INTRODUCTION

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.

EXPERIMENTAL PROCEDURES

BMSL2 cells (HOBMSL2 cells) (16) are derived from F₁ female liver cells of the cross C57BL/6-Hprt^a Pgk-1^a (strain AT29) × congenic strain Hprt^b-m3Pgk-1^b (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.

The 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.

DNA 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).

LMPCR 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.

Restriction map of the 5 region of the mouse Xist gene.

RESULTS

The mouse cell line BMSL2 is an established line derived from the liver of female F₁ hybrids with an active X chromosome containing the Hprt^b-m3 mutation (derived from strain BM3) and an Xi carrying the Pgk-1^a and Hprt^a 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 35 and 53 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).

In vivo footprinting analysis of the lower strand of the Xist 5 region (primer set A).

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.

In vivo footprinting analysis of the upper strand of the Xist 5 region (primer set B).

Summary of DMS and UV footprinting experiments on both strands of the 5 region of the expressed allele of the Xist gene.

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.

DISCUSSION

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.