| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
4
* Physical Molecular Biology, Laboratory of Receptor Biology and Gene Expression, National Cancer Institute, National Institutes of Health, Bethesda, Maryland 20892-5055, USA;
Department of Cellular Biotechnology and Haematology, University of Rome ‘La Sapienza’, 00161 Rome, Italy; and
Department of Chemistry and Chemical Engineering, Polytechnic University, Brooklyn, New York 11201, USA
4Correspondence: Department of Chemistry and Chemical Engineering, Polytechnic University, Six MetroTech Center, Brooklyn, NY 11201, USA. E-mail: jzlatano@duke.poly.edu; leuba{at}nih.gov
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Key Words: histone H1 • chromatin fibers • atomic force microscopy
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
. It is clear now that at least four different types of genes are controlled by DNA methylation (or the absence thereof) through a series of complex changes in their methylation pattern during development and differentiation: housekeeping genes, tissue-specific genes, imprinted genes, and X-chromosome-linked genes (recently reviewed in refs 2
3
4
5
). It is also becoming increasingly evident that DNA methylation is intimately linked to cancer and other diseases (for reviews, see refs 6
, 7
). The discovery during the past decade of proteins that preferentially bind to methylated DNA (8
9
10)
has turned attention to their possible participation in gene silencing, often through recruitment to specific methylated gene regions of histone deacetylases (for reviews, see refs 5
, 11
, 12
).
The involvement of chromatin structure in the DNA methylation-mediated regulation of transcription was demonstrated convincingly in the late 1980s by following the transcription activity of genes microinjected in cultured cells (13)
. Similar experiments were later performed in Xenopus oocytes (14)
. Ingenious experiments using transformation with patch-methylated plasmid constructs have suggested that decreased accessibility of chromatin DNA to restriction endonucleases (interpreted as chromatin compaction) spreads from focal points of methylation (15)
. The realization that chromatin structure may be involved in the effect of DNA methylation on transcription has led to studies on possible structural changes in chromatin caused by DNA methylation. Somewhat increased affinity of methylated DNA for histone octamers has been reported (16
, 17)
, but no significant changes in the fine structure of the core particle have been detected (18)
. Nucleosome placement or positioning was unaffected by DNA methylation in some sequences (17
18
19)
, whereas other sequences seemed to exclude nucleosomes when methylated (20)
. Earlier work reported the disappearance of nucleosome-free regions in gene promoters on methylation (21)
. Various studies agree that the level of m5C is higher in core than in linker DNA (ref 22
and references cited therein). More research is needed to understand chromatin structure changes brought about by DNA methylation.
To elucidate the effect of DNA methylation on chromatin structure, we used chromatin fibers isolated from control NIH/3T3 mouse fibroblasts and from fibroblasts that had been treated with 3-aminobenzamide (3-ABA), a drug that introduces new methyl groups into the 5' position of cytosine residues in CpG dinucleotides (23
24
25
26)
. AFM imaging and quantitative measurements of center-to-center internucleosomal distances, angles formed by consecutive linkers, and number of nucleosomes per unit fiber length showed that DNA hypermethylation causes chromatin fibers to compact.
To better understand the contribution of the different histone classes (core vs. linker) to the observed methylation-dependent chromatin compaction and to see whether the in vivo compaction requires methyl-CpG binding proteins, additional experiments were conducted on in vitro reconstituted chromatin fibers. The AFM results on these fibers were backed by analysis of micrococcal nuclease (MNase) digestion patterns and sucrose density gradient centrifugation. The data demonstrate unequivocally a requirement for linker histone (LH) binding in order for DNA methylation-dependent compaction to occur. The in vitro results also rule out drug-induced changes in poly(ADP-ribosyl)ation as a cause for the compaction seen in vivo.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
, 24
, 26)
.
Soluble chromatin fibers were prepared as described (27)
and characterized by MNase digestion (28)
. The protein content of the fibers was assessed by SDS-polyacrylamide gel electrophoresis (SDS-PAGE) (29)
.
Purification of DNA and histones
208–12 DNA was prepared by digestion of plasmid pPol1208 (30)
with HinpI and subsequent gel filtration on Sephacryl S500-HR using an FPLC system (Pharmacia, Piscataway, NJ). Chicken histone octamers and H1 were purified from frozen packed chicken erythrocytes (Pel Freeze) using hydroxyapatite and CM-Sephadex chromatography, respectively (31
, 32)
.
DNA methylation
208–12 DNA was methylated using SssI methylase (New England Biolabs, Beverly, MA) and subsequently purified by phenol-chloroform extraction and ethanol precipitation. The degree of methylation was checked by digestion with HpaII and MspI (New England Biolabs), both at 10 U/µg of DNA for 2 h. Only DNA fully protected from HpaII digestion was used for further experiments.
Chromatin reconstitution and characterization
Reconstitution was performed by salt dialysis (33)
; 10 µg of 208–12 DNA (control or methylated) was mixed with 10 µg of purified chicken octamers in 2 M NaCl, 10 mM Tris-HCl (pH 7.5), 0.5 mM EDTA and subsequently dialyzed at 4°C against 1 M, 0.75 M, 0.5 M (when LHs were also reconstituted) and 10 or 0 mM NaCl, each containing 10 mM Tris-HCl (pH 7.5), 0.5 mM EDTA, using Slide-A-Lyser (Pierce, Rockford, IL). Each dialysis step was carried out for at least 3 h, the last one usually for 9 h.
To reconstitute chromatin fibers containing LH, histone H1 was added at the 0.5 M NaCl dialysis step and the sample was further dialyzed against 0.5 M NaCl and the low-salt buffer. The presence of bound LHs was verified by purification of the reconstituted fibers by Centricon centrifugation, analyzing the protein content by SDS-PAGE, and MNase verification of the chromatosome pause (Fig. 3E
).
|
The quality of reconstitution was checked by methidiumpropyl-EDTA-iron(II) (MPE) hydrolysis (34
35
36)
. The reaction was performed for 10 min in 10 µl, using freshly prepared 3 µM MPE complex and stopped by adding 4x SDS loading buffer containing 20 mM bathophenanthroline disulfonic acid (Sigma, St. Louis, MO). Samples were electrophoresed on a 1.6% agarose gel in 1x TAE. Gels were stained with SYBR Gold (Molecular Probes, Eugene, OR) and washed in water before scanning on Storm 860 (Molecular Dynamics, Sunnyvale, CA).
For MNase digestions, EDTA was omitted from the dialysis buffers and Tris-HCl was replaced by TEA (triethanolamine-HCl, pH 7.5) to facilitate subsequent glutaraldehyde fixation needed for the AFM imaging. One to five units of MNase (Worthington, Freehold, NJ) were added to 8–9 µl of chromatin sample and incubated at 37°C for 2–30 min. The reaction was stopped by adding 4x SDS loading buffer containing 50 mM EDTA. Agarose gel electrophoresis and subsequent analysis were as for MPE. Before analyzing the samples on 6% PAGE (in 1x TBE), samples were treated with proteinase K, phenol/chloroform extracted and ethanol precipitated.
Sucrose density gradient centrifugation
Unmethylated 208–12 DNA was end-labeled by Cy5-dCTP (Pharmacia) using Klenow fragment (New England Biolabs). Cy5-labeled control and unlabeled methylated chromatin fibers were mixed and applied onto the top of a 15–30% sucrose gradient in 10 mM TEA-HCl (pH 7.6). The gradient was prepared in 11 ml ultracentrifugation tubes by self-diffusion of equal volumes of 15% and 30% sucrose solutions in the cold (http://133.71.125.239/english/methods/gradient.htm). The sample was centrifuged at 36,000 rpm in SW41 Ti rotor for 11 h at 4°C and fractions were collected from the bottom of the tube. Ten microliters of each fraction was made 0.3% in SDS, proteinase K was added to a final concentration of 100 µg/ml, and the samples were incubated at 37°C for 2 h. Analysis was by 1.2% agarose gel electrophoresis run in 1x TAE for 90 min at 80 V. A blue fluorescence scan of total DNA (SYBR Gold staining) and a red fluorescence scan (Cy5 fluorescence) were obtained sequentially on Storm 860.
AFM imaging and analysis
Chromatin fibers (A260=2) in 10 mM TEA-HCl, pH 7.5, 0.1 mM EDTA were fixed with 0.1% glutaraldehyde overnight. Fixation was done to avoid possible effects of shearing forces during attachment to the surface and the subsequent washing step (37)
. Two microliters of sample was deposited on freshly cleaved mica for 1 min, rinsed with five drops of Milli-Q (Millipore, Bedford, MA) water, and fluxed with argon to remove the visible liquid (38)
. Imaging was performed on a MacMode AFM (Molecular Imaging), using magnetically coated silicon nitride probes oscillated above the surface at a frequency of 
100 kHz. The amplitude of oscillation was kept constant at 5 nm during the scanning by piezo height compensation. Each set of experiments was repeated at least three times using, whenever possible, the same tip to image both control and treated samples. Using the same tip equalizes the broadening effect of the tip over images to be directly compared.
Center-to-center distances between adjacent nucleosomes and angles formed by the intersection of the two lines connecting the centers of three consecutive nucleosome centers were measured as described (32
, 39)
. Only regions of fibers clearly separated from other fibers were selected for measuring the number of nucleosomes per 10 nm of fiber contour length.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
, and the reduced poly(ADP-ribosyl)ation of a specific LH subtype may be responsible for the increased DNA methylation level (41)
. The 3-ABA-induced DNA hypermethylation causes cytologically detectable increases in both the number and the areas of heterochromatic regions, coincident with an increase in the density labeling of these regions with anti-m5C monoclonal antibodies (25)
.
Chromatin fibers solubilized by MNase treatment of nuclei were obtained from control and 3-ABA-treated NIH/3T3 cells. Prolonged MNase digestion was used to exclude from the analysis the most extended chromatin fibers whose structure is expected to be affected most by drug-induced inhibition of poly(ADP-ribosyl)ation (42)
. Such prolonged digestion was expected to result in chromatin fractions enriched in methylated cytosines, as reported (reviewed in ref 43
). The MNase digestion ladders of the solubilized chromatin fibers were indistinguishable for the control and 3-ABA-treated cells (not shown), indicating no major changes in the nucleosomal repeat length between the two cell populations. The protein patterns (not shown) demonstrated that 3-ABA treatment did not lead to any significant alterations in the histone complement either. Histone H1 was present in equal stoichiometries in both the control and treated cells.
Chromatin fibers were fixed with glutaraldehyde and imaged in air, at ambient humidity, with the MacMode AFM. By necessity, fibers were deposited on the surface from low ionic strength buffer. Imaging of chromatin fibers at moderate salt concentrations (40 mM NaCl and beyond) does not allow the nucleosome resolution needed to quantify fiber parameters (36
, 44)
. Representative images are shown in Fig. 1
A (control fibers) and B (fibers from 3-ABA-treated cells). Control fibers closely resembled fibers isolated from HeLa cells (44)
and exhibited a more extended arrangement of nucleosomes than fibers from 3-ABA-treated cells.
|
Normalized distributions of center-to-center distances between adjacent nucleosomes in chromatin fibers are presented in Fig. 2
A, B. The quantitative data agree with the visual inspection of images. The center-to-center distances for the control fibers centered around 28 nm, as expected on the basis of the known repeat length in cultured cells (45)
. There was a 4 nm reduction in this parameter for fibers isolated from treated cells. Similar differences between control and experimental samples were observed in two other experiments, although the absolute values differed somewhat from experiment to experiment, possibly as a result of collecting the cells at slightly different growth stages (46)
.
|
Figure 2C, D
are histograms of distributions of angles between successive linkers. For control fibers, angles spread from 80° to 180° (Fig. 2C
). This broad distribution is typical for linker histone containing chromatin, and significantly differs in shape from that of linker histone-depleted fibers (see fig. 4
in ref 44
). The half bell-shaped form of this distribution (see legend to Fig. 2
) is an indication that no major loss of LHs occurred during isolation (in accordance with the gel analysis) and imaging procedures. Fibers from 3-ABA-treated cells exhibited a noticeable shift of the distribution to lower angles (Fig. 2D
). Such a shift would indicate fiber compaction, in accordance with earlier predictions (47
, 48)
and experimental measurements (49)
.
|
Finally, we measured the number of nucleosomes per unit contour length of the fiber (Fig. 2E
). This parameter has been used in earlier physical and electron microscopy studies (45)
. The mean value for the control fibers was 0.46 ± 0.08 and that for the 3-ABA-treated cells was 0.7 ± 0.1 nucleosomes per 10 nm, pointing to a more compact structure in the latter.
The 208–12 chromatin fiber reconstituted in vitro: biochemical characterization and AFM imaging
Since 3-ABA was first described to affect the activity of poly(ADP-ribose)polymerase (40)
and it is well known that poly(ADP-ribosyl)ation is involved in chromatin organization essentially through modification of LHs (50)
, it was necessary to discriminate between the direct effect of the drug on poly(ADP-ribosyl)ation and its indirect effect on DNA methylation. Such discrimination could be achieved by using a system in which the only difference in chromatin fibers would be the extent of cytosine methylation, with otherwise identical histone complement. In the absence of any known in vivo system that would possess such characteristics, we turned to the widely used system for in vitro reconstituted fibers based on the tandemly repeated sequence of the 5S rRNA gene from the sea urchin Lytechinus variegatus (51)
(Fig. 3
A, B). The overall structure of these fibers is considered to be quite regular, since each 208 bp repeat positions a single nucleosome at one (or several closely situated) position(s) (51
52
53
54
55)
. Moreover, it contains 12 CpG methylation sites (Fig. 3B
), eight in the (major) core particle and four in the linker DNA. Thus, when fully methylated on both strands, each 208 bp repeat will contain 24 methyl groups. Such density of methylatable sites is unusually high, greatly exceeding the density in bulk chromatin and similar to that in CpG islands in promoters of housekeeping genes (56)
. A similar 5S rDNA sequence from the frog Xenopus borealis was used in earlier experiments aimed at understanding the effect of DNA methylation on the structure of reconstituted mononucleosomal particles and their affinity for LH binding (17)
.
Reconstitution of nucleosomal arrays on the 208–12 sequence was performed by salt dialysis. The quality of reconstitution was verified by MPE hydrolysis. MPE is a small chemical endonuclease (34)
that preferentially hydrolyses chromatin DNA in the linker regions (35)
. Indeed, when reconstituted chromatin fibers were partially cleaved with MPE, a nucleosomal ladder that extended < 12 nucleosomes was observed on agarose gels (Fig. 3C
). The nucleosome ladder was not affected by performing the reconstitution on the methylated DNA template, or by the presence of histone H1 (not shown).
Additional analysis ensured that histone H1 bound to the nucleosomal arrays correctly, i.e., that reconstitutes containing H1 showed the chromatosome pause in the course of MNase digestions (57)
. As Fig. 3E
shows, the addition of H1 to the arrays did produce the expected protection of a DNA fragment larger than the core particle size DNA (the latter is protected by the core histones). Note that both the core and the chromatosome DNA fragments are seemingly longer on these gels than the expected 146 bp and 168 bp for the core and chromatosome DNA, respectively. This anomalous electrophoretic behavior of the L. variegates 208 bp sequence has been observed before (53
, 58)
and attributed to a slight curvature in the sequence.
Reconstituted chromatin fibers were imaged under the same conditions as fibers isolated from cells (Fig. 4
). Fibers were again deposited on the surface from low ionic strength buffer, i.e., they were in their most extended state. In the case of control nucleosomal arrays [in the absence (Fig. 4A
) or presence of LHs (Fig. 4C
)] or arrays reconstituted on SssI methylase-methylated DNA (Fig. 4B
), the individual nucleosomes can be seen as discrete entities, well separated from each other. The broadening effect of the AFM tip causes all objects to look larger in the x-y dimension than they actually are, the extent of broadening depending on the dimensions of the apex of each individual tip. Thus, nucleosomes touching each other in images are actually spatially separated on the mica. The only morphological difference recognized by eye is in the case of reconstitutes on methylated DNA in the presence of LHs (Fig. 4D
). In these fibers, the nucleosomes are much more crowded than in any of the three other cases, often with some particles partially hidden by others.
Internucleosomal center-to-center distances in reconstituted fibers
To gain quantitative characterization of the various reconstitutes, we measured center-to-center internucleosomal distances and constructed frequency distribution histograms (Fig. 5
). In all cases, there was a bimodal distribution in this parameter, with two peaks centered 25 nm and 33 nm (see double Gaussian fits); the only difference among the various chromatin preparations was in the relative proportion of the two peaks (see below).
|
One possible explanation for the appearance of the 25 nm and 33 nm peaks lies in the existence, in each 208 bp repeat, of two or more minor nucleosome positions near the major one (Fig. 3B
) (53
, 54)
. Determination of the location of the 146 bp fragment protected from MNase hydrolysis in nucleosomal arrays reconstituted on the 208–18 repeat showed that a large proportion of the nucleosome particles occupy a unique position on each repeat, with some less populated sites located around this major site (53
, 54)
(Fig. 3B
).
From measurements in AFM images, our mean center-to-center distance was 29 nm for Fig. 5A-C
. In simple modeling, starting with a perfect fiber with an exact internucleosomal center-to-center spacing of 29 nm (Fig. 5E
, 5a
), a bimodal distribution became visible if > 30% of the nucleosomes occupy minor positioning sites ± 10 bp from the major site (Fig. 5E
, 5b
). A further increase in the number of nucleosomes occupying minor positions to 42% (see legend to Fig. 5
) may lead to an increase in the relative proportion of the 25 nm peak (Fig. 5E
, 5c
). The number of nucleosomes occupying minor positions used in our modeling agrees generally with the experimental finding that only 50% of the nucleosomes occupy the major position (53)
.
Thus, both the bimodal character of the center-to-center distance distribution curves and the peak height variations in these distributions (Fig. 5)
may be explained by differences in the relative occupancy of the available nucleosome positions. Relative occupancies can be modeled by Ni/No = exp(-(Ei - Eo)/kbT) where Ni and No are the occupancies of minor position i and major position o, respectively, Ei and Eo are the free energies of these positions, kb is the Boltzmann constant, and T is the absolute temperature. Using a similar equation, Dong et al. (53)
suggested a range of energy differences (Ei - Eo) on the order of 0.84 to 2.2 kbT for the alternative nucleosome positions on this sequence. We used a somewhat narrower energy variation range of 0.7–1.3 kbT because our modeling took into account only two minor positions (Dong et al. took into consideration four and more positions). We found that such modeling of relative occupancies produces two peaks with the same two maxima in the center-to-center distance distribution as observed in the experiment (compare Fig. 5A, E
).
If the two peaks in the distribution histograms of center-to-center internucleosomal distances are due to occupancy of alternative nucleosome positions on successive repeats, as suggested by our modeling, then the differences in the distributions for the different reconstitutes (Fig. 5)
would be due to redistribution of the occupancy of these alternative positions. That LHs can affect the relative occupancy of alternative nucleosome particle positions (without creating new ones) has been convincingly reported for the same DNA construct (54)
. The relative proportions of the two peaks in the distribution histograms (Fig. 5C
) suggest that LH binding causes more nucleosomes to occupy the major position than in the control nucleosomal array (the observed profile in the presence of LH resembles the modeled distribution in Fig. 5E
, 5b
, whereas that in the control array resembles the distribution in Fig. 5E
, 5c
). DNA methylation apparently does the same thing as H1, as judged by the increased second peak in Fig. 5B
.
Once we are reasonably convinced that the relative magnitudes of the two peaks in the distribution histograms reflect redistribution of nucleosome positions, let us turn to the case where methylated DNA was reconstituted with both core histones and H1 (Fig. 5D
). Since we observed an increase in the 33 nm peak in the two separate cases of methylated chromatin and control chromatin with H1, the combination of methylation and H1 was expected to lead to at least the same, if not a greater, increase in the area of this 33 nm peak. Instead, the opposite was observed. We hypothesize that this peculiar distribution reflects an independent event: compaction of the fiber.
Such a LH/DNA methylation-mediated compaction is corroborated by the significant increase in the number of nucleosomes per 10 nm of fiber contour length (Fig. 6
A). The frequency distribution histograms show that the control (minus or plus H1) reconstitutes and, in the absence of H1, on methylated templates form a peak with a mean number of 0.48 ± 0.01 (the distributions were so close in the three separate cases that we present them as a combined distribution). In contrast, the distribution on the LH-containing methylated reconstitute centers 0.67 ± 0.02, indicating a compaction of a factor of 1.4 compared with the other three cases.
|
Linker histone/DNA methylation-mediated compaction as seen by MNase digestion or sucrose density gradient sedimentation
To confirm the compaction seen in the AFM images, we used more conventional approaches. First, we compared the kinetics of MNase digestion on the different reconstitutes. As Fig. 6B
demonstrates, the methylated reconstitute containing histone H1 was digested more slowly than the corresponding H1-containing control reconstitute. Quantitation of the rate of appearance of mononucleosome particles in the course of digestion (not shown) confirmed the visual impression from the gel patterns. Control digestion experiments performed on unmethylated or methylated naked DNA revealed no difference between the two DNA substrates (Fig. 6C
). Since the only difference between the digested reconstitutes was the methylation status of the DNA, we concluded that the MNase digestion results corroborated the compaction seen in the AFM experiments and that the presence of H1 alone was not sufficient to drive the compaction. Similar comparisons between control and methylated reconstitutes that did not contain histone H1 (not shown) also corroborated the AFM-based observation that methylation alone was not sufficient to compact the fiber. Thus, DNA methylation must work in conjunction with histone H1 binding to cause chromatin fiber compaction.
In a second approach, we looked for differences in the sedimentation behavior of control H1-containing fibers and H1-containing fibers reconstituted on methylated DNA, by using sucrose density gradient centrifugation. The control H1-containing fibers were reconstituted on Cy5-labeled 208–12 DNA, this fiber and the H1-containing methylated reconstitutes were mixed in equal amounts and sedimented in the same centrifuge tube. Gradient samples were analyzed by agarose gel electrophoresis, and the distribution of the total and control material was followed by fluorescence scanning of the gels (the total DNA stained with SYBR Gold produced blue fluorescence, whereas the Cy-5-labeled control DNA fluoresced in red). Figure 7
A presents the distribution of both fluorescence labels throughout the gradient, with the third line (solid) presenting the difference between the two profiles, corresponding to the distribution of the methylated material in the gradient. As expected, the methylated H1-containing chromatin fibers sedimented faster than the control fibers, again indicating higher degree of compaction for the former fibers. Two repetitions of this experiment gave the same result.
|
Unmethylated and methylated DNA templates are equally loaded with histone octamers
The above results could easily be explained if there were a difference in the affinity of the histone octamers to these two kinds of DNA templates. The MPE ladders were indistinguishable from each other, with the same level of background material, suggesting there were no significant differences in the saturation levels of the two templates with octamers. The importance of this issue demanded more stringent analysis. We used two restriction endonuclease-based assays. In the first assay, all four types of reconstitutes were digested with DraI, an enzyme that cuts close to the dyad axis of the major nucleosome position (see Fig. 3B
), and the resulting DNA fragments were analyzed on agarose gels (Fig. 7B
). All four reconstitutes were digested to a similar degree, with the majority of the material remaining undigested due to the protection of the cleavage site by the histone octamer.
In the second assay, introduced by Hansen et al. (52)
as a way to estimate the average number of nucleosomes on repetitive templates, the nucleosomal arrays were digested with EcoRI, which cuts in the linker DNA, and then run on polyacrylamide native gels. If every repeat contained a bound octamer (100% saturation), digestion would yield just a single monosome band. If templates were subsaturated, the repeats free of nucleosomes would yield naked DNA fragments of the length of the EcoRI fragment (195 bp; note that there are two EcoRI sites in the 208 repeat). The ratio of monosome particle DNA to free DNA serves as a measure of the degree of saturation. As clearly seen in Fig. 7C
, lanes 1 and 2, the methylated and control templates were equally saturated with octamers. The average number of nucleosomes per nucleosomal array was calculated to be 10.9 in the case of the control array and 10.5 for the methylated array, in general agreement with AFM observations. Digestion with DraI, although leaving the majority of the material undigested (see above), gave similar amounts of free DNA fragments for the control and the methylated template (compare lanes 4 and 5). Finally, lanes 6 and 7 demonstrate the sensitivity of the assay. Reducing the ratio of octamers to DNA from 1:1 in lanes 1, 2, 4, and 5, to 0.8:1 in lanes 6 and 7 led to a noticeable increase in the amount of free DNA (the average number of nucleosomes per control array dropped from 10.9 to 7.4).
Thus, it is clear that the LH/DNA methylation-mediated compaction observed is not due to different loading of the temples with nucleosomes. Nor apparently is it due to a difference in LH binding to both types of arrays, since the LH content was the same in both cases (results not shown), in agreement with published reports (17)
.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
We then used an in vitro reconstitution approach to dissect the molecular determinants of the DNA methylation-dependent chromatin compaction. The results based on examination of AFM images of different reconstitutes suggested that DNA methylation causes compaction of the chromatin fiber only in conjunction with the binding of LH to the fibers. Each condition alone, DNA methylation or LH binding, is necessary but not sufficient to compact chromatin fibers. The AFM conclusions were further substantiated by two independent biochemical approaches: MNase digestion and sucrose gradient density centrifugation.
It is gratifying that the degree of compaction observed (expressed as number of nucleosomes per 10 nm of contour fiber length) was very close in vivo and in vitro: 0.46 for the control vs. 0.70 for the hypermethylated chromatin fibers from the in vivo experiments and 0.48 for the control vs. 0.67 for the methylated H1-containing reconstituted fibers. Some discussion is warranted here. In the in vivo study we compared ‘normal’ fibers with hypermethylated ones; the in vitro experiments compare unmethylated substrates with highly methylated ones. If the effect of DNA methylation is dependent on its density along the DNA template, these seemingly different in vivo and in vitro templates may not be so different after all: the scattered methylatable CpGs in the genome may provide low methylation density, insufficient to affect the structure of bulk chromatin. On the other hand, the in vivo hypermethylated chromatin fiber may resemble, in its methylation density, the methylated reconstitution substrate used in the in vitro experiments.
A second point concerns the role played by LHs in the DNA methylation-dependent chromatin compaction. We have demonstrated that LH binding is crucial for the compaction of methylated templates to occur. We must note that in the in vivo experiments there was no detectable difference in the amount of LHs present in the chromatin fibers isolated from control and treated cells, so that the in vivo compaction we observed may have involved LH binding. This statement may seem at odds with the results from an in vivo study on transcription from methylated templates microinjected in Xenopus oocytes (14)
. As pointed out by Bird and Wolffe (5)
, this methylation-dependent inhibition of transcription on chromatin templates occurred in the absence of the types of H1 normally associated with transcriptional repression. It is important, however, to bear in mind that other LH subtypes are present in the oocytes (59)
, and these may cooperate with DNA methylation to confer the compaction needed to repress transcription.
One last point concerns the lack of visible structural effect of histone H1 binding to control nucleosomal arrays (see Fig. 4C
). We have demonstrated that histone H1 does bind properly to the fiber under the conditions used (e.g., Fig. 3E
). The lack of visible effect on the fiber morphology may be because the reconstituted fibers are relatively short, only 12 nucleosomes in length. The reported zig-zag morphology of H1-containing fibers (39
, 60)
may remain undetectable in AFM images of such short fibers that may experience considerable distortion due to surface interactions and end effects. The lack of visible morphological effect of LH binding alone does not negate our main conclusion about the cooperation between LH binding and DNA methylation in producing chromatin fiber compaction, since it is based on positive evidence coming from both AFM and biochemical experiments.
In summary, our data indicate that the combined action of DNA methylation and LH binding is required to bring about chromatin compaction. This compaction may affect transcription of large chromatin domains. Compaction that affects specific gene transcription may require more complex interactions involving targeted binding of methyl-DNA binding proteins, histone deacetylation, and probably other mechanisms.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
2 On leave from the Research Institute of Physics, St. Petersburg State University, 198904 St. Petersburg, Russia.
3 On leave from the Institute of Biophysics, Academy of Sciences of the Czech Republic, 612 65 Brno, Czech Republic.
Received for publication May 2, 2001.
Revision received August 27, 2001.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  N. Gilbert, I. Thomson, S. Boyle, J. Allan, B. Ramsahoye, and W. A. Bickmore DNA methylation affects nuclear organization, histone modifications, and linker histone binding but not chromatin compaction J. Cell Biol., May 7, 2007; 177(3): 401 - 411. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. F. Kepert, J. Mazurkiewicz, G. L. Heuvelman, K. F. Toth, and K. Rippe NAP1 Modulates Binding of Linker Histone H1 to Chromatin and Induces an Extended Chromatin Fiber Conformation J. Biol. Chem., October 7, 2005; 280(40): 34063 - 34072. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 H. Dote, D. Cerna, W. E. Burgan, D. J. Carter, M. A. Cerra, M. G. Hollingshead, K. Camphausen, and P. J. Tofilon Enhancement of In vitro and In vivo Tumor Cell Radiosensitivity by the DNA Methylation Inhibitor Zebularine Clin. Cancer Res., June 15, 2005; 11(12): 4571 - 4579. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. T. Wierzbicki and A. Jerzmanowski Suppression of Histone H1 Genes in Arabidopsis Results in Heritable Developmental Defects and Stochastic Changes in DNA Methylation Genetics, February 1, 2005; 169(2): 997 - 1008. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 |
