Methylation of expanded CCG triplet repeat DNA from fragile X syndrome patients enhances nucleosome exclusion.

Long tracts of CCG trinucleotide or CCGNN pentanucleotide repeats in DNA have previously been shown to resist assembly into nucleosomes. This may provide a molecular explanation for the nature of certain rare, folate-sensitive fragile sites in human chromosomes that contain expanded CCG triplet tracts. Further, it is known that methylation of CpG dinucleotides at or near these fragile sites enhances the fragile phenotype. Here DNAs containing 76 tandem CCG triplets or 48 CCGNN pentanucleotide repeats were methylated with SssI methylase at three different levels of methylation. Using competitive nucleosome reconstitution/gel shift assays, the ability of these DNAs and a mixed sequence DNA from the pUC19 plasmid were compared in their ability to assemble into nucleosomes. DNA methylation had no significant effect on nucleosome formation over the pUC 19 fragment. However, the highly methylated DNAs containing 76 CCG triplets or 48 CCGNN pentanucleotide repeats were 2.0 +/- 0. 2-fold and 2.1 +/- 0.3-fold less efficient in nucleosome assembly than the respective unmethylated forms, and 4.4 +/- 0.4-fold and 12. 6 +/- 1.6-fold less efficient than a pUC19 fragment of similar length.

In eukaryotic cells methylation of CpG dinucleotides by DNA methyltransferase can directly inhibit gene expression (reviewed in Ref. 1). This enzyme, which places a methyl group on the cytosine residues of 5Ј-CpG-3Ј dinucleotides, has been isolated from several species and shown to play an important role in embryonic development (2). The inhibition of transcription by methylation has been shown to be mediated by the binding of a methyl-CpG-binding protein (MeCP-1) 1 to DNA sequences containing methylated CpG dinucleotides, with the degree of inhibition depending on the degree of methylation (3,4). DNA methylation also plays a direct role in certain pathological processes, for example in the fragile X syndrome (FRAX). This disease results from the inactivation of the FMR-1 gene, which encodes an RNA-binding protein (5,6). A block of repeating CCG triplets, which varies from 5 to 50 repeats in normal individuals, was mapped to the 5Ј-untranslated region of this gene (7,8). In carriers of FRAX, the triplet block frequently contains 60 -190 triplet repeats and in individuals with the frank disease, expansions to 200 or more repeats are common.
Here there is a strong correlation between the severity of the disease and the degree of methylation of CpG dinucleotides in the region of the expanded CCG repeats, with higher levels of methylation being more deleterious (9,10).
The fragile X syndrome was named for the presence of a rare, folate-sensitive fragile site on the X chromosome (FRAXA) whose appearance correlates with the disease. Fragile sites are defined cytologically as loci that exhibit properties of unstable chromatin, and they stain poorly and exhibit uncondensed gaps or enhanced DNA fragmentation. At FRAXA, the CCG repeat block itself is the site of preferential chromosomal breakage (7,8). Recently, expanded CCG triplet blocks have been correlated with four other rare, folate-sensitive fragile sites in the human genome, FRAXE, FRAXF, FRA16A, and FRA11B (11)(12)(13)(14)(15). In all five cases, the expression of the fragile site depends directly on the size of the triplet repeat block and the degree of methylation of the CCG repeats and adjacent CpG islands.
To investigate the molecular basis of these fragile sites, we recently examined the ability of DNA containing long blocks of CCG triplet repeats to assemble into nucleosomes. Using a combination of electron microscopy and competitive nucleosome reconstitution assays, we showed (16) that blocks of Ͼ50 contiguous CCG repeats strongly exclude nucleosomes, providing a possible explanation for the unstable chromatin phenotype. To test a specific model of nucleosome exclusion, we synthesized a model DNA containing repeating CCGNN pentanucleotides and demonstrated that a tract of 48 such repeats also strongly excludes nucleosomes (17).
Because the expression of the five fragile sites described above depends on the degree of methylation, it seemed important to examine the effects of DNA methylation on the ability of DNAs containing CCG or CCGNN repeats to assemble into chromatin. In this study, we have used competitive nucleosome reconstitution/gel retardation assays to investigate this issue and show that methylation inhibits nucleosome formation 2-fold beyond the inherent exclusion due to the sequences themselves. This observation provides further support for a model in which nucleosome exclusion creates unstable chromatin at these fragile sites, which may further facilitate the binding of the MeCP-1 protein to inhibit transcription from this region.

EXPERIMENTAL PROCEDURES
DNAs-Plasmid p(CCGNN) 48 contains four head-to-tail copies of 5Ј-CCGTA CCGAT CCGAA CCGGA CCGCT CCGAG CCGTC CCGTG CCGCA CCGGC CCGTT CCGAT-3Ј. This 240-base pair (bp) repeat block is termed here (CCGNN) 48 . A 268-bp fragment containing this element and a few nucleotides from the vector was obtained by EcoRI and XbaI digestion of p(CCGNN) 48 (17). A 282-bp fragment containing the (CCG) 76 repeat was obtained by AvaI digestion of pRW3376 (18), and a 262-bp pUC 19 fragment was obtained by polymerase chain reaction amplification using primer pairs from nucleotide 239 to 263 and from nucleotide 477 to 500. Restriction endonuclease-treated fragments or polymerase chain reaction products were purified by electrophoresis on 5% polyacrylamide gels, eluted with a solution containing 0.5 M ammonium acetate, 1 mM EDTA, and 0.1% SDS, and the purified DNAs were dephosphorylated with calf intestinal alkaline phosphatase (New England Biolabs), then labeled with T4 DNA kinase (New England Biolabs) in the presence of [␥-32 P]ATP (Amersham).
Methylation of DNA-SssI methylase (New England Biolabs) was used * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
to methylate cytosines in CpG dinucleotides (19). Reaction buffer contains 160 M S-adenosylmethionine, 50 mM NaCl, 10 mM Tris⅐HCl, pH 7.9, and 10 mM EDTA and no Mg 2ϩ was included in the reaction to avoid the topoisomerase activity in SssI methylase preparations (20). The reactions were carried out at 37°C for 2 h using 0.25-6 units of SssI methylase, and DNAs were then extracted with phenol and chloroform and precipitated with ethanol. The degree of methylation was determined by resistance to cleavage with a methylation-sensitive restriction endonuclease HhaI for the (CCG) 76 -containing DNA and the pUC 19 fragment (Fig. 1) and BsrFI for the (CCGNN) 48 -containing DNA.
Competitive Nucleosome Reconstitution-DNA fragments labeled with 32 P (50 ng) were mixed with 10 g of unlabeled calf thymus DNA and 2.5 g of histone octamers (isolated from HeLa cells; Ref. 21) in a solution containing 2 M NaCl, 100 g/ml bovine serum albumin, and 0.1% Nonidet P-40 (Sigma). The total DNA added is roughly equivalent to a 4-fold molar excess over the histone octamers. The salt was slowly lowered in increments of 0.1 M to a final concentration of 0.1 M by adding a solution of 20 mM HEPES, 1 mM EDTA, pH 7.5 (5 min for each step at room temperature). The assembly mixtures were then directly electrophoresed on 5% polyacrylamide gels at 150 V for 4 h at room temperature to separate free DNA from the nucleosome-assembled DNA. Samples were visualized by autoradiography. The amounts of DNA in each band were determined by PhosphorImager scanning (Molecular Dynamics). Each DNA was reconstituted in three separate but identical experiments.

Methylation of DNAs Containing CCG and CCGNN Re-
peats-To examine the influence of DNA methylation on nucleosome formation over DNAs containing long tracts of CCG, or CCGNN repeats, SssI methylase was used to place a methyl group on the C 5 position of cytosines within the 5Ј-CpG-3Ј dinucleotides (19). Two parallel sets of experiments were carried out. In one, a 262-bp fragment from pUC19 was compared to a 282-bp DNA containing 76 tandem CCG repeats (16). In the second set, the same pUC19 fragment was compared to a 268-bp fragment containing 48 contiguous CCGNN repeats (17). The number of possible methylation sites for the pUC19 fragment, the CCG-containing DNA, and the CCGNN-containing DNA is 40, 164, and 102, respectively. For the first set, the two DNAs were incubated with 0, 1, or 6 units of SssI methylase for 2 h at 37°C. The level of methylation was estimated from the cleavage pattern by the methylation-sensitive restriction endonuclease HhaI (Fig. 1). These two DNAs each contain a single HhaI site, and thus the resistance to cleavage by HhaI provides a direct measure of the level of methylation of the DNA. The amount of DNA in each band of the HhaI treated samples was determined by PhosphorImager scanning. The level of methylation was 0, 95%, and ϳ100% for the pUC 19 fragment and 0, 45%, and 89% for the (CCG) 76 DNAs.
For the second set of DNAs, the pUC 19 fragment was treated with 0, 0.25, or 1 units of methylase to achieve degrees of methylation of 0, 36%, and 43% and the (CCGNN) 48 DNA was treated with 0, 0.5, and 4 units of methylase to methylate at 0, 12%, and 62%, respectively (data not shown).
Competitive Nucleosome Assembly-Competitive nucleosome reconstitution was used to measure the energetics of nucleosome formation over the CCG and CCGNN repeats as compared to a pUC 19 DNA of the same size. For these assays, 50 ng of each DNA fragment (labeled with 32 P) methylated at three different levels was mixed with a 200-fold excess of unlabeled calf thymus DNA (10 g) and 2.5 g of histone octamers in a solution containing 2 M NaCl. The salt was lowered stepwise to 0.1 M NaCl and the mixture electrophoresed on a 5% polyacrylamide gel to separate free DNA from the nucleosomeassembled DNA (Fig. 2) (see "Experimental Procedures"). The results, summarized in Table I (panel A) present the ratio of the nucleosome-assembled DNA to free DNA for the pUC19 fragment and the (CCG) 76 DNA fragment each with three different degrees of methylation. This analysis showed that DNA methylation had no measurable influence on nucleosome formation for the pUC 19 fragment, an observation noted in several previous studies (22)(23)(24). In contrast, the (CCG) 76 DNA methylated at 89% of the available sites was 2.0 Ϯ 0.2-fold less effective in nucleosome assembly compared to the same (CCG) 76 DNA fragment that was unmethylated, and 4.4 Ϯ 0.4-fold less effective relative to the pUC19 fragment. The difference in free energy between the methylated (at 89%) and the unmethylated (CCG) 76 DNA is 405 Ϯ 44 cal/mol. With respect to the pUC19 DNA, the methylated (at 89%) (CCG) 76 DNA is 880 Ϯ 54 cal/mol less favorable.  1-6) and a 282-bp fragment containing the (CCG) 76 repeat element (lanes 7-12) were treated with 0, 1, or 6 units of SssI methylase to methylate cytosines at CpG dinucleotides and the degree of methylation determined by resistance to cleavage by the methylation-sensitive restriction endonuclease HhaI (see "Experimental Procedures"). HhaI cleaves the unmethylated pUC19 fragment into 242-and 20-bp pieces and the unmethylated (CCG) 76 fragment into 234-and 48-bp pieces. The small fragments are not shown. The amount of DNA in each band was determined by Phosphor-Imager scanning. The percentage of uncleaved DNA in each HhaIcleaved sample is designated as the degree of methylation.

Methylation of CCG Triplet DNA 22938
Similar results were obtained when the pUC19 fragment was compared to the 268-bp fragment containing 48 tandem CCGNN repeats (Table I , panel B; Fig. 3). The (CCGNN) 48 DNA fragment methylated at 62% of the available sites was 2.1 Ϯ 0.3-fold and 12.6 Ϯ 1.6-fold less effective in nucleosome formation compared to the unmethylated (CCGNN) 48 DNA and the pUC 19 fragment, respectively. These efficiencies correspond to 425 Ϯ 93 and 1498 Ϯ 75 cal/mol differences in free energy, respectively. DISCUSSION In this study, the effect of methylation of CpG dinucleotides on the ability of DNAs containing CCG or CCGNN repeats to assemble into nucleosomes was investigated. This analysis showed that DNAs containing ϳ228 -240 bp of continuous repeats, which are already severalfold less efficient in nucleosome assembly over mixed sequence DNAs, were further repressed in assembly following methylation. The greatest difference was observed with a fragment containing the (CCGNN) 48 repeat block. When this DNA was highly methylated, it was nearly 13-fold less effective in nucleosome assembly as compared to a mixed sequence DNA of the same size. Here methylation further increased the nucleosome exclusion due to sequence effects alone by ϳ2-fold. These results combined with the previous data provide a likely explanation for the molecular basis of the rare, folate-sensitive fragile sites, since the appearance of the fragile sites is directly linked to the increased length of the CCG triplet blocks and the high degree of methylation at the G/C-rich regions.
The CCG and CCGNN repeating sequences are both members of a repeating motif: ((G/C) 3 NN) n . In a recent model study (17), we synthesized DNAs containing (CCGNN) n repeats to test the hypothesis that (G/C) 3 wedges, which bend preferentially only into the major groove of DNA, would, when spaced by 2 nucleotides, generate a DNA that would resist wrapping around the histone core of a nucleosome. This was based on the work of Shrader and Crothers (25,26), which showed that DNA containing alternating major groove ((G/C) 3 ) and minor groove ((A/T) 3 ) wedges spaced by 2 nucleotides will preferentially wrap around the histone octamer to form nucleosomes. Sequence match analysis of the GenBank data base against 240 bp of a ((G/C) 3 NN) 48 repeating sequence revealed many matches, and 75 examples showed Ն85% sequence match over 200 bp of this motif. Many of these were present in or near the control regions for eukaryotic genes (17). Among them, the 5Ј control regions of the human dihydrofolate reductase (DHFR) and ETS-2 genes have been probed for nuclease-hypersensitive sites in vivo (often identified as nucleosome-free regions) and the nucleasehypersensitive regions (27,28) were found to overlap with the regions containing the ((G/C) 3 NN) 48 motif. It remains unknown how heavily methylated these loci may be, however from the findings described here, methylation of these regions could influence gene expression by further promoting nucleosome exclusion.
The DHFR, ETS-2, and several other genes that show sequence matches with the ((C/G) 3 NN) 48 motif are genes regulated by promoters lacking a TATA box. The "TATA-less" family includes many housekeeping genes, several oncogenes, and genes encoding growth factors and transcription factors (29). We suggested that the ((C/G) 3 NN) 48 -like sequences in these The ratio of DNA bound by histones to free DNA for each sample was determined by measuring the amount of DNA in each radioactive band using PhosphorImager scanning. The ratios for all samples were normalized against that of the pUC19 DNA fragment with no methylation. The ratio for the pUC19 DNA fragment with no methylation was assigned a value of 100. The values are derived from three separate experiments. The free energy was calculated according to the equation E(the CCG-containing fragment) ϭ RT ln(ratio of DNA in complex to free DNA for pUC19 fragment) Ϫ RT ln(ratio of DNA in complex to free DNA for the CCG (or CCGNN)-containing fragment). The free energy for the unmethylated pUC19 DNA was defined as zero.  Table I) are shown. Each DNA was reconstituted in three separate but identical experiments, and the fraction of DNA in the nucleosome-assembled and nucleosome-free DNA bands was measured by PhosphorImager scanning.

Methylation of CCG Triplet DNA 22939
genes may provide a mechanism that keeps the promoter regions accessible to transcription factors in part by excluding nucleosomes. Pugh and Tjian (30) have demonstrated that Sp1 is required in the initiation of transcription from TATA-less promoters by TFIID. Sp1 is a general transcription factor that recognizes the consensus sequence: 5Ј-(G/T)GGGCGGG(G/ A) 2 (C/T)-3Ј (31). This sequence is very G/C-rich and thus resembles the ((G/C) 3 NN) n motif. Indeed a GenBank sequence analysis revealed many sites that had been shown to be Sp1 binding sites within the sequences identified as having high degree of sequence matches to the ((G/C) 3 NN) 48 sequence. It has been demonstrated that methylation at the CpG dinucleotides of the Sp1 binding sites does not affect binding of Sp1 (32,33). Therefore, methylation of the DNA may increase the accessibility of the Sp1 binding sites within the promoter regions by inhibiting chromatinization while simultaneously not altering its affinity for Sp1 protein. The overall effect would be to facilitate initiation of transcription from TATA-less promoters. The free energy differences in nucleosome assembly between the methylated and unmethylated DNA are relatively low (Table I). However, in the cell, at physiological ionic strength and in the presence of histone H1 and other factors, these differences may be large enough to have strong biological effects and are likely to increase in strength as the size of the repeat tract lengthens. Furthermore, little has been known about how these values translate into nucleosome stability under physiological salt conditions. However, the examples of the DHFR and ETS-2 genes (17) cited above argue that the free energy of nucleosome formation measured in vitro by gel retardation methods are physiologically relevant. In addition, the strong correlation between the unstable chromatin phenotype of the rare folatesensitive fragile sites and the unfavorable energies of nucleosome formation of repeating CCG DNA, in particular when methylated, further argues that the range of free energies measured in vitro provide insight into the range of chromatin stability in vivo.
Myotonic dystrophy is a human genetic disease also linked to expansions of repeating triplets, in particular a block of CTG triplet repeats found in the 3Ј-untranslated region of the myotonic dystrophy protein kinase gene (34,35). We previously demonstrated that DNA containing long CTG triplet repeats from myotonic dystrophy patients form hyperstable nucleosomes (36,37) and suggested that this could alter the local chromatin structure to create the pathology. Thus, two repeating triplets involved in human genetic diseases: CTG and (methylated) CCG generate chromatin that is either hyperstable or hypostable. Indeed, we estimate that there may be a 80-fold difference in the stability of chromatin containing long tracts of these two triplet repeats.
The data presented here support a general proposal for the role of CCG repeats and methylation in the generation and expression of fragile sites containing these triplet repeats (Fig.  4). Long repeating CCG triplet blocks inhibit chromatinization of DNA (16), which would increase the accessibility of these regions to DNA methyltransferase. Methylation in turn would further repress chromatinization and make the DNA more available for binding by methyl-CpG-binding proteins, which have been shown to repress transcription (3,4). Thus, these changes promoted by methylation would lead to the simultaneous generation of a fragile site and the repression of nearby genes such as the FMR-1 gene.