Inhibition of Histone Deacetylases Alters Allelic Chromatin Conformation at the Imprinted U2af1-rs1 Locus in Mouse Embryonic Stem Cells*

Most loci that are regulated by genomic imprinting have differentially methylated regions (DMRs). Previously, we showed that the DMRs of the mouse Snrpn andU2af1-rs1 genes have paternal allele-specific patterns of acetylation on histones H3 and H4. To investigate the maintenance of acetylation at these DMRs, we performed chromatin immunoprecipitation on trichostatin-A (TSA)-treated and control cells. In embryonic stem (ES) cells and fibroblasts, brief (6-h) TSA treatment induces global hyperacetylation of H3 and H4. In ES cells only, TSA led to a selective increase in maternal acetylation at U2af1-rs1, at lysine 5 of H4 and at lysine 14 of H3. TSA treatment of ES cells did not affect DNA methylation or expression of U2af1-rs1, but was sufficient to increase DNase I sensitivity along the maternal allele to a level comparable with that of the paternal allele. In fibroblasts, TSA did not alter U2af1-rs1 acetylation, and the parental alleles retained their differential DNase I sensitivity. AtSnrpn, no changes in acetylation were observed in the TSA-treated cells. Our data suggest that the mechanisms regulating histone acetylation at DMRs are locus and developmental stage-specific and are distinct from those effecting global levels of acetylation. Furthermore, it seems that the allelic U2af1-rs1acetylation determines DNase I sensitivity/chromatin conformation.

The allelic expression of imprinted genes in mammals depends on whether the allele is inherited from the mother or the father (1). Genetic experiments have established that allelic differences in DNA methylation at CpG dinucleotides are essential for the correct expression of most imprinted genes (2). The great majority of imprinted loci have defined regulatory sequences that are methylated predominantly on one or other of the two parental alleles. At several of these differentially methylated regions (DMRs), 1 the allelic methylation is established in the germ line and is maintained during embryonic and postnatal development (reviewed in Ref. 3). However, CpG methylation cannot be the sole determinant in the somatic maintenance of imprints. At constitutive DMRs, allelic methylation patterns must somehow be protected from the genomewide demethylation that occurs following fertilization and during early stages of development (reviewed in Ref. 4). Although some mouse DMRs lose methylation during preimplantation development, they can regain allelic methylation patterns at later stages (3). To account for such observations, it has been argued that the somatic maintenance of epigenetic marks at DMRs may involve multiple, interdependent, modifications including DNA methylation, nonhistone protein binding, and alterations to nucleosomes and chromatin (5)(6)(7).
Chromatin appears to be organized differently at the maternal and paternal alleles of DMRs. Several DMRs, for instance, display differential chromatin compaction when assayed by enzymatic digestion in nuclei (5). Along the splice factor-encoding, imprinted U2af1-rs1 gene on mouse chromosome 11 (8,9), the methylated and repressed maternal allele is severalfold more resistant to DNase I than the unmethylated, active, paternal allele (10). It has been suggested that the DMR comprising exon 1 of the human SNRPN gene on chromosome 15q11-q13 (11) also has differential chromatin compaction, based on assays that map matrix attachment regions (12). This DMR corresponds to the imprinting control center involved in the neuro-developmental Prader-Willi Syndrome (13,14); and both in humans and mice, it shows increased histone H3 and H4 acetylation on the unmethylated paternal allele (15,16). Histone H4 associated with the differentially methylated region-2 of the IGF2 receptor gene (Igf2r) on mouse chromosome 17 is heavily acetylated on the unmethylated paternal allele and underacetylated on the maternal allele (17). The DMR encompassing the imprinted U2af1-rs1 gene also shows pronounced acetylation differences between the methylated maternal and the unmethylated paternal allele. By using antisera specific for particular acetylated lysines on histones H3 and H4, we previously established that the underacetylation of H4 at the methylated U2af1-rs1 allele is confined to lysine 5, whereas for H3, at least three of the four acetylatable lysine residues were underacetylated. Similar results were obtained for the constitutive DMR comprising exon 1 of the mouse Snrpn gene (16). Furthermore, we found that by inducing high levels of CpG methylation on the paternal U2af1-rs1 allele, we could bring about underacetylation of H3, but not of H4, lysine 5 (16). Thus, allelic differences in histone acetylation can be both histonespecific and lysine residue-specific and can be linked differently to CpG methylation.
In the present study we explore how allelic patterns of histone acetylation are maintained at the DMRs of the imprinted mouse genes U2af1-rs1 and Snrpn by examining the in vivo effects of trichostatin-A (TSA), a highly specific inhibitor of histone deacetylases (18). We used ES cells and differentiated cells derived from interspecific hybrid embryos to compare directly the maternal and paternal alleles of the U2af1-rs1 and Snrpn genes in chromatin and expression assays. This analysis demonstrates that the DMRs at the imprinted Snrpn and the U2af1-rs1 genes are highly resistant to TSA treatments that cause global hyperacetylation of histones H3 and H4. In undifferentiated ES cells, but not in embryonic fibroblasts, and at the U2af1-rs1 DMR only, TSA treatment induces selective, lysine residue-specific changes in acetylation. These changes are associated with altered chromatin conformation along this imprinted locus.

EXPERIMENTAL PROCEDURES
Mice, Cells, and in Vitro Culture-Mice that were maternally (Matdi11) or paternally (Patdi11) disomic for chromosome 11 were produced by intercrossing animals heterozygous for the Robertsonian translocation Rb(11.13)4Bnr (19). Primary embryonic fibroblasts were derived from day 14 fetuses (line EF1; Ref. 16) and were cultured in DMEM medium containing 20% fetal calf serum. For chromatin assays, early passage (Ͻpassage 5) EF1 fibroblasts were used. ES line SF1-1 was cultured in ES medium with 10 3 units/ml of leukemia inhibitory factor (LIF) (20). For chromatin studies, semiconfluent early passage (Ͻpassage 15) SF1-1 cells were used that were morphologically undifferentiated. For TSA treatment, exponentially growing cells were cultured for 6 h in medium supplemented with TSA (at 300 nM).
Nuclease Sensitivity Assays, Southern and Northern Hybridization, and Reverse Transcriptase PCR-Nuclei were isolated from tissue or cultured cells and were resuspended in DNase I, MNase, or MspI digestion buffer at ϳ10 7 nuclei/ml, as described previously (10). For the DNase I assay, 200-l aliquots of nuclei suspension were incubated for 10 min at 25°C at increasing concentrations of enzyme (Roche Molecular Biochemicals). DNA extraction and Southern hybridization were performed as described previously (10). Hybridized filters were analyzed by phosphorimager (FLA3000, Fuji) and intensities were determined using the Quantity-One imaging software (Bio-Rad). Probe 7 is a 397-bp fragment (nucleotides 3556 -3953 of the sequence for Gen-Bank TM accession number AF309654) and probe 8 is 389-bp (nucleotides 7336 -7725 of the sequence for GenBank TM accession number AF309654). For reverse transcriptase PCR analysis, poly(A) ϩ RNA was extracted from cells using a Qiagen "Oligotex direct mRNA kit." First strand DNA synthesis was from 100 -200 ng of mRNA using random primers and Superscript II reverse transcriptase (Invitrogen). cDNA was used as template for U2af1-rs1 amplification (forward, cgcagatcagacatactgcgg; reverse, tgtggtacggccagcctatg) and Snrpn amplification (forward, gagagggagccggagatg; reverse, ttgctgttgctgagaacgtc) in a mixture containing [␣-32 P]dCTP, and the resulting products were migrated through an SSCP gel. For total RNA extraction we used a Qiagen "RNeasy kit." Northern hybridization was with a 250-bp HindIII-PstI fragment from the 5Ј end of the mouse Gapdh gene, a 499-bp fragment comprising exon 7 of Snrpn (16), and U2af1-rs1 probe 1 (10).

TSA Alters the Differential Acetylation of Maternal and Paternal U2af1-rs1 Alleles in ES Cells but Not in Fibroblasts-
Acetylation studies were performed on interspecific hybrid cell lines. In all ChIP assays, the parental alleles were compared directly by using a combination of PCR amplification and electrophoretic detection of SSCP (25). For our studies, we selected a primary embryonic fibroblast line, EF1, that is (C57BL/6 ϫ Mus spretus)F1 for proximal chromosome 11 on a homozygous C57BL/6 background (16) and a (C57BL/6 ϫ M. spretus)F1 embryonic stem (ES) cell line (SF1-1; Ref. 20). In both cell lines, the two imprinted genes that we analyzed, U2af1-rs1 and Snrpn, had maternal DNA methylation at their DMRs and were expressed exclusively from the paternal allele (16,20).
To explore the role and regulation of the paternal allelespecific H3 and H4-lysine 5 acetylation at U2af1-rs1 gene, we set out to alter levels of acetylation by treatment of the interspecific hybrid cells with TSA. To minimize pleiotropic or cell cycle effects of TSA (18), we restricted the treatment time to 6 h at a concentration of 300 nM. In initial experiments, we found that in undifferentiated SF1-1 ES cells, prolonged TSA treatment (12 or 24 h at 100 or 300 nM) led to extensive detachment of cells from the culture dish and severe restriction of cell growth after removal of the drug. In contrast, TSA treatment of SF1-1 ES cells (and EF1 fibroblasts) for only 6 h did not give rise to gross morphological changes or cell detachment, and cells continued to grow normally after TSA removal (data not shown). We analyzed global levels of H3 and H4 acetylation in untreated cells and in cells harvested immediately after the 6-h treatment. In the SF1-1 ES cells, the short treatment was sufficient to induce a major increase in histone acetylation, detected by Coomassie Blue staining and Western blotting of bulk histones separated on acetic acid/urea/Triton X-100 gels ( Fig. 1). Gel scanning showed that ϳ80% of all histone H4 in the TSA-treated ES cells was present in the tetra-, tri-, or diacetylated forms, as compared with less than 5% before treatment. Based on immunostaining with antisera to H3 acetylated at either lysine 14 or lysines 9 and/or 18 (the antiserum used does not discriminate between H3 acetylated at lysines 9 and 18), TSA treatment also induces a dramatic increase in global levels of H3 acetylation ( Fig. 1A and data not shown). Very similar results were obtained with the embryonic fibroblast line EF1 (Fig. 1B).
The effects of TSA on acetylation at U2af1-rs1 were investigated by performing chromatin immunoprecipitation (ChIP) on untreated and TSA-treated SF1-1 and EF1 cells. Antibodybound fractions were assayed for paternal and maternal DNA from the 5Ј-and 3Ј-UTRs of the U2af1-rs1 gene, by PCR amplification and electrophoretic detection of SSCP ( Fig. 2A). Of critical importance for the application of PCR-SSCP to allelic acetylation studies is the demonstration that PCR amplifications from (C57BL/6 ϫ M. spretus)F1 genomic DNA give equal amounts of C57BL/6 and M. spretus-specific fragments on SSCP gels. This is shown in Fig. 2 for the U2af1-rs1 regions analyzed.
We have shown previously that, in untreated ES cells, the methylated, maternal U2af1-rs1 allele is underacetylated at H4 lysine 5 and at H3 lysines 14 and 9/18 but not at H4 lysines 8, 12, and 16 (16). In TSA-treated ES cells, ChIP/PCR-SSCP assays show that levels of histone H4 acetylated at lysine 5 (H4Ac5) became similar on the maternal and paternal U2af1-rs1 alleles at both the 5Ј-and the 3Ј-UTR (Fig. 2, B and C, respectively). TSA also has an effect on H3Ac14 levels; paternal/maternal ratios are about 2-fold lower in TSA-treated cells, although the paternal allele remains more highly acetylated. In contrast, TSA did not effect the relatively low levels of H3Ac9/18 on the maternal U2af1-rs1 allele. If anything, the measured ratios of paternal over maternal H3-K9/18 acetylation were even higher than in the untreated cells (Fig. 2, B and  C). These findings, summarized in Table I, suggest that in ES cells, TSA induces a significant gain of maternal acetylation on H4-lysine 5 and H3-lysine 14. In contrast to the specific effects in ES cells, in the EF1 embryonic fibroblasts, the preferential acetylation of both H3 and H4 on the paternal U2af1-rs1 allele remained essentially unaltered after growth for 6 h in the presence of TSA (Table I). Hence, despite its pronounced effects on global levels of H3 and H4 acetylation in these differentiated cells (Fig. 1B), TSA did not affect the relative allelic levels of acetylation along U2af1-rs1.
U2af1-rs1 Expression and DNA Methylation Are Unaltered in TSA-treated Cells-TSA treatment did not affect the expression of the U2af1-rs1 gene. Levels of U2af1-rs1 mRNA measured by Northern hybridization were unaltered by TSA treatment of ES cells and fibroblasts (Fig. 3A). When assayed by the more sensitive reverse transcriptase PCR amplification technique, expression in the TSA-treated ES cells continued to be from the paternal chromosome exclusively (Fig. 3B). Before treatment of the SF1-1 cells with TSA, all CpG methylation at the U2af1-rs1 locus was present on maternal chromosomes. Specifically, about 85% of SF1-1 cells showed maternal methylation of a unique NotI restriction site in the 5Ј-UTR and of 24

FIG. 1. TSA increases global H3 and H4 acetylation in ES cells and fibroblasts.
Histones were extracted from untreated (Ϫ) and TSA-treated (ϩ) SF1-1 ES cells (A) and EF1 embryonic fibroblasts (B) and resolved on acetic acid/ urea/Triton X-100 gels. Subsequently, gels were stained with Coomassie Blue (left panels) or transferred to nylon filter and immunostained with antisera to H4Ac16, H4Ac5, and H3Ac9/18 (right panels). The migration of histones is as described by Bonner et al. (28) and was confirmed by immunostaining gene, shown as a box with its coding part in black. The line above represents the domain of maternal DNA methylation and differential generalized DNase I sensitivity (10). Small bars indicate the regions analyzed by PCR-SSCP. B, acetylation at the 5Ј-UTR of U2af1-rs1. ChIP was performed simultaneously on nontreated (Ϫ) and TSAtreated (ϩ) SF1-1 ES cells, and PCR on the corresponding DNA samples was with primers from the 5Ј-UTR. The first three lanes show PCR products (after SSCP electrophoresis) from control liver DNAs of C57BL/6 (m), M. spretus (s), and (C57BL/6 ϫ M. spretus)F1 (F1), respectively. Subsequent lanes show PCR amplifications following ChIP with antisera to H4Ac16, H4Ac5, H3Ac14, and H3Ac9/18, respectively. Maternal (M) and paternal (P) allele-specific fragments are indicated. C, acetylation at the 3Ј-UTR of U2af1-rs1. Amplification from the same DNAs was with primers from the 3Ј-UTR. HpaII restriction sites distributed along the locus. This was not altered by TSA treatment, and U2af1-rs1 methylation in EF1 fibroblasts was also unaltered by TSA ( Fig. 3C and data not shown). TSA Alters the Differential Sensitivity to DNase I of Maternal and Paternal U2af1-rs1 Chromatin in ES Cells but Not Fibroblasts-In adult brain and liver (10), and in kidney (Fig. 4B), chromatin along the repressed maternal U2af1-rs1 allele is severalfold less sensitive to DNase I in vivo than its paternal counterpart. A similar differential was observed in the ES cells and the fibroblasts, at the passages that were analyzed in this study. Hence, when incubating nuclei purified from these cells at a range of increasing concentrations of DNase I, the repressed maternal allele became fully digested only at severalfold higher enzyme concentrations than the active paternal allele (Fig. 4, C and D). To investigate whether this allelic difference in generalized sensitivity along U2af1-rs1 is associated with paternal allele-specific histone acetylation, we studied DNase I sensitivity in the TSA-treated SF1-1 ES cells. Using a Southern blotting approach, we found that the maternal U2af1-rs1 copy invariably becomes more DNase I-sensitive upon TSA treatment, acquiring a generalized sensitivity to DNase I similar to that of the paternal chromosome. This was observed using a BglII ϩ SacI restriction fragment length polymorphism (RFLP) (Fig. 4B) that encompasses the 3Ј-half of the gene, in which no hypersensitive sites are present (10). In contrast, TSA did not have a predominant effect on the differential sensitivity in the EF1 fibroblasts, in which the maternal U2af1-rs1 allele remained more resistant to DNase I than the paternal chromosome (Fig. 4D). This agrees with our finding of unaltered allelic acetylation in these TSA-treated differentiated cells.
To analyze nuclease sensitivity at the opposite end of the gene, we made use of the PCR-SSCP polymorphism at its 5Ј extremity (see Fig. 2), encompassing 293 bp in which no hyper-sensitive sites are present (10). Hence, nuclei from control and TSA-treated SF1-1 and EF1 cells were incubated at increasing concentrations of DNase I, and extracted DNA samples were used to PCR amplify from the 5Ј-UTR, followed by migration of the PCR products through a nondenaturing polyacrylamide gel (precisely as for the PCR-SSCP analysis of immunoprecipitated chromatin, Fig. 2). This showed that in the untreated ES cells and fibroblasts, the paternal allele was more readily digested by DNase I than the maternal allele (Fig. 5). In the TSAtreated ES cells, however, similar amounts of maternal and paternal PCR products were amplified at all but the highest nuclease concentration used, indicating that the maternal and the paternal alleles had become much more similar in their sensitivity to DNase I. In the EF1 fibroblasts, TSA treatment did not change the differential PCR amplification at the 5Ј-UTR; the repressed maternal allele remained more resistant to DNase I digestion than the expressed paternal allele (Fig. 5).
The differential, generalized sensitivity to DNase I seems not to be associated with significant differences in the positioning of nucleosomes along the maternal and paternal U2af1-rs1 alleles. This was apparent from analysis of mice that were maternally (Matdi11) or paternally (Patdi11) disomic for chromosome 11. Purified liver nuclei from Matdi11 and Patdi11 mice were incubated for increasing lengths of time with micrococcal nuclease. Genomic DNA was extracted from these MNase series, digested with the restriction enzyme HindIII, Southern blotted, and hybridized with small probes (probes 7 and 8, respectively) from the opposite extremities of a 4.2-kb HindIII fragment that comprises most of the U2af1-rs1 gene (Fig. 6). The MNase digestion profiles revealed by hybridization with both these probes appeared identical for the Matdi11 and Patdi11 nuclei (Fig. 6). This finding agrees with our earlier observation (10) that the parental alleles of U2af1-rs1 have a similar sensitivity in vivo to MNase. We also analyzed nuclei from early-passage androgenetic and parthenogenetic ES cells, and in these monoparental cells the U2af1-rs1 MNase digestion profiles were similar (data not shown).
In adult tissues, the U2af1-rs1 locus displays differential sensitivity to the restriction endonuclease MspI, with chromatin on the silent and methylated maternal chromosome being highly resistant to this methylation-insensitive enzyme (10). This difference may be attributed to the presence of methyl-CpG-binding domain (MBD) proteins on the maternal allele (16). In contrast to its pronounced effects on the generalized DNase I sensitivity in ES cells, TSA did not significantly alter the differential MspI sensitivity along the U2af1-rs1 locus (data not shown).
At DMR1 of Snrpn, TSA Treatment Does Not Affect Allelic Differences in Histone Acetylation or DNA Methylation-The 5Ј part of the imprinted Snrpn gene has a DMR (DMR1, Fig. 7A) at which methylation is established in the female germ line and is maintained in all embryonic lineages (26). In a previous study, we established that the DMR1 of Snrpn has paternal allele-specific patterns of histone acetylation. As for U2af1-rs1, the differential acetylation at histone H4 is most pronounced at lysine 5, whereas at histone H3, all lysine residues analyzed show paternal acetylation (Ref. 16; Fig. 7B).
We find that TSA causes no detectable change in the relative levels of acetylation on the maternal and paternal Snrpn alleles. With antibodies to acetylated H3 (lysines 14, 9, and 18) and H4 (lysine 5), most of the chromatin precipitated from the DMR1 originated from the paternal chromosome, as in the untreated cells (Fig. 7B). The expression of Snrpn appeared also unaltered by TSA treatment of ES cells (Fig. 7, C and D). We were unable to perform allelic acetylation studies on the EF1 fibroblasts, since these are homozygous C57BL/6 for chro- mosome 7 (where Snrpn resides). However, TSA did not lead to any detectable increase in Snrpn expression in these cells (Fig. 7C). DISCUSSION A key finding in this study is that whereas TSA induces global hyperacetylation on core histones H3 and H4, only partial (or no) effects are observed at the constitutive DMRs of the imprinted loci analyzed. Only in undifferentiated ES cells, at U2af1-rs1 but not at Snrpn, did brief TSA treatment lead to a selective increase in the relative levels of histone acetylation on the repressed maternal chromosome. The TSA-induced changes in ES cells are confined to specific lysine residues and are associated with increased sensitivity to DNase I along the imprinted locus. These findings raise questions about the regulation of histone deacetylation at DMRs and its role in chromatin organization and gene repression.
Maintenance of Differential Histone Acetylation at the U2af1-rs1 and Snrpn DMRs-The observed effects of TSA on U2af1-rs1 were cell type-and lysine residue-specific. In ES cells, TSA abolished the paternal/maternal difference in H4Ac5 at U2af1-rs1 and reduced, but did not eliminate, the difference in H3Ac14. In contrast, there was no evidence for a gain in H3Ac9/18 on the maternal allele, despite the fact that TSA treatment led to a major increase in overall levels of H3-K9/18 acetylation in the ES cells. One interpretation of this result is that continuous HDAC activity is necessary to maintain the allelic acetylation differences in H4Ac5 and H3Ac14 in ES cells. In contrast, in primary embryonic fibroblast cells, TSA had no effect on the relative levels of H3Ac14, H3Ac9/18, or H4Ac5 on the maternal and paternal U2af1-rs1 alleles. Although underlying mechanism(s) need to be determined, this difference be- tween the two cell types would imply that acetylation patterns at U2af1-rs1 become somehow stabilized and resistant to TSA upon differentiation.
In contrast to the results with U2af1-rs1, at the constitutive DMR of the Snrpn gene, we found no evidence for allelic changes in H3 and H4 acetylation in TSA-treated ES cells. One possible interpretation of this finding is that, in contrast to the regulation of global histone acetylation, which seems to involve continuous HDAC activity, there is a strongly reduced, or possibly cell cycle-regulated, turnover of histone acetate groups at these two DMRs. Alternatively, the HDAC activities that maintain the allelic acetylation patterns at U2af1-rs1 and Snrpn could differ in their sensitivities to TSA, either because the enzymes involved are distinct or because they are associated with different proteins that alter their catalytic properties. For example, the NAD-dependent deacetylase SIR2 is highly resistant to TSA (27). The selective effect of TSA on H4-lysine 5 and H3-lysine 14 acetylation at U2af1-rs1 raises the possibility that the enzyme(s) involved are specific for these lysine residues. Several HDACs show preferences for specific histones or lysine residues. Histone deacetylase Hda1p in yeast, for instance, preferentially deacetylates H3 in vitro (28), while histone acetylation patterns in mutants lacking this activity suggest a preference for H3 and H2B in vivo (29). Yeast HOS3p has a preference for H4Ac5 and H4Ac8 and H3Ac14 and H3Ac23 (30). Specificity can also be substrate-dependent. HDAC1, as part of the NuRD complex, for example, deacetylates all H4 lysines in free histones, but only lysines 5, 8, and 12 in chromatin (31). To our knowledge, so far no HDACs have been shown to be specific for H4Ac5 or H3Ac14 in chromatin.
Histone Acetylation, Chromatin Conformation, and DNA Methylation-In ES cells, TSA abolished paternal/maternal differences in generalized DNase I sensitivity along U2af1-rs1, while at the same time, allelic differences in H4Ac5 and H3Ac14 were reduced or abolished. In differentiated embryo fibroblasts, in contrast, there was no detectable change in relative levels of H3-K14 or H4-K5 acetylation and no allelespecific increase in DNase I sensitivity. These correlations do not, in themselves, establish H4-lysine 5 and/or H3-lysine 14 as mediators of DNase I sensitivity. They do, however, demonstrate that DNase I sensitivity on the maternal U2af1-rs1 allele is not dependent on global changes in histone acetylation but may be regulated by the selective acetylation/deacetylation of specific lysine residues on H3 and H4. Our study did not consider core histones H2A and H2B, and we do not exclude a possible co-involvement of acetylation at these core histones. Now that suitable antisera are becoming available, this can be investigated. It was reported recently that chromatin at silenced transgenes acquires increased DNase I sensitivity in vivo after only a few hours of TSA treatment (32), and a correlation between histone underacetylation and chromatin compaction has also been demonstrated at the chicken ␤-globin chromosomal domain (33). It is unclear how precisely deacetylation of lysine residues on H3 and H4 leads to compaction of chromatin. The N-terminal tail of histone H4 links neighboring nucleosomes in core particle crystals, and such interactions might influence chromatin compaction in vivo (34). Indeed, it has been established that the N terminus of H3 is essential for the formation of condensed chromatin fibers (35), and several recent in vitro studies show that the extent of histone deacetylation at tail domains influences chromatin condensation (36,37). Alternatively, acetylation of specific residues on H3 and/or H4 could prevent, either directly or indirectly, the association of nonhistone proteins that are involved in chromatin compaction. Centromeric chromatin is particularly susceptible to the effects of TSA and loses its ability to retain heterochromatin protein-1 (HP1) on prolonged treatment with this HDAC inhibitor (38).
TSA treatment did not induce methylation changes at U2af1-rs1, and the chromatin on the methylated maternal chromosome remained highly resistant to the restriction endonuclease MspI (which recognizes sites that can be methylated but is not methylation-sensitive). One possible interpretation of the unaltered MspI resistance of chromatin is that there is continued binding of proteins to methylated CpG dinucleotides. Several studies have demonstrated physical association between HDACs and MBD proteins (reviewed in Ref. 6), and in a previous study we demonstrated in vivo association of MECP2 with the methylated maternal U2af1-rs1 allele (16). Such association of specific MBD proteins to methylated DNA represents an attractive targeting mechanism that could, at least partially, account for the observed low acetylation at histones on the maternal U2af1-rs1 allele.
We found no evidence that TSA induces expression of U2af1-rs1 or Snrpn. A few hours of incubation with TSA also failed to de-repress the silent parental alleles of the imprinted Igf2 and H19 genes on mouse chromosome 7 (39). In contrast, prolonged treatments of cultured cells with TSA has been reported to induce expression of the normally silent allele of the imprinted Igf2 gene and the Igf2-receptor gene on mouse chromosome 17 (17,40,41). In these experiments, cells were grown for 24 h in the presence of the HDAC inhibitor. Perhaps, passage through S phase, or even a complete cell cycle, is necessary before the switch to a new transcriptional state can be accomplished at imprinted genes. We note, however, that TSA treatments of up to 72 h do not lead to de-repression of the silent maternal alleles of the U2af1-rs1 and Snrpn genes (15,41), and this supports our main finding that the DMRs at these imprinted FIG. 7. Allelic acetylation at DMR1 of Snrpn is unaltered by TSA. A, map of the Snrpn gene, with exons 1-10 (filled boxes) and the differentially methylated region comprising exon 1 (DMR1, horizontal bar) as defined by Shemer et al. (26). The small bar below indicates the sequences analyzed by PCR-SSCP. B, acetylation of Snrpn in untreated (ES) and TSA-treated (TSAϩTSA) ES cells. PCR-SSCP was performed on the same ChIP assays as described in the legend to Fig. 2B, and amplification was with primers from the DMR1 of Snrpn. Lanes C correspond to amplification from input chromatin without the addition of antiserum. C, Snrpn expression in TSA-treated ES cells. Northern blot hybridization was with an Snrpn, exon 7 (upper panel) and a Gapdh probe, respectively. Snrpn:Gapdh ratios of band intensities are indicated. This yielded the same relative intensities as compared with a probe hybridizing to 18 S and 28 S RNA (data not shown). D, unaltered paternal Snrpn expression. The lane to the left shows amplification (201-bp fragment covering exons 1-3) from C57BL/6 cDNA (from liver). To the right, amplification from cDNA samples corresponding to untreated (ES) and TSA-treated (ESϩTSA) SF1-1 cells is shown; parallel assays were performed without the addition of reverse transcriptase (Ϫ). After denaturation, Mus musculus (Mus.)-and M. spretus (Spret.)specific amplification products were separated by SSCP electrophoresis. loci are particularly resistant to the effects of this HDAC inhibitor.