ABSTRACT
We have characterized the structure of transcriptionally active nucleosome
subunits using electron spectroscopic imaging. Individual nucleosomes were
analyzed in terms of total mass, DNA and protein content, while the ensemble of
images of active nucleosomes was used to calculate a three-dimensional reconstruction. Transcriptionally active nucleosomes were
separated from inactive nucleosomes by mercury-affinity chromatography thus making it possible to compare their
structures. The chromatographic results combined with electron spectroscopic
imaging confirm that active nucleosomes unfold to form extended U-shaped particles. Phosphorus mapping indicated that the nucleosomal DNA
also underwent a conformational change consistent with particle unfolding. The
three-dimensional structure of the Hg-affinity purified nucleosomes determined using quaternion-assisted angular reconstitution methods unites and resolves
the different electron microscopic views of the particle and is concordant with
a sulphydryl-exposing disruption of the H3-H4 tetramer.
The canonical conformation of the nucleosome particle, as determined by X-ray crystallographic analysis of nucleosomes from transcriptionally
quiescent genes, is a disc, 110 Å in diameter and 57 Å in height (
1
-
3
). Its centre is a wedge-shaped protein octamer containing two each of the histones H2A, H2B, H3
and H4, around which 146 bp of DNA are wound in 1.8 negative superhelical
turns. The X-ray structure of the octameric protein core shows a centrally located (H3-H4)
2
tetramer which contains buried sulphydryls, flanked by two H2A-H2B dimers (
4
). The three subunits are assembled in the form of a left-handed protein super helix, with a pitch of 28 Å and form a solid object with a small central cavity. A
distribution of positive charges on the surface of the octamer appears as a
left-handed spiral at the calculated path of the DNA. A model of the nucleosome
core particle, based on this data, shows that the two DNA strands coincide with
the path of the histone positive charges, and that the central 12 turns of the
DNA double helix contact the surface of the octamer at repetitive structural
motifs (
5
). Contacts between the individual histones and the encircling DNA have been
mapped by chemical cross-linking experiments (
6
,
7
). Histones H3 and H4 are cross-linked to DNA within 30 bp on either side of the dyad axis. Histones H2A,
H2B and H3 are cross-linked to the DNA where it enters and exits the nucleosome. Both chemical
(
7
) and UV-laser cross-linking methods (
8
) show that the N-terminal domains of the histones make numerous contacts with the DNA coil
at the periphery of the nucleosome (
9
). The cysteine residues of the two histone H3 molecules lie within a few
Angstroms of each other at the centre of the core, where they are not
accessible to SH-reactive agents (
10
).
This compact structure would be expected to restrict the accessibility of DNA to
RNA polymerases and transcription factors. Although most transcriptionally
active genes are packaged into nucleosomes, they are organized into a less
condensed chromatin structure. This is achieved in part by a depletion of
histone H1 (
11
), which normally binds the DNA strand where it enters and exits from the
nucleosome core, and by a relaxation of histone-imposed constraints on DNA within the core particle. Active regions of
chromatin are highly accessible to hydrolytic cleavage by DNase I (
12
,
13
), DNase II (
14
,
15
), and micrococcal nuclease (
16
,
17
). Moreover, the sensitivity of a gene to micrococcal nuclease (MNase) is
related to the timing of its transcription. The MNase-sensitive domain does not include the non-transcribed DNA sequences flanking the gene, and this sensitivity is
lost when transcription stops (
16
,
17
). On the chemical level the individual histone proteins become hyperacetylated
in their N-terminal regions, resulting in a number of transcriptionally-associated changes in physico-chemical properties (
18
-
20
).
Whereas it is recognized that changes in DNA topology and altered histone-DNA contacts occur in active chromatin, the corresponding structural
basis for conformational changes of the nucleosome has not yet been fully
addressed. Elongated forms of the nucleosome core believed to be associated
with transcriptional activity have been detected by electron microscopy (
21
). Similarly, nucleosomes from transcribing ribosomal genes of
Physarum
seem to be unfolded into a so-called lexosome structure stabilized by non-histone proteins (
10
).
Since transcriptionally active nucleosomes can be separated specifically from
inactive particles by mercury-affinity chromatography, it is possible to analyze the active structure in
two and three dimensions without interference from inactive chromatin fragments
(
22
-
25
). To characterize the structure of nucleosomes associated with
transcriptionally active genes, we have chosen to use electron spectroscopic
imaging (ESI), a technique that permits high resolution structural studies and
elemental microanalysis (
26
-
32
). An advantage of this technique is its ability to quantify the characteristics
of individual nucleosomes, specifically their total mass, DNA and protein
contents. Moreover, it can be anticipated that structural changes will alter
the path of the DNA within the nucleosome. The ability to map the phosphorus
organization, and thereby the DNA path, makes ESI particularly sensitive for
detecting alterations in nucleoprotein structure. Though ESI requires a high
exposure of the specimen to the electron beam, the stability of DNA-protein complexes that have been so analyzed permits the collection of
images that have resolution in the 10-20 Å range (
29
,
32
). Other high resolution methods for studying highly modified nucleosome
particles that can provide such information are not currently available.
Following ESI, the three-dimensional structure of nucleosomes associated with transcriptionally
active genes was determined using image processing methods based on quaternion-assisted angular reconstitution. This technique permits the three-dimensional structures of biological macromolecules to be calculated
from randomly oriented particles (
33
-
37
). The results of Hg-affinity chromatography combined with this microscopical analysis confirm
that active nucleosomes are unfolded, forming extended U-shaped particles, consistent with a disruption of the H3-H4 tetramer. The changes in nucleosome conformation are consistent
with a model where only unfolding occurs, with no additional requirement for
displacement of the histone core (
38
).
COLO 320 DM cells, human adenocarcinoma cells (obtained from ATCC), were grown
in DMEM supplemented with 10% newborn calf serum (Gibco), 100 [mu]g/ml penicillin, 100 [mu]g/ml streptomycin, 10 [mu]g/ml gentimycin sulphate (Gibco). The cells were harvested in log
phase at a concentration of 10
5
cells/ml. Nuclei were prepared as described (
24
), and nucleosomes were released by a limited digestion with micrococcal
nuclease. To minimize further degradation of the unfolded monomers (
24
), conditions were modified to release only 3-5% of the total nuclear DNA into the S1 fraction. The released chromatin
fragments were fixed in 1% buffered, EM grade formaldehyde for 5 min at room
temperature in 50 mM Tris-HCl, 50 mM NaCl, 0.5 mM MgCl
2
, 1 mM [beta]-mercaptoethanol and 5 mM sodium butyrate and immediately applied to
a mercury-affinity column. After washing the column to elute the unbound
nucleosomes, the Hg bound nucleosomes were eluted in two steps, as described (
24
). Two samples of each nucleosome fraction were submitted for electron
spectroscopic imaging. One was not further treated, whereas the other was
treated with 50 [mu]g/ml RNase A for 1 h at 37oC. All samples were then dialyzed against a buffer of 10 mM Tris-HCl, 1 mM EDTA pH 8.0 at 4oC, and imaged by ESI.
Nucleosomes fractionated by a 2-step Hg-affinity chromatography procedure were spread onto a 2-3 nm-thick carbon support film supported on 1000-mesh copper grids. The carbon films were exposed
to glow discharge immediately before use. Linearized plasmid DNA was spread
with the nucleosomes to serve as an internal reference for both mass and
phosphorus analyses (
27
,
32
). A 4 [mu]l drop of nucleosomes and a 4 [mu]1 drop of DNA solution were applied together to the grid. The total
concentration of the chromatin and DNA was ~20 [mu]g/ml. If necessary, a final dilution on the grid surface was carried
out by injection of an equal volume of buffer into the chromatin/DNA droplet on
the grid surface. After 30 s the grid was washed with buffer to remove excess,
unbound sample, followed by a wash with distilled H
2
0. Excess was removed with filter paper, after which the specimen was air dried.
To provide a dose-sparing darkfield-like image, grids were examined at about 30 eV energy loss.
Phosphorus-enhanced images were recorded by imaging with electrons corresponding to
the phosphorus ionization peak at 155 eV energy loss. Reference images were
recorded before the phosphorus edge at energy losses of 105 and 120 eV.
Micrographs were recorded on Kodak SO-163 electron image film at a magnification of 13 000 or 20 000 *. The film was developed for 10 min in full-strength Kodak D-19 developer. Micrographs were digitized on a
computerized image capture system and analyzed (
39
). The digitized pixel size corresponded to 10 [mu]m on the film, permitting a number of nucleosomes to occupy a field, usually
also with a portion of a molecule of control DNA. Background regions were used
to normalize the reference and the phosphorus-enhanced images in order to offset slight differences in micrograph
exposure density. Differences greater than 10% resulted in rejection of the
image sets to avoid potential non-linearity problems between the films. In order to delineate the objects,
the average background value was chosen as a threshold for the reference and
phosphorus enhanced images. Objects were then outlined interactively and the
integrated optical densities were calculated. The integrated optical density of
an object is relatively insensitive to the size of the delineation mask in this
procedure. An integrated optical density of a nucleosome and a defined length
of linearized plasmid DNA are both obtained from the 120 eV reference image
which is essentially a `mass' image, obeying Beer's law. Knowing the length of
the DNA and its molecular weight and assuming an average molecular mass of 650
Da per base pair and an experimentally determined value of 3.4 Å/bp, the integrated optical densities were converted to mass, so that by
comparison, the mass of the nucleosomes could be estimated. The same masks, for
the standard DNA and for the nucleosomes, were then used to calculate the
phosphorus content of the same nucleosomes in the net phosphorus images (
39
). To calculate the mass and phosphorus content on either side of an apparent
plane of symmetry in the Hg-affinity column bound fraction, masks were created on both sides of the
chosen plane. To determine the symmetry plane, the particle boundary was
delineated at a background value plus one standard deviation measured over the
background. The two longest axes were drawn by hand using a mouse and a line
joining the midpoints of these lines or a line orthogonal to it at its midpoint
formed the axis of symmetry. In some cases the line through the bisection of
the angle at the point of intersection of the two longest axes created a
symmetry axis. To minimize error from the determination of this plane, average
values of mass and phosphorus were obtained by moving the line ~10 Å to either side of the chosen plane.
Three-dimensional reconstructions were calculated using an approach based on
quaternion-assisted angular reconstruction (
33
-
37
). The method permits the three-dimensional reconstruction of biological macromolecules and their
complexes from sets of images of single particles at random orientations. It
has been used to determine the three-dimensional structures of proteins (
35
,
36
). Using ESI images of the nucleosome purified from inactive chromatin
fractions, the method has been applied to calculate the three-dimensional structure of different conformational states of the particle
with respect to different ionic environments (
40
). In the current study, ESI images of Hg-affinity purified nucleosomes were selected on the basis of their
molecular mass. Outliers with a mass of more than one standard deviation from
the mean were excluded. In order to obtain images with a higher signal to noise
ratio, three-dimensional image processing was carried out using the phosphorus contrast
enhanced images acquired at 155 eV energy loss. Three-dimensional reconstructions were calculated as before (
35
,
36
). The quaternion-assisted angular reconstitution approach (
36
) is based upon the principle of common lines (
37
). This principle states that projection images of a particular macromolecule at
different orientations share common lines of integrated intensity. By
determining these common lines it is possible to determine the relative angular
orientations amongst many projection images. The method is robust, has been
tested with simulated images and a macromolecule of known structure and in
protein structure determination (
34
-
36
).
In the reconstruction process, 150 nucleosome images were used. This is a
greater than sufficient number to permit 3D reconstructions to be calculated at
high signal-to-noise ratios and at resolutions of 15-30 Å, which permit gross nucleosome conformation to be
determined (
36
). To determine the resolution, the set of 150 images was split randomly into
two sets of 75 images and two independent reconstructions were determined. A
phase residual approach was used to determine the level of consistency between
the two reconstructions, a measure of resolution. A phase residual based
approach provides a measure of average structural overlap at particular spatial
frequencies and is effectively a consistency measure between features of
specific sizes in the reconstructions. This method and comparisons to other
methods is extensively discussed elsewhere (
36
). At a phase residual of 60o, a conservative measure of average overlap or resolution, the
reconstructions correlated to 30 Å. At a phase residual of 90o, the point at which average overlap no longer occurs, the
reconstructions correlated to 18 Å. This latter measure is equivalent to the Fourier ring correlation with
a cut-off at two standard deviations of the noise. These numbers are consistent
with other studies carried out using the quaternion-assisted angular reconstruction approach (
36
,
40
).
The ability to separate transcriptionally active from inactive nucleosomes by
mercury affinity chromatography makes it possible to compare their structures
without the confounding characteristics inherent in a mixture of active and
inactive chromatin fragments.
Typical conformations of the nucleosomes in the Hg column flow-through and the DTT eluted fractions imaged in energy loss mode are shown
in Figure
1
. The length:width ratios of the particles show that nucleosomes of the mercury-bound fractions are considerably extended relative to the compactly beaded
nucleosomes of the unbound fraction (Table
1
). The length:width ratios of the nucleosome fractions eluted in 0.5 M NaCl and
in 20 mM DTT, relative to the unbound nucleosomes, are compatible with an
unfolding mechanism that involves a relaxation of constraints on the DNA coiled
around the histones of the nucleosome core. This view is supported by the
differences in morphology of the particles from the unbound and Hg-bound fractions.
The nucleosomes released from the mercury-column in 0.5 M NaCl were not all compactly beaded (not shown). Instead
their images revealed a mixture of three classes of chromatin subunits: (i)
approximately 27% resembled toroids (circular profiles with reduced mass in the
centre); (ii) 50% had U-shaped or elongated profiles; and (iii) 23% had a distorted or `swollen'
circular profile with a diameter 10-20% greater than the toroids of the unbound fraction but without reduced
mass in the centre. The expanded state of this Hg-bound nucleosome is not surprising, in light of the evidence that all thio-reactive nucleosome fractions have consistently shown very high
levels of histone acetylation (
22
-
25
,
41
-
43
) and that hyperacetylation weakens and shifts histone-DNA interactions at the centre of the nucleosome core (
8
,
44
). Most of the U-shaped particles of the 0.5 M NaCl wash fraction have DNA contents in the
range expected for mononucleosomes of COLO320 cells (average = 158 bp) (Fig.
2
C). They exhibit, however, a larger spread in the protein mass distribution in
comparison to control nucleosomes (Fig.
2
C), and the average protein mass, 88 kDa, falls below that expected for a
complete histone octamer (108 kDa).
We are confident that the elongated or U-shaped profiles are not the result of exposure to 0.5 M NaCl. First, the
exposure of the particles to fixative before the chromatography should prevent
unfolding. Nevertheless, to test whether 0.5 M NaCl alone could lead to a
detectable disruption of the canonical structure, we exposed a sample of the
pre-fixed, flow-through nucleosome fraction to 0.5 M NaCl. The appearance of these
particles in the subsequent ESI analysis was indistinguishable from the same
fraction not exposed to this salt concentration (Fig.
1
C). We conclude, therefore, that the elongated or U-shaped appearance of the column bound, salt eluted fraction is not due to
exposure to 0.5 M NaCl.
Unfolded particles were even more prevalent in the DTT-eluted nucleosome fraction eluted with one step, without exposure to 0.5 M
NaCl. Approximately 62% of the particles were elongated or U-shaped (Fig.
1
B). Conversely, only 13% of the DTT-eluted fraction appeared toroidal and 25% had a distorted or `swollen'
profile. The presence of the U-shaped nucleosomes in the DTT-eluted fraction is of particular interest, since that fraction
consists only of nucleosomes linked to the Hg-column through accessible thiols of histone H3 (
24
,
43
). It is not necessary, therefore, for the nucleosome to unfold completely to
reveal the H3-thiol groups known to be located at the centre of the core (
10
). Again, we argue that it is not likely that either the nuclease digestion or
column buffers are responsible for the unfolded appearance of the column-bound fractions for the following reasons. First, the nuclease digestion
buffer has an ionic strength close to the physiological level. Secondly,
fixation before the Hg column would make further subsequent disruptions
unlikely. And thirdly, even if the column buffer (10 mM Tris, 25 mM NaCl, 25 mM
KCl, 2% sucrose, 5 mM sodium butyrate) was capable of causing a structural
transition, exposure to it did not unfold the control fraction nucleosomes.
They were compact, whereas the column-bound nucleosomes were elongated. Because the bound and unbound fractions
were treated identically, the different appearance of particles in these
fractions indicates an underlying structural difference and is not due to
exposure to particular buffers or to other steps in the imaging process.
Estimates of the DNA and protein content of the particles in this bound
fraction revealed an average DNA content of 155 +- 40 bp and a protein content of 111 +- 28 kDa (Fig.
2
B).
A three-dimensional reconstruction for the DTT eluted nucleosomes (Fig.
3
) reveals an open or U-shaped conformation. The central core region of the nucleosome, normally
inaccessible as indicated by X-ray crystallography (
4
,
5
), is open and accessible in this three dimensional reconstruction. This is
consistent with biochemical observations of internal sulphydryl accessibility
in nucleosomes associated with transcriptionally active genes. One half of the
reconstruction also appears more massive than the other, consistent with mass
measurements of individual nucleosomes given below. This is evident in the
lower portion of the reconstruction shown in Figure
3
. Profiles of the reconstruction are consistent with the images showing U shaped
and extended projections.
The reconstruction described here was determined using 150 images of nucleosomes
selected on the basis of their relative molecular mass. Two additional
reconstructions were calculated from independent sets of 75 images. The 75-image reconstructions were similar to the 150-image reconstruction, appearing U-shaped and elongated (not shown). The use of a phase residual
based approach, a measure of structural overlap, to compare these two
reconstructions indicated a resolution of 30 Å at a phase residual of 60o, a conservative measure of resolution. This approach also indicated
a resolution of 18 Å at a phase residual of 90o, the limit at which structural detail exhibits zero overlap. A more
detailed discussion of these measures is given in reference
36
.
A reconstruction has not yet been calculated for the particles of the flow-through eluate of the column, which corresponds to transcriptionally
inactive nucleosomes. A reconstruction of transcriptionally inactive
nucleosomes from calf thymus nuclei has been calculated, indicating a closed
compact structure, consistent with the crystallographic form of the
macromolecular complex (
40
). We have, however, compared length:width measurements and principle component
analysis (
45
) results of the flow through fraction and calf thymus-derived nucleosomes. Both populations are nearly indistinguishable in the
correspondence of their projections to that of an oblate ellipsoid (data not
shown). We, therefore, expect that the reconstruction of the flow-through fraction will be similar to that of the calf thymus nucleosomes,
indicating a closed, compact structure.
ESI permits the visualization of the phosphorus distribution in nucleosome
particles, from which the path of the DNA can be inferred (Fig.
4
) (
29
). Particles with high length:width ratios, such as those seen in Figure
1
B, always show an extended DNA distribution. Because of the variety of
projections that are possible, even particles with a lower length:width ratio
can have a DNA distribution that indicates an unwinding of the supercoil. In
some of the images, the DNA profile is U-shaped (Fig.
4
A, rows 1 and 2). Phosphorus maps of many of the more elongated chromatin
subunits strongly support the suggestion that the DNA around the U-shaped and elongated chromatin subunits has been stretched like the coil
of a spring to form an S shaped projection of the DNA (Fig.
4
A, rows 3 and 4). In this state, it is not surprising that the associated
histone H3-thiol groups were readily accessible for binding to the Hg-column. As stated earlier, a small fraction of the particles have a
compact circular profile similar to those in the control fraction. In fewer
than one-third of these, the phosphorus or DNA appears to be confined to a
trajectory at the periphery of the protein core (Fig.
4
A, row 5), indistinguishable from those projections that were regularly observed
with the control nucleosome particles (Fig.
4
B).
To characterize these elongated particles further, we estimated the DNA and
protein content of each half of a number of particles that had an apparent
mirror plane of symmetry in the mass-sensitive images. The results are summarized in Table
1
. Representative measurements made from images of individual nucleosomes are
given in Table
2
. The mass in each half of a nucleosome was measured and designated as the high
or low mass side. The phosphorus content of each half was measured. On average,
the amount of DNA on each side was equivalent. In contrast, the ratio of the
DNA to the protein mass on the low mass side of the Hg-bound particles, was ~20% higher than that on the high mass side. This indicates that a
loss in protein may be the cause of the altered DNA to protein ratio on the low
mass side. Variability was observed, however, between individual particles
(Table
2
). For example, the three particles analyzed in Table
2
each had 80 +- 5 bp DNA on each side of the mirror-symmetric plane. For particle 1, no significant difference in DNA
to protein content from that obtained for control particles was observed on
either side of the mirror-symmetric plane. For particles 2 and 3, however, there is a higher protein
to DNA ratio on the high mass side and a lower protein to DNA ratio on the low
mass side than that obtained for control particles. Since we cannot identify
proteins with this analysis, we cannot determine whether altered protein to DNA
ratios are due to altered core histone levels or to the presence of other
factors. Electrophoretic analyses have shown that the Hg-bound nucleosomes contain all four histones in stoichiometric proportions
(
24
,
42
,
43
). This, however, is a bulk property of all of the particles in the fraction. In
contrast, the analysis presented in Table
2
indicates that variability in DNA:protein stoichiometry between individual
particles and between symmetrical halves of particles is observed. But even
this variability is not reflected in the average protein and phosphorus levels
of the entire fraction nor would this variability be detectable in Coomassie-stained protein gels or gels to assay DNA length.
REFERENCES
Return