5-Azadeoxycytidine-induced chromatin remodeling of the inactive X-linked HPRT gene promoter occurs prior to transcription factor binding and gene reactivation.

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.

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 nucleasesensitive 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) hemidemethylation 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)(2)(3)(4)(5)(6)(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)(12)(13), genomic sequencing (14 -16), and the DNA demethylating agent 5-azacytidine (6,13,(17)(18)(19). All of these studies support the notion that hypermethylation of the 5Ј 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-2Ј-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 5Ј 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 5Ј 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 hemidemethylation 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 5Ј 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.

EXPERIMENTAL PROCEDURES
Cell Lines-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 Treatment-X8 -6T2 cells were treated with 5aCdr as described previously by Sasaki et al. (22). Briefly, cells were treated with 0.4 g/ml 5aCdr in growth medium (RPMI 1640 medium with 10% fetal bovine serum and 40 g/ml gentamicin) for 24 h. Cells were then washed with phosphate-buffered saline and returned to normal medium.
MspI Treatment of Chromatin-Assaying 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-3ЈPst sub-band, and the transcription factor binding sites are shown in Fig. 1.
Quantitation of hybridization signals in Southern blots was per- 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 (TCCTCCT-GAGCAGTCAGC) is shown in Fig. 1. The reverse primer (GGCGAT-GTCAATAGGACTC) 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 g 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 ϭ 100 ϫ [(units 5aCdr Ϫ background)/micrograms of RNA]/[(units GM06318 Ϫ 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 g of RNA was used for all calculations), and units GM06318 is the number of Phos-phorImager 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.
Preparation of DNA: In Vitro DMS Treatment and DNA Isolation-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 Footprints-LMPCR 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 Ϫ91 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).

5-Azadeoxycytidine Reactivation of the HPRT Gene
Quantification of Footprints by Autoradiogram Densitometry-Quantification of transcription factor binding at position Ϫ91 was carried out by densitometry of the DNA sequencing autoradiogram. The densitometric value for the band intensity at position Ϫ91 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 Ϫ91 in each lane. The intensity of the Ϫ91 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 Ϫ91 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 Ϫ198 and Ϫ210 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.

RESULTS
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 5Ј 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 5aCdrtreated cells.
Changes in Chromatin Structure of the 5Ј Region during 5aCdr-induced Reactivation of the HPRT Gene-Discreet regions of nuclease sensitivity in the 5Ј 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 5Ј 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 5Ј region, as re-flected 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 5Ј 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 5Ј 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 5Ј region of the human HPRT gene was detected at position Ϫ91 (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 mediumselected 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 Ϫ91 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 Ϫ91 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 5aCdrtreated cells were treated with DMS to generate in vivo footprints as described previously, and the Ϫ91 footprint was visualized by LMPCR using primer set E (24). Given the unusually strong signal exhibited by the footprint at position Ϫ91, 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  3. RT-PCR analysis of human HPRT mRNA in X8 -6T2 cells after treatment with 5aCdr. Levels of human HPRT mRNA were assayed by RT-PCR at 0, 12, 24, 32, 48, and 60 h after initiating 5aCdr treatment of X8 -6T2 cells (lanes marked by numbered brackets). Numbers above each lane denote micrograms of RNA used in RT-PCR reaction; RNA from each 5aCdr-treated sample was assayed by RT-PCR in duplicate. C, control lanes in which no RNA was added to the RT-PCR reactions. GM06318 is a rodent-human somatic cell hybrid carrying an active human X chromosome.

5-Azadeoxycytidine Reactivation of the HPRT Gene
position Ϫ91 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 Ϫ91 region. Lanes 7 and 8 show the very intense DMS modification and cleavage of the guanine residue at position Ϫ91 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 5Ј 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 Ϫ91 relative to naked DNA. Thus, no in vivo footprint at position Ϫ91 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 Ϫ91 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 Ϫ91 by densitometric analysis is shown in Fig. 6. Thus, the binding of a transcription factor(s) at position Ϫ91 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 5Ј 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 Ϫ91.
A similar result is also seen upstream of position Ϫ91 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 Ϫ163, Ϫ198, and Ϫ210 (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 Ϫ198 and Ϫ210 in X8 -6T2 cells after 5aCdr treatment. As with the footprint at position Ϫ91, all samples assayed from 0 -24 h (Fig. 5, lanes 1-3) showed no increase in band intensity (i.e. footprints) at position Ϫ198 and Ϫ210. A detectable increase in Ϫ198 and Ϫ210 band intensities is observed beginning at 32 h (Fig. 5, lane 4) relative to adjacent bands such as positions Ϫ199 and Ϫ211, respectively, and continues to gradually increase through 60 h (Fig. 5, lane 6). These relative increases in band intensity at positions Ϫ198 and Ϫ210 are consistent and reproducible. Densitometric analysis of relative band intensities at positions Ϫ198 and Ϫ210 during 5aCdr reactivation is shown in Fig. 6.
Therefore, as seen with the Ϫ91 footprint, these data also indicate that the binding of transcription factors at positions Ϫ198 and Ϫ210 (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 Ϫ163 in the GC box region does not exhibit a clear increase in intensity (relative to the adjacent band at position Ϫ164) 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.
Summary of Nuclease Sensitivity, HPRT mRNA, and Transcription Factor Binding-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 Ϫ91, Ϫ198, and Ϫ210, and HPRT mRNA levels are plotted as a function of time after initiating 5aCdr treatment. The appearance of the Ϫ91, Ϫ198, and Ϫ210 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 Ϫ91 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

5-Azadeoxycytidine Reactivation of the HPRT Gene
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.

DISCUSSION
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 5Ј region as a result of 5aCdr treatment does not require transcription factor binding in the immediate promoter region (footprints associated with the Ϫ91, Ϫ198, and Ϫ210 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 Ϫ77 and Ϫ227 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 5Ј 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 5Ј 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.  Fig. 2. hHPRT mRNA indicates the percentage of human HPRT mRNA in 5aCdr-treated samples relative to human HPRT mRNA detected in cultured cells containing an active X chromosome (GM06318); calculations for the graph are based on the average value for two samples using 0.5 g of RNA in the RT-PCR reaction (see Fig. 3). Transcription Factor Binding indicates the intensity of the footprinted bands at positions Ϫ91, Ϫ198, and Ϫ210 relative to the intensity of the same band in cells carrying an active human X chromosome; see "Experimental Procedures" for detailed descriptions of the methods used to quantitate these three parameters. The shaded box denotes the period of 5aCdr treatment.