that report, in vitro
gel mobility shift assays and reporter gene assays were used to
characterize several functional elements, including a potential
TATA-binding protein element
25 to
30 base pairs from the
transcription start site and a second element just upstream of this
site.
Here we report results of an in vivo footprinting study of the Xist promoter region, in which we separately analyzed the expressed and silent Xist alleles after in vivo treatment of the mouse female cell line, BMSL2 (16). Intact cells were treated with dimethyl sulfate (DMS) or UV, and then the DNA was analyzed by use of ligation-mediated PCR (LMPCR). We find evidence for several specific protein-DNA interactions on the expressed Xist allele, but none on the silent allele.
BMSL2 cells (HOBMSL2 cells) (16) are derived from F1 female liver cells of the cross C57BL/6-Hprta Pgk-1a (strain AT29) × congenic strain Hprtb-m3Pgk-1b (strain BM3). The cells were maintained in Dulbecco's modified Eagle's medium containing 4.5 g/liter glucose, supplemented with 1 × nonessential amino acids and 10% fetal bovine serum. They were grown as monolayers to about 70% confluency and then were used for DMS treatment or UV irradiation. After treatment of the cells with 0.1% DMS (Aldrich) in serum-free medium at room temperature for 5 min, they were washed twice with serum-free medium, washed once with phosphate-buffered saline, and then were lysed. UV irradiation was performed using a Stratalinker 2400 UV cross-linker (Stratagene). The cells were washed once with phosphate-buffered saline, irradiated with 1000 J/m2 UV (254 nm), and lysed immediately.
Isolation and in Vitro Treatment of DNAThe cells were lysed by the direct addition of 10 mM Tris (pH 7.5), 10 mM EDTA, 50 mM NaCl, 0.5% SDS, and 0.3-0.5 mg/ml proteinase K. The lysates were incubated at 37 °C overnight. In the case of DMS treatment, the incubation time was shortened to 3 h to minimize cutting of the DNA at modified residues at this incubation temperature. After the addition of NaCl to a final concentration of 290 mM, the lysates were extracted once with phenol and once with chloroform. Nucleic acids were precipitated with an equal volume of 2-propanol, resuspended in 10 mM Tris (pH 7.5), 1 mM EDTA, and 5 µg/ml DNase-free RNase (Boehringer Mannheim), and incubated at room temperature overnight. After phenol extraction and chloroform extraction, DNA was precipitated with 2-propanol.
As a control, DNA extracted from untreated cells was treated in vitro with DMS or UV. In vitro DMS treatment was performed essentially as described by Maxam and Gilbert (17). DNA (100 µg) was treated with 0.2 µl of DMS in a volume of 202 µl at room temperature for 2 min. In vitro UV treatment was performed by exposing 50-µl droplets containing 0.1 mg/ml DNA, 10 mM Tris (pH 7.5), and 1 mM EDTA to 1000 J/m2 UV.
Separation of Expressed and Silent AllelesDNA of each sample (40 µg) was digested with restriction endonucleases HpaI, PvuII, and XbaI (200 units each) at 37 °C for 5 h. The resulting restriction fragments were separated by electrophoresis through a 0.8% agarose gel in 50 mM TBE buffer containing 2 µg/ml ethidium bromide. To minimize photodamage, DNA was visualized by brief use of a long wavelength ultraviolet transilluminator. Slits were cut in the gel above and below the regions of interest, and the desired DNA fragments (1.0-1.8 kb for Xi; 2.3-4.4 kb for active X chromosomes) were collected by electrophoresis onto NA45 DEAE-cellulose membranes (Schleicher and Schuell) (18). The fragments were eluted by incubating the membrane three times for 30 min in 10 mM Tris (pH 7.5), 1 mM EDTA, and 2 M NaCl at 37 °C.
After phenol-chloroform extraction and 2-propanol precipitation, samples (3 µg) were treated with piperidine to induce cleavage at the sites of methylguanines or (6-4) photoproducts, or were incubated first with T4 endonuclease V and then with photolyase to induce cleavage at the sites of cyclobutane pyrimidine dimers, both as described by Tornaletti and Pfeifer (19).
LMPCRLMPCR was done essentially as described (19-21),
except that Vent (exo) DNA polymerase (exo
;
New England Biolabs Inc.) was used in the first strand synthesis as
well as in the PCR step of LMPCR, and a post-PCR booster step was done
using AmpliTaq DNA polymerase (Perkin-Elmer). Six primers were used:
A1, GATATATTGAAATTTTGCATAGACA; A2, TTTTGCATAGACAGGTGTGTGACCTAATGT; A3,
CATTATTTAATGTTTATGTGGAAGTTCTACA; B1, CCATAAGGCTTGGTGGTAGG; B2,
CTTGGTGGTAGGGGAACTAAAAATGTTC; B3, AAAAATGTTCCCCCAAAGCTCCTTAG. The A
primers (upstream primers) were used to analyze the lower strand of the
Xist promoter. The B primers (downstream primers) were for
the analysis of the upper strand.
Cleaved DNA (0.3 µg) in 28 µl of first strand mix was denatured at
95 °C for 5 min and annealed for 20 min at 47 °C (with primer A1)
or at 51 °C (with B1). After the addition of 1.2 units of
exo (diluted to 2 µl) on ice, each sample was incubated
at the annealing temperature for 3 min and then at 72 °C for 2 min.
This reaction mixture (30 µl) contained 20 nM primer A1
or B1, 250 µM each dNTP, 5 mM Tris (pH 7.5),
0.5 mM EDTA, and 5 mM MgSO4 in
addition to 1 × ThermoPol Buffer (New England Biolabs Inc.).
After the addition of 40 µl of ligation solution (48 mM
Tris (pH 7.5), 19 mM MgCl2, 19 mM
dithiothreitol, 1.9 mM ATP, 48 µg/ml bovine serum
albumin, and 4.5 units (Weiss) of T4 DNA ligase) and 5 µl of 20 µM double-stranded linker in 250 mM Tris (pH
7.7) on ice, the mixture was incubated at 17 °C overnight. DNA was
precipitated by addition of 20 µl of 10 M ammonium
acetate, 2 µl of 0.5 M EDTA, 1.5 µl of 20 mg/ml glycogen, and 250 µl of ethanol. The precipitate was dissolved in 50 µl of water. After the addition of 50 µl of amplification mixture
consisting of 2 × ThermoPol Buffer, 8 mM
MgSO4, 0.5 mM each dNTP, 0.4 µM
primer A2 or B2, 0.4 µM linker primer, and 4 units of
exo
, the sample was subjected to PCR using 23 cycles of 1 min at 95 °C (4 min at 95 °C for the first cycle), 2 min at
61 °C (with primer A1) or 59 °C (with B2), and 1 min at 72 °C.
After thermal cycling, AmpliTaq DNA polymerase (1 unit) was added, and
the sample was further incubated at 72 °C for 8 min prior to
phenol-chloroform extraction and ethanol precipitation.
Following published procedures (20, 21), the PCR-amplified fragments
were resolved by electrophoresis through a denaturing polyacrylamide
gel (8%), transferred to a nylon membrane by electroblotting, and
visualized by autoradiography after hybridization with a
single-stranded probe. The probe was made as described (22) from the
plasmid pGPT2 (14), consisting of the vector pBluescript II KS(+)
(Stratagene) and a 4.4-kb EcoRI insert that includes the 5
region of the Xist gene (see Fig. 1). After
HaeIII digestion, pGPT2 (50 ng) was mixed with 5 units of
Taq polymerase, GeneAmp 1 × PCR buffer (Perkin-Elmer), 0.5 µM primer A3 or B3, 0.67 µM
[
-32P]dCTP (3000 Ci/mmol), and 20 µM
each of dATP, dGTP, and dTTP in a volume of 50 µl. Repeated run-off
synthesis was performed with 30 cycles of 1 min at 94 °C (3 min at
94 °C for the first cycle) and 2 min at 52 °C (with primer A3) or
at 58 °C (with B3) and 2 min at 72 °C. The theoretical length of
the resulting single-stranded probes was 202 and 150 bases for
synthesis directed by primers A3 and B3, respectively. After visual
inspection of the autoradiogram, band ratios were measured by use of a
PhosphorImager (Molecular Dynamics, Inc.). Only regions showing at
least 50% differences between the in vivo sample and the
purified DNA control were considered to be footprints.
The mouse cell line BMSL2 is an established line derived from the
liver of female F1 hybrids with an active X chromosome
containing the Hprtb-m3 mutation (derived from
strain BM3) and an Xi carrying the Pgk-1a and
Hprta alleles (derived from strain AT29). The
cells contain normal steady-state levels of Xist
RNA.2 Sequence analysis after reverse
transcriptase PCR has shown that Xist RNA is transcribed
only from the AT29 X chromosome (16), whereas Pgk-1 RNA is
transcribed only from the BM3 X
chromosome.3 Because only the AT29 allele
has an extra PvuII site near the transcription start site
(6) (see Fig. 1), the promoter region of the BM3 allele
of the Xist gene can be distinguished from that of the AT29
allele by Southern blotting (Fig. 2). When DNA from BMSL2 cells was digested with PvuII in addition to
HpaI and XbaI (Fig. 2, lane 2), a
1.5-kb fragment (the AT29 allele) and a 3.5-kb fragment (the BM3
allele) were detected. Fig. 2 also shows that the BM3 Xist
allele is methylated at all HhaI sites in the 5 region
(Fig. 1), unlike the AT29 allele. When the methylation-sensitive enzyme
HhaI was added (Fig. 2, lane 3), the AT29 allele
was further digested, whereas the BM3 allele remained intact. Since
previous studies (14) have established that the silent allele is
methylated in somatic cells whereas the expressed allele is
unmethylated, these results further confirm that the BM3 allele is
silent, whereas the AT29 allele is expressed. Treatment of BMSL2 cells
with DMS (Fig. 2, lanes 4 and 5) or with UV
(lanes 6 and 7) before DNA extraction did not
interfere with agarose gel separation, as expected because breakage of
the DNA phosphodiester bonds does not take place until subsequent
chemical or enzymatic cleavage (19, 23). After electrophoresis and
recovery of separated fragments, the DNA was treated with piperidine or
with T4 endonuclease V and photolyase. Piperidine cleaves DMS-treated
DNA predominantly at sites of methylated guanines and UV-irradiated DNA
predominantly at sites of (6-4) photoproducts. T4 endonuclease V
cleaves UV-irradiated DNA at the sites of cyclobutane pyrimidine
dimers, and photolyase renders fragment ends ligatable.
The positions of cleavage were determined at single-nucleotide
resolution by LMPCR. To increase sensitivity, we used Vent (exo) DNA polymerase both in the first strand synthesis
and in the PCR step, instead of Sequenase and Taq DNA
polymerase (20, 21) or Vent DNA polymerase (24, 25). When the amount of
input DNA was small, the exo
enzyme gave more intense
signals (see Fig. 3), perhaps because of the lack of
both 3
5
and 5
3
exonuclease activities. This enzyme, however,
generates two kinds of DNA fragment ends, blunt ends and single-base 3
overhangs, sometimes producing strong doublet bands in the sequencing
ladder (Fig. 3, lanes 5 and 6). To solve this
problem, we added after the last cycle of PCR a final treatment with
Taq polymerase. This booster step efficiently converts ends
into single-base overhangs (Fig. 3, lanes 7 and 8).
In vivo footprints in the 5 region of the Xist
gene were discerned by comparing the pattern of the sequencing ladder
of the in vivo treated DNA sample with that of in
vitro treated purified DNA. Fig. 4 shows the result
of in vivo DMS and UV footprinting of the lower strand of
the Xist 5
region from nucleotide
180 to +9 relative to
the major transcription start site. Regions reproducibly showing 50%
or more reduction (or enhancement) in band strength relative to that of
the control are considered to be footprints and are marked by
vertical bars. All of the footprints detected in this study
were on the expressed allele; no footprints were observed on the silent
allele. We found weak protection from DMS in the regions from
49 to
50 and from
65 to
73 (Fig. 4, lanes 7 and
8). These two regions bind the transcription factor Sp1 with
low affinity in vitro.4 There
was a dipyrimidine showing strong hyper-reactivity for the in
vivo formation of cyclobutane pyrimidine dimers at nucleotides
102 and
103 only on the expressed allele. This TT dipyrimidine is
in the sequence ATTGG, hence in a consensus CCAAT box. The enhancement
of the formation of photoproducts in vivo in the CCAAT box
has been reported for the human PGK1, c-JUN, and
PCNA genes (19, 23). We also find a hyper-reactive site for
the in vivo formation of (6-4) photoproducts in the sequence
TTT at nucleotides
26 to
28, despite the fact that TT dinucleotides
are not preferred substrates for the formation of this photoproduct
(26). It has been suggested that this region (TTAAAG,
30 to
25) can
be recognized by TATA-binding protein but may bind another protein with
higher affinity (15). Assuming an interaction with TATA-binding
protein, the unusual formation of (6-4) photoproducts at this
dinucleotide might be due to the severe bending of DNA caused by the
interaction of TATA-binding protein with the TATA box (27, 28).
Fig. 5 shows the analysis of the upper strand. No DMS
footprints are apparent on this strand, but three regions show UV
photofootprints. The strongest footprint is seen at the CC dipyrimidine
at position 105 to
104, where an enhancement of (6-4) photoproducts
is found (Fig. 5, lanes 15 and 16). These
nucleotides are in a consensus CCAAT box and just opposite the hot spot
for cyclobutane dimer formation (see Figs. 4 and 6).
Protection from the formation of both (6-4) photoproducts and
cyclobutane dimers is seen at nucleotides
69 to
68.
Hyper-reactivity for cyclobutane dimer formation is also seen at
nucleotides
66 to
65. These two dipyrimidine sites are opposite one
of the DMS footprint regions on the lower strand. We also detected a
weak hyper-reactive site for cyclobutane dimer formation (TTC at
nucleotides
48 to
46) opposite to the other DMS footprint region on
the lower strand.
The results of our footprinting experiments on both strands are summarized in Fig. 6. Four regions with in vivo footprints were found, and all of them were on only the expressed allele. Three of the footprints can be detected on both upper and lower strands.
The function(s) of the Xist gene is not yet clearly understood, but several studies suggest that it is a key element necessary for the establishment of X inactivation (6, 13, 29). Establishment is likely to have several stages, four of which are (i) counting of the X to autosome ratio; (ii) initiation of the inactivation process if X/autosome = 1 or more; (iii) choosing which X remains active; and (iv) the spreading of inactivation to cover most genes on the X. Xist may not be necessary for the maintenance of X inactivation, at least in cultured cells (8), but on the other hand, Xist RNA is found in female tissues at all ages and has been postulated to be involved in chromosome architecture (10).
Correct Xist function during the establishment of X inactivation requires at least one other element to be located in cis within 1-2 megabases of the Xist locus (30). An interesting region showing differential methylation correlated with different X-controlling element alleles has, in fact, been identified by Courtier et al. (7) 15 kb downstream of Xist. Recently, Lee et al. (13) obtained evidence for a "counting/choosing" element located within 450 kb of Xist and have speculated that this cis-acting regulatory element could sense an X chromosome number and then function as an enhancer of Xist expression. Considering a somewhat different question, Rastan (4) has speculated that the Xist gene is part of the mechanism that, after initiation of the X inactivation process, leads to the retention of one and only one active X chromosome per diploid autosomal set. The activation triggered by counting the X/autosome ratio is postulated to lead to the expression or activation of a blocking factor that binds to one copy of the Xist gene. In a subsequent step, all copies of the X chromosome that are not "blocked" become inactivated. In its simplest form, this model suggests that the silent Xist gene (which is on the active X chromosome) could be kept silent by the binding of a special "blocking" factor. Both of these models suggest the need for characterization of protein-nucleic acid interactions at the Xist promoter. Also, since the Xist gene is expressed from the Xi, it is to be expected that a transcription complex will be present on the Xist promoter located on the Xi. But will this complex be present only in heterochromatin, as is found for some genes in Drosophila (31, 32)? For these reasons, in vitro and in vivo studies of the Xist promoter region may give substantial mechanistic insight.
The mouse Xist promoter region has been characterized in a
previous study (15). Reporter gene constructs containing various deletions or mutations in the promoter region were used to assay transcription after transfection. The results indicated that the major
promoter activity is contained in a region from 82 to +20 relative to
the transcription start site. Gel shift assays also were done, and
sites for protein binding were identified. In no case were differences
seen between male and female cells. However, in vitro DNA
binding studies do not necessarily give an accurate picture of in
vivo protein-DNA interactions, since the in vivo chromatin structure can not yet be mimicked in vitro. We
therefore sought to do an in vivo analysis of the
Xist promoter region, but this is not straightforward
because Xist is only expressed in female somatic cells, and
these have an equal mixture of expressed and silent alleles. To study
the nucleoprotein structure of the expressed allele and to compare it
with the silent allele, it was necessary to separately assay the two
alleles. To accomplish this we used a hybrid mouse cell line (BMSL2)
that has Xist alleles which are electrophoretically
separable because of a restriction fragment length polymorphism.
Isolation of DNA fragments after in vivo treatment, but
prior to LMPCR, permitted separate analysis of silent and expressed
alleles from the same nucleoplasm. Thus a picture of the minimally
perturbed true in vivo chromatin configuration should be
obtained by these studies.
DMS is the most commonly used agent for both in vitro and in vivo footprinting. This alkylating agent is convenient to use, but detects DNA-protein interactions mainly at guanine residues, and DMS reactivity is not at all affected by some protein contacts, for example those found in nucleosomes (33, 34). So far UV photofootprinting (19) has been less commonly used for in vivo footprinting. UV makes two major photoproducts at sites of neighboring pyrimidines, cyclobutane dimers and (6-4) photoproducts. Photoreactivity is affected by alterations in DNA conformation induced by the presence of bound protein. Photodimer sites then can be converted to strand breaks and assayed by LMPCR (35, 36).
Using the complementary methods of DMS footprinting and UV
footprinting, we examined both strands of the promoter region of the
mouse Xist gene (Fig. 6). This region is not very rich in guanines but does have numerous dipyrimidines (Fig. 6). We detected four regions with in vivo DMS or UV footprints in the 5
region of the expressed allele (Figs. 4, 5, 6): a canonical CCAAT box (nucleotides
105 to
102), two weak Sp1 sites (
73 to
65 and
50
to
46), and a weak TATA box (
28 to
26). The in vivo
footprints we observed are not dramatic (<8-fold difference in signal
intensity versus the in vitro control), which is
as one might expect for a gene expressed at relatively low level. The
in vivo footprints we detected are in sequences rather well
conserved between mouse and human and correspond to the location of
in vitro footprints4. We did not see any
in vivo footprints at the major transcription start site,
which contains a consensus sequence for an initiator element (37),
although an in vitro footprint has been
observed4. However, it should be noted that no in
vivo footprints have been found at the transcription start site of
human PGK1, a highly expressed X-linked gene that also has a
consensus initiator near the start site (38). There was a clear
photofootprint in the weak TATA box about 30 nucleotides upstream of
the major start site, but we found no footprint in the TATA consensus
sequence (nucleotides
129 to
123), which is located ~40
nucleotides upstream of a minor start site (nucleotide
82). Thus, the
Xist promoter seems to work in vivo as a promoter
with a weak TATA box as has been suggested (15) and perhaps as a weak
initiator element.
Although both the expressed allele and the silent allele of Xist are in the same nucleus, and potentially exposed to the same transcription factors, there is no evidence for the presence of any transcription factors or special blocking factors on the silent allele. This result is consistent with in vivo footprinting results for other X-linked genes in that the silent allele has no transcription factors in the promoter region (38, 39). For normal X-linked genes, it is assumed that methylation and/or a somatically heritable chromatin structure prevents access of transcription factors to the silent allele. This same mechanism thus seems likely to apply to the Xist promoter region. The silent allele in somatic cells is highly methylated, while the expressed allele is not methylated (14). This difference is likely to be involved in maintenance by stabilizing alternate chromatin structure.
We thank Drs. J. M. LeBon, S. Tommasi, R. Dammann, and G. P. Pfeifer for their kind help, A. Sancar for photolyase, and S. Lloyd for T4 endonuclease V.