(Received for publication, September 12, 1996, and in revised form, April 4, 1997)
From the Department of Biochemistry and Molecular
Biology, ** Center for Mammalian Genetics, and
Division of Pediatric Genetics, University of Florida
College of Medicine, Gainesville, Florida 32610 and the Departments of
§ Medicine and
Genetics, University of Washington,
Seattle, Washington 98195
During the process of 5-aza-2-deoxycytidine
(5aCdr)-induced reactivation of the X-linked human hypoxanthine
phosphoribosyltransferase (HPRT) gene on the inactive X
chromosome, acquisition of a nuclease-sensitive chromatin conformation
in the 5
region occurs before the appearance of HPRT mRNA.
In vivo footprinting experiments reported here show that
the 5aCdr-induced change in HPRT chromatin structure precedes the
appearance of three footprints in the immediate 5
flanking region that
are characteristic of the active HPRT allele. These and
other data suggest the following sequence of events that lead to the
reactivation of the HPRT gene after 5aCdr treatment:
(a) hemi-demethylation of the promoter, (b) an
"opening" of chromatin structure detectable as increased nuclease
sensitivity, (c) transcription factor binding to the
promoter, (d) assembly of the transcription complex, and
(e) synthesis of HPRT RNA. This sequence of events supports
the view that inactive X-linked genes are silenced by a repressive
chromatin structure that prevents the binding of transcriptional
activators to the promoter.
A unique system of differential gene expression in mammals is
established during female embryogenesis by X chromosome inactivation (1, 2). The inactivation of one X chromosome within each female somatic
nucleus generates a transcriptionally active and inactive allele of
most X-linked genes and results in dosage compensation for X-linked
genes between males and females. A variety of molecular mechanisms have
been implicated in regulating the initiation, spreading, and
maintenance of X inactivation (1-7). The involvement of DNA
methylation in this process has been established by studies using
methyl-sensitive restriction enzymes (8-10), DNA-mediated transformation (11-13), genomic sequencing (14-16), and the DNA demethylating agent 5-azacytidine (6, 13, 17-19). All of these studies
support the notion that hypermethylation of the 5 CpG island
associated with many X-linked housekeeping genes is involved in the
transcriptional silencing of these genes on the inactive X
chromosome.
The ability to demethylate and reactivate individual genes on the human
inactive X chromosome in rodent-human somatic cell hybrids by treatment
with 5-azacytidine or 5-aza-2-deoxycytidine (5aCdr)1 (6, 20) suggests that
transcriptional regulation of X-linked genes by X chromosome
inactivation involves some measure of local control either at the level
of individual genes or at the level of chromatin domains. Reactivation
of inactive X-linked genes such as the hypoxanthine
phosphoribosyltransferase (HPRT) and phosphoglycerate kinase
(PGK-1) genes after 5-azacytidine or 5aCdr treatment is
associated with both a change in chromatin structure from a
nuclease-inaccessible to a nuclease-accessible conformation and a
reduction in DNA methylation levels in the 5
CpG island (17, 21).
In previous studies, Sasaki et al. (22) assayed four
parameters during 5aCdr reactivation of the human HPRT gene
in a hamster-human somatic cell hybrid cell line (X8-6T2) containing
the inactive human X chromosome. The parameters examined were HPRT
mRNA levels and three properties of the 5 region, including hemi-
and symmetrical demethylation of DNA, and MspI nuclease
sensitivity of chromatin. Hemi-demethylation and MspI
sensitivity were detectable 6 h after the addition of 5aCdr and
reached maximum levels at 24 h, whereas symmetrical demethylation
and HPRT mRNA levels became detectable at 24 h and reached
maximum levels 48 h after exposure to 5aCdr. Thus, the initial
events during reactivation of the HPRT gene by 5aCdr
treatment are the hemi-demethylation and alteration of chromatin
structure in the promoter region, followed by symmetrical demethylation
and transcription of the gene. A similar sequence of events is reported
for 5aCdr-mediated reactivation of the mouse APRT gene (23).
The major question we address here is whether the binding of
transcription factors to the promoter region upon 5aCdr reactivation is
correlated with the early change in chromatin structure or with actual
transcription of the gene (i.e. appearance of mRNA).
Analysis by in vivo footprinting shows that the promoters of
the active HPRT (24) and PGK-1 (7, 15) alleles
are bound by transcription factors, whereas the promoters of the
corresponding inactive alleles are devoid of these factors. On the
active human HPRT allele, in vivo footprints are
associated with each of five potential Sp1 binding sites, a potential
AP2 binding site, and a region near the multiple transcription
initiation sites (24). For both the HPRT and
PGK-1 genes, no evidence has been found for the binding of
sequence-specific repressors to the promoters of the inactive alleles.
Furthermore, there is no evidence for the interaction of methylated
DNA-binding proteins (25) with the 5 regions of these genes on the
inactive X chromosome. These in vivo footprinting studies
indicate that a major component of transcriptional silencing on the
inactive X chromosome is the exclusion of transcription factors from
promoter regions.
To determine the timing of transcription factor binding during 5aCdr-induced reactivation of the human HPRT gene on the inactive X chromosome, we have performed dimethyl sulfate (DMS) in vivo footprinting on X8-6T2 cells at various times after initiating 5aCdr treatment. We now demonstrate that the binding of transcription factors to three sites in the promoter region correlates with the appearance of HPRT mRNA rather than with the preceding change in nuclease sensitivity of chromatin. Thus, the remodeling of chromatin structure during 5aCdr reactivation of the HPRT gene on the inactive X chromosome precedes and thus does not require the binding of at least three sequence-specific transcription factors to the promoter region.
DNA samples were prepared from cultures of cell lines described previously (24, 26). Briefly, GM00468 (National Institute of General Medical Sciences Human Genetic Mutant Cell Repository) is a normal diploid human male fibroblast cell line containing an active X chromosome. Cell line 4.12 (generously provided by David Ledbetter) is a hamster-human somatic cell hybrid containing only the active human X chromosome in the HPRT-deficient hamster cell line RJK88; RJK88 carries a deletion of the endogenous hamster HPRT gene. GM06318 is also a human-rodent somatic cell hybrid containing an active human X chromosome. Cell lines X8-6T2 and 8121 are hamster-human somatic hybrids containing an inactive human X chromosome (20, 27, 28).
5aCdr TreatmentX8-6T2 cells were treated with 5aCdr as described previously by Sasaki et al. (22). Briefly, cells were treated with 0.4 µg/ml 5aCdr in growth medium (RPMI 1640 medium with 10% fetal bovine serum and 40 µg/ml gentamicin) for 24 h. Cells were then washed with phosphate-buffered saline and returned to normal medium.
MspI Treatment of ChromatinAssaying chromatin structure
changes during 5aCdr reactivation was performed by nuclease digestion
of isolated nuclei as described by Sasaki et al. using the
restriction enzyme MspI (22). Briefly, nuclei were isolated
from X8-6T2 cells 0, 12, 24, 32, 48, and 60 h after initiating
treatment with 5aCdr. Nuclei from each time point were treated with 0, 200, and 600 units/ml MspI, and genomic DNA was isolated and
digested to completion with PstI and then subjected to
Southern blot analysis with human HPRT hybridization probe PB1.7. The
approximate positions of the clustered MspI sites, the PB1.7
hybridization probe, the 1.6-kb Msp-3Pst
sub-band, and the transcription factor binding sites are shown in Fig.
1.
Quantitation of hybridization signals in Southern blots was performed
by PhosphorImager analysis (Molecular Dynamics) in the PhosphorImager
Analysis Facility of the Markey Molecular Medicine Center at the
University of Washington. The radioactivity in each 3.3-kb
5Pst-3
Pst band and 1.6-kb
Msp-3
Pst sub-band was quantitated by
PhosphorImager analysis and expressed in PhosphorImager units. The
relative levels of nuclease digestion in Fig. 6 were calculated as
follows: percentage maximal = 100 × [Msp
sub-band
background]/([Msp sub-band
background] + Pst band). Msp sub-band is the
value of PhosphorImager units of the 1.6-kb
Msp-3
Pst sub-band using 200 units of
MspI; background is the PhosphorImager units of the 1.6-kb
Msp-3
Pst sub-band for 0 units of
MspI; and Pst band is the PhosphorImager units of
the 3.3-kb 5
Pst-3
Pst band at 200 units of
MspI.
RT-PCR of Human HPRT mRNA
Detecting the appearance of human HPRT mRNA was performed by RT-PCR as described by Sasaki et al. (22). The position of the forward HPRT RT-PCR primer (TCCTCCTGAGCAGTCAGC) is shown in Fig. 1. The reverse primer (GGCGATGTCAATAGGACTC) is located in exon 9. First-strand cDNA was reverse-transcribed from total RNA with random hexamer primers and amplified by PCR with human HPRT-specific primers. PCR products were fractionated on agarose gels, Southern blotted, and hybridized with a radiolabeled human HPRT cDNA probe; the human HPRT-specific RT-PCR product is 794 bp.
Quantitation of hybridization signals in Southern blots was performed
by PhosphorImager analysis as described previously. The radioactivity
in each 794-bp HPRT RT-PCR product was quantitated by PhosphorImager
analysis for each 5aCdr-treated sample (at 0, 12, 24, 32, 48, and
60 h) using 0.5 µg of RNA in the RT-PCR reaction and averaging
the results of duplicate RT-PCR reactions. The relative levels of human
HPRT mRNA in representative samples on the Southern blot were
calculated in Fig. 6 as follows: percentage maximal = 100 × [(units 5aCdr background)/micrograms of RNA]/[(units GM06318
background)/micrograms of RNA]. Units 5aCdr is the
PhosphorImager units of the 794-bp HPRT RT-PCR product, background is
the average PhosphorImager units at seven separate points on the
autoradiogram that did not contain experimental samples, micrograms of
RNA is the micrograms of RNA used in the RT-PCR reaction (0.5 µg of
RNA was used for all calculations), and units GM06318 is the number of
PhosphorImager units of the GM06318 samples after the total RNA for the
GM06318 sample was normalized to that of the experimental samples
according to the amount of MIC2 RT-PCR product. HPRT values for the
GM06318 sample were considered to represent 100% reactivation.
DMS treatment of cells, DNA purification, and piperidine cleavage of DMS-treated DNA were performed essentially as described previously (24, 29). 5aCdr-treated cells were footprinted in vivo by treatment with 0.2% DMS for 5 min in RPMI growth medium.
Ligation-mediated PCR (LMPCR) and Detection of in Vivo FootprintsLMPCR was carried out essentially as described by
Hornstra and Yang (24, 26, 29). For LMPCR, primer set E was used to analyze the upper strand encompassing the 91 footprint, and primer set C was used to analyze the upper strand in the region of the GC
boxes, as described previously for in vivo footprinting of the human HPRT gene (24). The position of the LMPCR primers (and the regions they analyze) and the location of the footprinted sites used in this study are shown in Fig. 1. Reaction conditions for
first-strand synthesis, ligation, and PCR amplification were identical
to those described previously (26).
Subsequent gel electrophoresis and electroblotting were carried out as described previously, using a 5% Long Ranger gel (AT Biochem) substituted for the standard polyacrylamide DNA sequencing gel (24, 26, 29). To visualize the final DNA sequencing ladder, strand-specific hybridization probes were synthesized from M13 clones containing the human HPRT promoter region. Probe synthesis, hybridization, washing, and autoradiography were performed as described previously (24, 26, 29).
Quantification of Footprints by Autoradiogram DensitometryQuantification of transcription factor binding at
position 91 was carried out by densitometry of the DNA sequencing
autoradiogram. The densitometric value for the band intensity at
position
91 in each sample was first normalized for loading
differences using the average band intensity of eight nonfootprinted
bands flanking both sides of the band at position
91 in each lane.
The intensity of the
91 footprint was then expressed as a percentage
of the average intensity of the same band in control cell lines 4.12 and GM00468 (Fig. 4, lanes 6 and 7), each of
which carries an active human X chromosome. The basal band intensity at
position
91 in naked DNA (Fig. 4, lane 1) was subtracted
from each normalized value (lanes 2-5) and averaged control
samples (lanes 6 and 7) before calculating the
final percentage. Quantification of bands at positions
198 and
210
was carried out in the same manner, except that the average band
intensity of four nonfootprinted flanking bands was used to normalize
for loading differences.
During the course of 5aCdr-mediated reactivation of genes on the
inactive human X chromosome, the human HPRT gene was
monitored for changes in chromatin structure, appearance of human HPRT
mRNA, and binding of transcription factors to the promoter region.
A hamster-human somatic cell hybrid containing an inactive human X
chromosome (X8-6T2) was first treated with 5aCdr for 24 h and then returned to normal growth medium without 5aCdr. At various times
from 0 to 60 h after the addition of 5aCdr, chromatin structure of
the 5 region was examined by nuclease sensitivity, human HPRT mRNA
was detected by RT-PCR, and the binding of transcription factors was
assayed by DMS in vivo footprinting. These assays were
performed simultaneously on the same cultures of 5aCdr-treated cells.
Discreet regions of
nuclease sensitivity in the 5 region of HPRT are known to
be present only on the active gene, and these can be detected with
either MspI or DNase I (21, 22, 30, 31). To examine the
effect of 5aCdr on the chromatin structure of the inactive
HPRT gene, treated and untreated X8-6T2 nuclei were
digested with MspI as described previously by Sasaki
et al. (22). As shown in Fig. 2, sensitivity
of chromatin in the 5
CpG island was negligible in X8-6T2 cells
before 5aCdr treatment (zero time), reached near maximal levels by
12 h, and reached maximal sensitivity to digestion with
MspI by 24 h after initiating treatment with 5aCdr.
Thus, the major chromatin structure changes in the HPRT gene
5
region, as reflected by the increased accessibility of
MspI sites in chromatin to cleavage by MspI,
occurred within 12-24 h of exposure to 5aCdr. These results are
similar to those reported by Sasaki et al. (22) in which
maximal sensitivity of chromatin to MspI digestion was
achieved by 24 h after initiating 5aCdr treatment.
Appearance of Human HPRT mRNA during 5aCdr-induced Reactivation of the HPRT Gene
The appearance of human HPRT mRNA in the
same populations of 5aCdr-treated X8-6T2 cells was assayed by RT-PCR
of total RNA. Fig. 3 shows Southern blot analysis of
human HPRT RT-PCR products amplified from total RNA of 5aCdr-treated
samples using a radiolabeled human HPRT cDNA hybridization probe.
Human HPRT mRNA first became detectable at 24 h after the
addition of 5aCdr and reached maximal levels at 60 h, when the
experiment was terminated. Thus, detectable HPRT mRNA levels did
not begin to appear until the chromatin structure of the 5 region had
nearly reached its maximal sensitivity to MspI, an
observation similar to that of Sasaki et al. (22).
Binding of Transcription Factors during 5aCdr-induced Reactivation of the HPRT Gene
The binding of transcription factors to the 5
region during reactivation of the inactive HPRT gene by
5aCdr treatment was assayed by LMPCR in vivo footprinting
(24, 26, 29). In previous studies, an in vivo DMS footprint
in the 5
region of the human HPRT gene was detected at
position
91 (relative to the translation start site) on the active
human X chromosome (24); this footprint was not detected on the
inactive HPRT allele. The identical footprint was also
observed in 5-azacytidine-treated cells that were hypoxanthine-, aminopterin-, and thymidine-containing medium-selected for reactivation of the human HPRT gene on the inactive X chromosome (24).
This footprint is characterized by a band of strongly enhanced
autoradiographic intensity at position
91 in the guanine-specific DNA
sequencing ladder (as compared with the intensity of the same band in
the transcriptionally inactive allele or in naked DNA), indicative of a
very DMS-reactive guanine residue on the active allele due to the
binding of a transcription factor in vivo.
To examine the binding of transcription factor(s) to the 91 region
during reactivation of the human HPRT gene, X8-6T2 cells were assayed by DMS in vivo footprinting at 0, 12, 24, 32, 48, and 60 h after the addition of 5aCdr in the same samples
assayed for MspI sensitivity and HPRT mRNA. Intact
5aCdr-treated cells were treated with DMS to generate in
vivo footprints as described previously, and the
91 footprint
was visualized by LMPCR using primer set E (24). Given the unusually
strong signal exhibited by the footprint at position
91, the binding
of transcription factor(s) to this site on the HPRT gene
promoter in 5aCdr-treated cells can be readily detected by an increase
in the relative intensity of the band at position
91 in the final
guanine-specific DNA sequencing ladder.
Fig. 4 shows the results of in vivo
footprinting assays on the 5aCdr-treated cells in the 91 region.
Lanes 7 and 8 show the very intense DMS
modification and cleavage of the guanine residue at position
91 that
is indicative of the in vivo footprint in two control cell
cultures containing only an active human X chromosome. Lane
1 displays the DMS modification and cleavage pattern of the HPRT
5
region in X8-6T2 cells (containing an inactive human X chromosome)
before 5aCdr treatment; this sequencing ladder is typical of the
in vivo guanine-specific modification and cleavage pattern
seen for the transcriptionally inactive human HPRT gene (24)
as well as for naked genomic DNA purified before DMS treatment. Samples
assayed from 0-24 h (lanes 1-3) show no increase in the band intensity at position
91 relative to naked DNA. Thus, no in vivo footprint at position
91 is detectable up to
24 h after addition of 5aCdr. However, beginning at 32 h
(lane 4) and continuing through 60 h (lane
6), a gradual increase in the intensity of the band corresponding
to the
91 footprint is detected relative to the intensity of adjacent
and surrounding nonfootprinted bands in the sequencing ladder
(lanes 4-6). Quantitation of relative band intensity at
position
91 by densitometric analysis is shown in Fig. 6. Thus, the
binding of a transcription factor(s) at position
91 seems to
correlate most closely with the appearance of HPRT mRNA, reaching
its maximum level at 60 h, rather than with the alteration of
chromatin structure, which reaches a maximum at 24-32 h. Thus, the
remodeling of chromatin structure of the 5
region to a more
nuclease-sensitive conformation in response to 5aCdr treatment does not
require binding of the transcription factor(s) to the region of the
promoter surrounding position
91.
A similar result is also seen upstream of position 91 in a region
containing five GC boxes, potential binding sites for the transcription
factor Sp1 (32). This region exhibits multiple footprints on the active
HPRT allele that include three guanines with enhanced DMS
reactivity on the upper strand at positions
163,
198, and
210
(24). These sites of enhanced DMS reactivity are not detected on the
inactive HPRT allele or on naked DNA purified before DMS
treatment. We chose to analyze sites of enhanced DMS reactivity because
they are more readily detectable in the subpopulation of 5aCdr-treated
cells that reactivate the HPRT gene than sites that exhibit
protection from DMS reactivity. Fig. 5 shows the results
of LMPCR in vivo footprinting of the region containing positions
198 and
210 in X8-6T2 cells after 5aCdr treatment. As
with the footprint at position
91, all samples assayed from 0-24 h
(Fig. 5, lanes 1-3) showed no increase in band intensity (i.e. footprints) at position
198 and
210. A detectable
increase in
198 and
210 band intensities is observed beginning at
32 h (Fig. 5, lane 4) relative to adjacent bands such
as positions
199 and
211, respectively, and continues to gradually
increase through 60 h (Fig. 5, lane 6). These relative
increases in band intensity at positions
198 and
210 are consistent
and reproducible. Densitometric analysis of relative band intensities
at positions
198 and
210 during 5aCdr reactivation is shown in Fig.
6.
Therefore, as seen with the 91 footprint, these data also indicate
that the binding of transcription factors at positions
198 and
210
(most likely Sp1) occurs late in the process of 5aCdr-mediated
reactivation (well after maximal levels of nuclease sensitivity have
been achieved at 24-32 h) and correlates more closely in time with
active transcription of the HPRT gene rather than alteration
in the chromatin structure of the HPRT locus.
In contrast, the site of enhanced DMS reactivity at position 163 in
the GC box region does not exhibit a clear increase in intensity
(relative to the adjacent band at position
164) during the course of
5aCdr reactivation (data not shown). This site is not as strongly
footprinted in cells that express HPRT fully (24), and the percentage
of reactivated cells at any of the time points examined after 5aCdr
treatment is relatively low. This is also true for the AP2 site and the
remaining Sp1 sites. Therefore, the inability to demonstrate clear
evidence of these footprints during 5aCdr reactivation most likely
reflects limitations on the sensitivity of the in vivo
footprinting assay at these sites.
A graphical summary of the events following 5aCdr
treatment of the inactive X hybrid is shown in Fig. 6, in which
chromatin structure (nuclease sensitivity), transcription factor
binding at positions 91,
198, and
210, and HPRT mRNA levels
are plotted as a function of time after initiating 5aCdr treatment. The
appearance of the
91,
198, and
210 footprints are correlated with
the appearance of HPRT mRNA rather than with the earlier change in chromatin structure. This change in chromatin structure, therefore, does not require binding of a factor(s) to the
91 region, a region that is near the multiple sites of transcription initiation and in a
location similar to regions previously reported to be critical for
silencing other genes by DNA methylation (33, 34). Our data demonstrate
that chromatin remodeling in response to 5aCdr treatment does not
require transcription factor binding to multiple transcription factor
binding sites in the HPRT promoter region.
The 5aCdr-induced reactivation of the inactive X-linked HPRT gene involves an initial hemi-demethylation of the promoter region that is associated with a change in chromatin from a nuclease-resistant to a nuclease-sensitive structure. After the alteration in chromatin structure, symmetrical demethylation occurs and HPRT mRNA appears (22). We show here that transcription factor binding to at least three sites in the promoter region is correlated with the appearance of HPRT mRNA rather than with the preceding remodeling of chromatin structure.
From this sequence of events, we propose that the change in chromatin
structure of the 5 region as a result of 5aCdr treatment does not
require transcription factor binding in the immediate promoter region
(footprints associated with the
91,
198, and
210 sites).
Reactivation of HPRT apparently requires a 5aCdr-induced remodeling of
chromatin structure such that DNA binding sites in the promoter region
become accessible to transcriptional activators. The binding of these
activators, which are known to be present in the nucleus before 5aCdr
treatment (and are bound to the active HPRT allele in female
cells), would then affect further changes in chromatin structure of the
promoter region that potentiate transcriptional activity
(e.g. an alteration in the nucleosomal structure). The
primary mechanism by which DNA methylation maintains the silence of the
HPRT gene on the inactive X chromosome may therefore involve
a role in organizing or stabilizing chromatin into a conformation that
prevents the accessibility of transcriptional activators or otherwise
precludes their binding to DNA.
Although we were only able to analyze three of the seven
sequence-specific footprints characteristic of the transcriptionally active HPRT gene (24), reports by Martinez-Balbas et
al. (35) and Lee and Garrard (36) that nuclease hypersensitivity
is independent of transcription factor binding suggests that if it were
possible to analyze the remaining footprints, these would also appear
after the induction of chromatin remodeling by 5aCdr treatment. It is possible that DNA-binding proteins not detected by our earlier DMS
in vivo footprinting studies of the promoter region of the active and inactive HPRT alleles could be responsible for
remodeling of chromatin structure in response to 5aCdr. However, recent
DNase I in vivo footprinting of the human HPRT
gene on the active and inactive X chromosome shows no evidence for
sequence-specific DNA-protein interactions on the upper strand between
positions 77 and
227 on the inactive allele, nor does it show
additional footprints in regions not previously footprinted by DMS on
the active allele (24).2 These DNase I
studies are consistent with our original findings that only the GC
boxes, AP2 site, and the transcription initiation region are bound by
factors on the active HPRT allele and that no DNA-binding
proteins are found on the inactive allele in this region (including
methyl DNA-binding proteins). The same conclusions were reached by
Pfeifer and Riggs (7) and Pfeifer et al. (15) in similar
in vivo footprinting studies of the X-linked human PGK-1 gene using DMS and DNase I.
It has been postulated that DNA methylation affects transcription by
directly altering the interaction of DNA-binding regulatory proteins
with their binding sites (37) or by altering chromatin structure and
secondarily altering sequence-specific DNA-protein interactions (38,
39). Recent studies of high-resolution methylation patterns in the
human HPRT gene and the human and mouse PGK-1 gene 5 regions suggest that direct, methylation-induced alteration of
sequence-specific DNA-protein interactions in the promoter region is
unlikely to be the primary mechanism by which X-linked genes are
silenced by DNA methylation. In these studies methylation patterns in
the 5
region of the human and mouse PGK-1 genes (15, 16,
34) and human and mouse HPRT genes (14, 40) reveal no strict
correlation between methylated CpG dinucleotides and binding sites for
transcriptional activators and no discernible conserved pattern of CpG
methylation among the inactive alleles of these genes (other than a
generally higher level of methylation compared with the active allele).
In particular, the methylation pattern seen in the mouse
PGK-1 gene makes it unlikely that DNA methylation functions
by directly modifying individual interactions between transcriptional
activators and their DNA binding sites in the immediate promoter
because only a single CpG dinucleotide is fully methylated on the
inactive X chromosome (16). The results we report here support the
concept that DNA methylation and demethylation primarily affect
chromatin structure and secondarily (as a result of changes in
chromatin structure) influence sequence-specific DNA-protein
interactions of the promoter region within native chromatin.
We thank Michael Goldman and Chien Chen for helpful comments on the manuscript and Kathleen Carlisle and Alan Fjeld for technical assistance.