The Unmethylated State of CpG Islands in Mouse Fibroblasts Depends on the Poly(ADP-ribosyl)ation Process*

In vivo and in vitroexperiments carried out on L929 mouse fibroblasts suggested that the poly(ADP-ribosyl) ation process acts somehow as a protecting agent against full methylation of CpG dinucleotides in genomic DNA. Since CpG islands, which are found almost exclusively at the 5′-end of housekeeping genes, are rich in CpG dinucleotides, which are the target of mammalian DNA methyltransferase, we examined the possibility that the poly(ADP-ribosyl)ation reaction is involved in maintaining the unmethylated state of these DNA sequences. Experiments were conducted by two different strategies, using either methylation-dependent restriction enzymes on purified genomic DNA or a sequence-dependent restriction enzyme on an aliquot of the same DNA, previously modified by a bisulfite reaction. With the methylation-dependent restriction enzymes, it was observed that the “HpaII tiny fragments” greatly decreased when the cells were preincubated with 3-aminobenzamide, a well known inhibitor of poly(ADP-ribose) polymerase. The other experimental approach allowed us to prove that, as a consequence of the inhibition of the poly(ADP-ribosyl)ation process, an anomalous methylation pattern could be evidenced in the CpG island of the promoter fragment of the Htf9 gene, amplified from DNA obtained from fibroblasts preincubated with 3-aminobenzamide. These data confirm the hypothesis that, at least for the Htf9 promoter region, an active poly(ADP-ribosyl)ation protects the unmethylated state of the CpG island.

During pre-implantation development, most DNA sequences undergo extensive demethylation. This unmethylated state is maintained through the blastula stage to the time of embryo implantation, when a burst of de novo methylation generates a bimodal pattern, characterized by unmethylated CpG islands versus the bulk of genomic DNA, which is highly methylated (1). A problem yet to be solved is the identification of different cis-acting signals and trans-acting protein factors that may play a key role in defining the bimodal pattern of methylation involved in cell differentiation and gene expression.
Our previous in vitro experiments, carried out with the aim of individuating chromatin proteins involved in determining and/or maintaining in some way the DNA methylation pattern, have shown that histone H1 (2,3), through its variant H1e (4,5), is a chromatin protein that is able to greatly inhibit (Ͼ90%) methylation of double-stranded DNA. Moreover, gel retardation experiments have emphasized that H1e is the only variant able to bind CpG-rich sequences, both unmethylated CpG-rich double-stranded oligonucleotides and double-stranded DNA purified from chromatin fractions enriched in CpG islands. Further experiments have evidenced that the inhibition of in vitro enzymatic DNA methylation by histone H1 was essentially due to the poly(ADP-ribosyl)ated isoform of this protein and/or to the long and branched protein-free ADP-ribose polymers (6).
In vivo experiments carried out on L929 mouse fibroblasts preincubated for 24 h with or without 8 mM 3-ABA, 1 a well known inhibitor of poly(ADP-ribose) polymerase, confirmed the existence of a negative correlation between poly(ADP-ribosyl)ation and DNA methylation processes. A block of the poly-(ADP-ribosyl)ation process allowed, in the isolated nuclei, a considerable increase in the susceptibility of DNA to methylation by endogenous DNA methyltransferase. Subsequent methylation by exogenous enzymes was, as a consequence, severely reduced.
The aim of this paper was to verify if, in cells preincubated with 3-aminobenzamide, the DNA methyltransferase becomes able to modify the unmethylated state of CpG islands, which normally remain untouched by the action of DNA methyltransferase even though they are located in the promoter region of housekeeping genes (7)(8)(9), which are permanently accessible to the transcription factors. Our results, obtained using two different experimental approaches, allowed us to observe that the block of the poly(ADP-ribosyl)ation process modifies the methylation pattern of CpG islands in general and in particular introduces an anomalous methylation pattern in the Htf9 promoter region (10) when it is purified from cells treated with 3-aminobenzamide. These data confirm the hypothesis that, at least for the Htf9 promoter region, an active poly(ADP-ribosyl)ation protects its unmethylated state. Klenow fragment of DNA polymerase I, and 50-bp ladder molecular size markers were from Boehringer Mannheim. The BstUI restriction enzyme and SssI methylase were from New England Biolabs Inc. MspI and HpaII restriction enzymes, Taq DNA polymerase (EC 2.7.7.7), and the Wizard DNA Clean-Up system were from Promega. 3-Aminobenzamide and S-adenosyl-L-methionine were from Sigma. The nylon Gene-Screen membrane was from NEN Life Science Products. The PCR purification kit, nucleotide removal kit, and QIAEX II gel extraction kit were from QIAGEN Inc. All other chemicals used were of the highest purity commercially available.
Fluorescence-activated Cell Sorter Analysis-Fluorescence-activated cell sorter analysis of L929 mouse fibroblast cells, performed according to the previously described method (6), indicated that the cell cycle of synchronized fibroblasts was not influenced by the presence of 3-ABA.
Evaluation of the "Residual" DNA Methyl-accepting Ability-In nuclei, obtained from 6.5 ϫ 10 6 L929 mouse fibroblasts (11) and preincubated for 24 h with or without 8 mM 3-ABA, the endogenous methylaccepting ability was saturated by adding 16 M unlabeled S-AdoMet for 1 h at 37°C. The DNAs purified from these nuclei (2 g) were used as substrates to evaluate their residual methyl-accepting ability in a final volume of 50 l in the presence of 2.5 units of bacterial SssI methylase using, as methyl donor, 80 M S-AdoMet plus 100 Ci/ml S-[ 3 H]AdoMet. The enzymatic reaction was carried out for 2 h at 37°C and then was stopped by addition of 1% (w/v) SDS and 250 g/ml proteinase K at 37°C overnight. Purified DNAs were analyzed for their methyl-accepting ability (2), and incorporation of labeled methyl groups was evaluated in a Beckman LS-6800 liquid scintillation spectrometer.
Bisulfite Reaction-The bisulfite reaction was carried out according to PCR Amplification-The bisulfite-modified DNAs were used to PCRamplify fragment 1482-1773 of the Htf9 promoter region on the 5Ј-3Ј strand. The same pair of modified primers was used to amplify DNAs purified from cells preincubated either with or without 3-ABA. The sequences of the primers were as follows: forward, 5Ј-AGGTGTTGTT-GGTTTGGTGGTTTTTGGG-3Ј; and reverse, 5Ј-AATACACACATAAAT-CACCAAAAAAACACACTCC-3Ј.
The PCR mixture contained 150 ng of bisulfite-treated genomic DNA, primers (50 pmol both in frame and in reverse), dNTPs (final concentration of 0.2 mM), and 2.5 units of Taq DNA polymerase in 67 mM Tris-HCl (pH 8.8.), 6.7 mM MgCl 2 , 170 mg/ml bovine serum albumin, 16.6 mM ammonium sulfate, and 0.5 mM tetramethylammonium chloride. The reaction (50 l) was carried out under the following conditions: denaturation at 96°C for 5 min, 95°C for 1 min, 63°C for 2 min, and 72°C for 2 min for 37 cycles and a final cycle of 95°C for 1 min, 63°C for 2 min, and 72°C for 6 min. Primers were removed from the PCR mixture using the QIAGEN PCR purification kit.
Methylation-dependent Restriction Analysis-Methylathion-dependent restriction analysis was performed on 70 -100 ng of DNA (20 l) purified from nuclei (obtained from cells treated with or without 8 mM 3-ABA) in which the DNA methyl-accepting ability was saturated in the presence of 16 M S-AdoMet, exploiting endogenous DNA methyltransferase activity. Samples were digested with MspI and HpaII restriction enzymes (30 units added three times) for 36 h at 37°C. DNA fragments were end-labeled with 5 Ci of [␣-32 P]dCTP in the presence of DNA polymerase I Klenow fragment (2 units) for 30 min at 37°C, and the reaction was stopped by addition of NaEDTA to a final concentration of 25 mM. DNAs were coprecipitated in the presence of 3 g of tRNA and stored at Ϫ80°C. Samples (dissolved in 10 mM Tris-HCl (pH 7.8)) were analyzed by 2% agarose gel electrophoresis, and the bands were autoradiographed on Kodak x-ray film.
Sequence-dependent Restriction Analysis-Sequence-dependent re-striction analysis was performed on DNA fragments amplified after bisulfite modification. Before the restriction analysis, the amplified DNA was identified by 2% low melting point agarose gel electrophoresis, and the band, correspondent to a 291-bp fragment, was excised from the gel and purified using the QIAEX II gel extraction kit. Each DNA sample (30 l) was incubated with 10 units of BstUI restriction enzyme for 16 h at 60°C. DNA samples were loaded onto 2.5% agarose gel, and electrophoresis was carried out at 120 V for 3 h in 0.5% Tris borate/ EDTA buffer (pH 7.8). DNA samples were transferred from the agarose gel to a nylon GeneScreen membrane using the capillary transfer method (13), and to identify the products of the digestion mixture, they were hybridized with a uniformly 32 P-labeled probe obtained using random oligonucleotide primers. A mixture of amplified undigested DNA fragments was used as a labeled probe. A 50-bp ladder end-labeled with 10 Ci of [␥-32 P]dATP in the presence of T4 polynucleotide kinase was used as marker after removing unincorporated [␥-32 P]dATP with the QIAGEN nucleotide removal kit. Bands were evidenced by autoradiograpy on Kodak x-ray film.

RESULTS
Experiments were done to verify the possible direct role of poly(ADP-ribosyl)ation in maintaining the characteristic unmethylated state of CpG islands. The DNA used in the experiments was purified from nuclei obtained from cells in which the block of poly(ADP-ribosyl)ation allowed the introduction of new methyl groups. The DNAs used in the experiments were different in their residual methyl-accepting ability as this was reduced, according to previous results (6), to ϳ30 Ϯ 10% in the three different cell preparations examined (in comparison with the control taken as 100%) when DNA was purified from cells preincubated with 3-aminobenzamide.
The first experimental approach was carried out in accordance with the method used by Bird et al. (7)(8)(9) to evidence the clusters of unmethylated CpG dinucleotides in genomic DNA named CpG islands. The methylation state of CpG islands was investigated by digesting, with methylation-dependent restriction enzymes, the genomic DNAs purified from nuclei (obtained from cells treated with or without 3-aminobenzamide) in which the DNA methyl-accepting ability was saturated in the presence of 16 M S-AdoMet, exploiting endogenous DNA methyltransferase activity. As shown by gel electrophoresis, those "HpaII tiny fragments," which typically appear following digestion of genomic DNA with HpaII, were present when the DNA was purified from control cells, but were greatly decreased if the DNA was purified from cells preincubated with 3-ABA (Fig.  1).
In the second experimental approach, fragment 1482-1773 of the mouse CpG island Htf9 promoter region was amplified by PCR after the bisulfite reaction, which converts cytosine to uracil, but 5-methylcytosine does not react (12). Since this reaction immortalizes the methylation state of CpG sites on DNA, the fragment, even after amplification, retains the memory of the original methylation pattern. Following amplification of the DNA fragment, uracil was amplified as thymine, whereas the 5-methylcytosine residues were amplified as cytosine, so that we replaced the use of methylation-dependent restriction enzymes with sequence-dependent ones.
As the sequence-dependent restriction enzyme, we chose the BstUI enzyme, which recognizes and cuts CGCG sequence. The use of this enzyme (14,15) allowed us to observe alterations in the methylation pattern only if both cytosines were methylated in the sequence since the methylation of only one cytosine or the absence of 5-methylcytosine produces TGCG, CGTC, or TGTG sequences that BstUI cannot recognize and cut.
Following digestion of PCR-amplified DNA fragments with the BstUI restriction enzyme, it was possible to observe an anomalous methylation pattern when the Htf9 promoter region was purified from fibroblasts preincubated with 3-ABA. In fact, Southern blot analysis of digestion products showed the pres-ence of a 55-bp fragment only in the 3-ABA sample (Fig. 2a,  lane 2), suggesting that the CGCG sequence involved in the methylated state could be the sequence at position 1536 (Fig.  2b). An additional control experiment is reported in Fig. 2a  (lane 3), showing the BstUI restriction pattern of the 291-bp fragment amplified from DNA purified from nuclei obtained from cells that had not been treated with 3-ABA and that had not undergone the unlabeled S-AdoMet incubation step, but had been subjected to the bisulfite reaction. DISCUSSION How the CpG islands maintain their unmethylated state despite being rich in CpG dinucleotides is still an intriguing unanswered question. The problem is made interesting by the fact that the pattern of CpG islands remains unmethylated despite the fact that they are correlated mainly with the housekeeping genes (8), which are located in the decondensed chromatin structure. In this chromatin region, to which the transcription factors and the enzymes involved in the transcription process have easy access, the DNA methyltransferase should have easy access, too (17). Taking into consideration the importance that this biological process assumes for the regulation of gene expression (16), many researchers are working with the aim to individuate some cis-acting and/or trans-acting factors that directly or indirectly play a role in regulating the methylation pattern of CpG islands, specially since in vitro experiments (18,19) have shown that CpG islands are not by themselves unmethylatable. "Centers of methylation" able to prevent the methylation pattern of flanking DNA sequences (20 -28) as well as some sequence motifs that are intrinsically protected against de novo methylation (20, 29) have been identified. Parallel research carried out to individuate trans-acting factors capable of binding methylated DNA has met with great success, but although these proteins (30 -44) are considered to be important in mediating the methylation-dependent repression of the genes, the simplest possibility that there are trans-acting factors directly associated with CpG islands, capable of preventing the access of DNA methyltransferase to those DNA regions, has been difficult to demonstrate up to now.
The results shown in this paper demonstrate how the regulatory role played by poly(ADP-ribosyl)ation in DNA methylation does not involve only the in toto genomic DNA (6), but also the CpG islands. While these results clearly indicate the direct influence of the poly(ADP-ribosyl)ation process on the regulation of the methylation pattern of CpG islands, much still has to be done to clarify the mechanism by which this protection takes place and to identify the chromatin protein(s) that assumes this role.
As this post-synthetic modification introduces several negative charges into the modified proteins (45), it must be clarified if the poly(ADP-ribosyl)ation carries out this regulatory role by introducing the ADP-ribose polymers into the proteins able to bind methylated or unmethylated DNA. A third possibility is that the long and branched polymers of poly(ADP-ribose) by themselves directly, or positioned by a specific protein, act as trans-acting factors in this process (6). FIG. 2. Sequence-dependent restriction analysis. a, DNAs were purified and subjected to the bisulfite reaction following saturation of nuclei methyl-accepting ability by the endogenous DNA methyltransferase. These nuclei were obtained from L929 mouse fibroblasts preincubated with or without 3-aminobenzamide for 24 h. Sequence-dependent restriction analysis was performed, using 10 units of BstUI restriction enzyme at 60°C for 16 h, on the 291-bp fragment of the Htf9 gene amplified by PCR after the bisulfite reaction from control and 3-ABA-treated samples. Southern blot analysis was carried out using a mixture of all 291-bp amplified products as a 32 P-labeled probe. (For more details, see "Experimental Procedures.") Lanes 1 and 2, BstUI restriction patterns of the 291-bp fragment amplified from control and 3-ABA-treated samples, respectively; lane 3, BstUI restriction pattern of the 291-bp fragment amplified from DNA subjected to the bisulfite reaction, but obtained from nuclei purified from cells that were not treated with 3-aminobenzamide and that were not subjected to the DNA methylation step; lane M, DNA molecular size markers (50-bp ladder). b, shown is a diagram of the 291-bp amplified fragment and the sizes of fragments expected following BstUI restriction enzyme digestion.
Starting from the simplest hypothesis that there is a protein or another kind of molecule directly involved in protecting the unmethylated state of CpG islands, we propose histone H1 either covalently (46) modified by ADP-ribose polymers and/or noncovalently (47) linked to long and branched ADP-ribose polymers as the protein responsible for this protection. A possible mechanism is that histone H1 in its covalently modified isoform could position itself on the CpG islands, for which it shows a greater affinity (6), and, when there, attract the long and branched polymers that inhibit methylation of doublestranded DNA (6), thus preventing DNA methyltransferase from having access to these DNA regions. Histone H1 could participate in this role through its genetic variant H1e, which is (i) the only one involved in inhibiting the in vitro DNA methylation process (4, 5); (ii) the only one capable of binding CpG-rich DNA regions (4, 5); (iii) the only one involved in chromatin condensation and, in its poly(ADP-ribosyl)ated isoform, chromatin decondensation (48). Using methyl-accepting ability assays, we have been able to demonstrate that the poly(ADP-ribosyl)ation of the histone H1e variant facilitates chromatin decondensation, even though this modification does not remove histone H1 from chromatin (48).
Analyzing the data in light of the above, it is possible to suppose that the poly(ADP-ribosyl)ated isoform of histone H1 is present in a decondensed chromatin structure with the aim of safekeeping the expression of the housekeeping genes, through maintaining the unmethylated state of CpG islands. How histone H1 could be involved in modulating the transcription is "still an enigma" (49), but the idea that such a protein is almost absent in transcriptionally active chromatin regions (50 -53) conflicts with some data reported in the literature (54 -56) and also with our results. This contrast could only be apparent if indeed the small amount of histone H1 found associated with these DNA regions is the variant H1e in its poly-(ADP-ribosyl)ated isoform that is specifically involved in this protector role.