The histone chaperone DAXX maintains the structural organization of heterochromatin domains

H3.3 is enriched in a fraction of H3K9me3-enriched domains

The discovery of DAXX as an H3.3-specific chaperone responsible for its deposition
in telomeres and pericentric heterochromatin 19], 25] indicated that H3.3 was not restricted to active regions of the genome. To determine
the DAXX-dependent nuclear localization of endogenous H3.3, we examined wild type
and DAXX null MEFs using immunofluorescence microscopy (Additional file 1). We observed that, within the resolution limit of fluorescence microscopy, H3.3
correlated strongly with the periphery of a fraction of H3K9me3-enriched domains but
not significantly with others (Fig. 1a). Line scan intensity plots show H3.3-associated and non-associated H3K9me3-enriched
domains from wild type and DAXX null cells. In this context, association is defined
as overlap of the H3K9me3 and H3.3 fluorescence signals. The percent of H3K9me3-enriched
domains associated with H3.3 is slightly greater in DAXX null cells with a mean of
74.8 (SE 1.5), yet only 60.4 (SE 1.1) in the control wild-type cells (Fig. 1b). This is not due to changes in the total number of chromocentres as the average
number per nucleus are similar in both cell lines, with means of 30.5 (SE 5.1) in
wild type and 33.0 (SE 5.1) in DAXX null cells (Fig. 1c). Western blot analysis of whole-cell lysates (Fig. 1d, e) showed that the expression levels of H3.3 and H3K9me3 are similar in both cells
lines. As a control for epitope accessibility within compact chromatin domains, wild
type and DAXX null cells were transiently transfected with a FLAG-H3.3 plasmid and
immunolabeled with anti-FLAG primary antibodies (Additional file 2A). The sub-nuclear distributions of both endogenous and over-expressed H3.3 were
equivalent, and line scan intensity plots again revealed that FLAG-H3.3 remained peripherally
associated with H3K9me3-enriched domains; only a fraction of chromocentres were associated
with FLAG-H3.3. The forced expression of H3.3 is thus not sufficient to drive H3.3
into the core of constitutive heterochromatin domains, but rather remains peripherally
associated with H3K9me3-enriched regions. The peripheral association of H3.3 with
H3K9me3-enriched domains was confirmed using correlative LM/ESI and immunogold labeling
of H3.3 (Additional file 2B). We did not observe H3.3 in the core of the compact chromatin domains, rather it
was found at the periphery of the chromocentres and in the euchromatic space.

Fig. 1. The association of H3.3 with H3K9me3-enriched chromatin is increased in the absence
of DAXX. a Wild type and DAXX null fibroblasts immunolabeled for H3.3 (red) and H3K9me3 (green). Line scan intensity plots of H3.3 and H3K9me3 for magnified chromocentres, H3.3
associated (I, III) and non-associated (II, IV) chromocentres. Scale bar 5 µm. b Quantification of the percent association of H3.3 and H3K9me3-enriched chromatin
domains in wild type (n = 52) and DAXX null (n = 54) cells. Data calculated as the mean percent of association per cell. c Quantification of the average number of chromocentres (H3K9me3-enriched foci). Error bars represent SEM from three independent experiments. d, e Western blot analysis of whole-cell lysates of H3.3 (d) and H3K9me3 (e)

DAXX maintains the structural integrity of pericentric heterochromatin domains

We next sought to determine if the significant increase in association of H3.3 with
the periphery of H3K9me3-enriched domains observed in the absence of DAXX (Fig. 1a, b) affected the underlying chromatin structure of chromocentres. Therefore, wild
type and DAXX null MEFS were prepared for correlative LM/ESI with antibodies specific
to the H3K9me3 histone modification to mark pericentric constitutive heterochromatin
domains. Physical sections on EM grids were imaged for fluorescence microscopy (Fig. 2a, top left panel) prior to imaging by ESI. The H3K9me3-positive regions were identified
by correlation to the fluorescence image (Fig. 2a, bottom left panel) and imaged (Fig. 2a, arrowhead and white square). ESI micrographs of the H3K9me3-enriched domain were
obtained and the approximate boundaries of the H3K9me3 signal outlined by dashed lines
(Fig. 2a). The process was repeated for a total of 74 H3K9me3-enriched regions from wild
type and 95 such regions from DAXX null cells. Since some centric and pericentric
regions of the genome can be found adjacent to the nucleolus 11], 28], the H3K9me3-enriched regions were classified as either non-nucleolar or perinucleolar
(nucleolar-associated) heterochromatin.

Fig. 2. Disruption of chromocentres in the absence of DAXX. Correlative LM-ESI micrographs.
H3K9me3 fluorescence (left panel, top) overlaid on the mass sensitive image generating the correlative image (left panel, bottom). High magnification ESI micrograph of the region shown in the white box (correlative image) of the corresponding H3K9me3-containing structure (arrowhead). Approximate boundaries of the H3K9me3 region are indicated by a dashed line. a Non-nucleolar and b perinucleolar heterochromatin domain of wild-type cell. c–g H3K9me3 correlative LM-ESI micrographs of DAXX null cells. (c, e, and g) Non-nucleolar and (d and f) perinucleolar heterochromatin structures. In all ESI images, chromatin is represented
by levels of yellow and protein-based structures as cyan. Nu nucleolus. Scale bar 0.5 µm. h Quantification of the number of typical non-nucleolar and perinucleolar H3K9me3-containing
heterochromatin domains in wild type and DAXX null cells

Consistent with previous reports, typical chromocentres from wild-type cells are radially
symmetric structures that are comprised of compact chromatin, occupy discrete regions
within the nucleus, and are readily discernible from the surrounding euchromatin (Fig. 2a) 20], 23]. Furthermore, the boundaries of the chromocentres, as approximated by the H3K9me3
fluorescence signal, correlate strongly with the regions of compact chromatin. A similar
correlation between H3K9me3-enrichment and compact chromatin can be observed in perinucleolar
heterochromatin domains (Fig. 2b). Importantly, despite the close spatial associations between H3K9me3-enriched chromatin
and the nucleolus (Nu), the two regions form discrete structures. What should be appreciated
in Fig. 2b is the clear “respect” for spatial boundaries, where both structures occupy discrete
regions, albeit immediately juxtaposed. As such, the following four characteristics
were used as criteria to classify non-nucleolar and perinucleolar-associated (adjacent)
heterochromatin domains as either typical or atypical (1) compact chromatin (2) radial
symmetry, (3) enriched in H3K9me3, and (4) sharp boundaries with other nuclear structures
(i.e., surrounding euchromatin and nucleoli). Regions that met all four criteria were
classified as typical, and those that did not, atypical. Mitotic cells, including
those in early prophase were excluded from the analysis. With either fluorescence
microscopy or ESI it is very easy to identify a cell in early prophase (and metaphase)
by virtue of the chromatin/chromosome morphology. Therefore, the compaction observed
through the loss of DAXX is not related to the chromosome condensation associated
with mitosis.

We found the majority of non-nucleolar (87.5 %) and perinucleolar (88.5 %) chromocentres
from wild-type cells were typical in nature (Fig. 2a, b). The H3K9me3-enriched regions (Fig. 2a, b; dashed lines) correlated with compact chromatin and were discernible from both
the surrounding euchromatin and from the nucleolus. In contrast, only 23.0 % of non-nucleolar
and 26.7 % of perinucleolar chromocentres from DAXX null cells were typical (Fig. 2c–g). Figure 2 shows representative images of the range of chromatin-related phenotypes observed
in non-nucleolar (c, e, and g) and perinucleolar (d and e) chromocentres caused by
the loss of DAXX. The chromocentre shown in Fig. 2c contains levels of non-chromatin protein-based structures not typically observed
in chromocentres, even though the radial nature of the structure was retained. In
this example, the H3K9me3 signal does not span the entire area of compaction of the
non-nucleolar domain. Similarly, a loss of spatial correlation between H3K9me3-enrichment
and compact chromatin can be seen in the non-nucleolar chromocentre of Fig. 2e. In this example, the H3K9me3 signal shows two regions of enrichment (arrowhead,
dashed lines), yet a third connected domain of compact chromatin can be discerned
that is depleted in the mark. Despite the absence of H3K9me3 in this region, the domain
contains compact chromatin and can be readily distinguished from the surrounding chromatin.
The example in Fig. 2g shows another phenotype characterized by large regions of compact chromatin interspersed
with protein-based structures and compact chromatin that is no longer restricted to
H3K9me3 domains; no morphologically distinct boundaries exist between the marked and
unmarked compact chromatin.

We also observed a loss of spatial boundaries between perinucleolar heterochromatin
and the nucleolus in the absence of DAXX (Fig. 2d, f). For example, in Fig. 2d, not only do the three regions of H3K9me3 signal (dashed lines) correlate with ranges
of chromatin compaction, but the distinction between the nucleolus and the surrounding
heterochromatin is lost. The chromocentre shown in Fig. 2f is found between two nucleoli. Although compact chromatin domains containing H3K9me3
labeling are observed, they contact a compact chromatin domain that does not correlate
with the H3K9me3 modification. The two domains of compact chromatin, one containing
H3K9me3 and one not, can be distinguished morphologically in the phosphorus distribution
image. Moreover, the discrete boundary between compact chromatin and the nucleolus
is lost. A quantification of the percentage of typical non-nucleolar and perinucleolar
H3K9me3-enriched chromatin domains in the control and DAXX null cells is shown in
Fig. 2h.

The observed phenomena are not cell-cycle dependent as different phenotypic classes
were sometimes observed in a single cell. We conclude that in the absence of DAXX,
the compact chromatin-containing chromocentres, both non-nucleolar and perinucleolar,
are no longer restricted to H3K9me3 boundaries. As well, the spatial boundaries between
perinucleolar heterochromatin domains and the nucleolus are severely compromised.

Aberrant spatial relationships of H3K9me3-enriched chromatin and major satellite DNA

We further sought to determine whether H3K9me3-enriched chromatin domains that form
in the absence of DAXX (Fig. 2) maintained the typical signatures of constitutive heterochromatin. Wild type and
DAXX null cells were immunolabeled with antibodies specific to the repressive histone
modifications H3K9me3 and H4K20me3 and the heterochromatin-associated protein HP1
6], 10], 40]. We found that chromocentres from both cell lines are enriched in H3K9me3, H4K20me3,
and HP1 (Additional file 3A, B).

Given the enrichment of major satellite DNA in chromocentres 28], we determined whether that association was conserved in the absence of DAXX. Using
antibodies against H3K9me3 and DNA FISH probes against major satellite DNA, and consistent
with previous reports 28], we observed that chromocentres from wild-type cells contain a core of major satellite
repeat DNA packaged as H3K9me3-enriched chromatin (Fig. 3a). Line scan analysis of representative chromocentres demonstrates strong correlations
in intensity peak profiles between the H3K9me3 and major satellite signals. Independent
of the direction of the line scan, the major satellite signal is contained with radial
symmetry in relation to the H3K9me3 area. In DAXX null cells, however, while approximately
half of the chromocentres display the typical spatial relationship between H3K9me3
and major satellite DNA observed in control cells, approximately 50 % display aberrant
spatial relationships (Fig. 3b). In these regions, major satellite DNA peak profiles are no longer centered on
the core of the H3K9me3-defined region but extend asymmetrically beyond the H3K9me3
distribution. The mean percentage of the typical chromocentres in DAXX nulls (centrally
located major satellite DNA) is 63.5 (SE 10.3, n = 1301) compared to wild-type cells where 92.4 (SE 7.6, n = 1222) chromocentres display typical H3K9me3 and major satellite spatial relationships
(Fig. 3c). We conclude that pericentric satellite repeat DNA becomes uncoupled from its association
with the H3K9me3 histone mark when DAXX is absent. Taken together, these data led
us to conclude that DAXX plays a role in maintaining spatial relationships between
compact chromatin, compact chromatin biochemically marked as constitutive heterochromatin
by H3K9me3, and major satellite repeat DNA.

Fig. 3. Aberrant spatial relationships between H3K9me3 and major satellite DNA. FISH of major
satellite DNA (red) and immunofluorescence (IF) microscopy of H3K9me3 (green) of wild type (a) and DAXX null (b) fibroblasts. Magnified chromocentres are marked by white boxes in the merged images. Two independent line scan intensity plots are shown for each
enlarged chromocentre. Dashed solid arrows indicate the line scan plot and direction of the first (left) and second (right) histograms, respectively. Scale bar 5 µm. c Quantification of the percentage of typical versus atypical H3K9me3-enriched chromatin
domains from each cell line. Error bars SEM

Increased number of cells containing mini nucleoli

A structural relationship exists between centric and pericentric heterochromatin and
the nucleolus 11], 28], 29], 49]. In Su(var)3–9 mutants that lack H3K9me chromatin, the cells displayed disorganized
nucleoli 49]. We therefore wanted to determine if the observed disruptions in the organization
of heterochromatin in the absence of DAXX, including the frequently observed loss
of a discrete boundary between perinucleolar heterochromatin and nucleoli, result
in changes in the structural integrity of the nucleolus. Wild type and DAXX null cells
were labeled with antibodies against B23 (Fig. 4a), a protein enriched in the granular component (GC), and also found in the dense
fibrillar component (DFC) of nucleoli 7], 8]. We observed a similar number of large nucleoli with a mean of 6.0 (SE 0.15) in wild
type and 5.6 (SE 0.13) in DAXX null fibroblasts. However, we also observed, in some
cells, between 1 and 9 very small accumulations of B23 which we refer to as “mini
nucleoli” (Fig. 4a, arrowheads). Since the GC forms the outermost region of the nucleolus, B23 distribution
has a ring-like pattern. Line scan analysis of intact nucleoli shows the doublet pattern
of B23 intensity whereas mini nucleoli, due to their small size, appear as singlets
in the line scan intensity plots. Using this feature of B23, the quantification revealed
that 83 % of DAXX null cells contain one or more mini nucleolus in contrast to the
15 % of wild type cells having at least one mini nucleolus (Fig. 4b).

Fig. 4. DAXX maintains the structural integrity of nucleoli and the organization of rDNA.
a B23-labeled IF images. Arrowheads indicate the mini nucleoli. Arrows indicate the line scan intensity plot and direction. Scale bar 5 µm. b Quantification of the percentage of cells containing a minimum of one mini nucleolus.
Error bars represent SEM of a minimum of 100 cells. c FISH of rDNA (red) and IF microscopy of B23 (green). Arrowheads indicate rDNA foci found outside of the B23-defined nucleolar boundaries. Scale bar, 5 µm. d Box plot of the fraction of rDNA foci found outside of the nucleolar boundaries

Dispersed rDNA genes in the absence of DAXX

Ribosomal DNA (rDNA) genes associate with the cytogenetically discrete nucleolus organizer
regions (NORs) 45]. As nucleoli form from NORs, it is assumed that rDNA is sufficient to establish a
functional nucleolus 27], 38]. Since perinucleolar heterochromatin is intimately linked to rDNA gene regulation
and stability 29], 49], we wanted to determine if the organization of rDNA genes was disrupted in the absence
of DAXX. To test this, we performed an immuno-FISH experiment using antibodies against
B23 as a marker for the nucleolus and FISH probes against rDNA genes (Fig. 4c). In wild-type cells, rDNA repeats are clustered and localized within the confines
of the nucleolus as visualized by B23. In the absence of DAXX, however, we observed
an increase in the fraction of rDNA foci localized outside the nucleolar boundaries
(Fig. 4c, d). Taken together, these data demonstrate that DAXX-dependent heterochromatin
organization and the structural integrity of the nucleolus are intimately linked.
Furthermore, we conclude that the observed increase in the number of mini nucleoli
is likely caused, in part, by the dispersal of rDNA genes.

DAXX-dependent chromatin accessibility

We showed that the loss of DAXX has consequences, at a local level, on subnuclear
structures (Figs. 2, 3, 4). Next, we aimed to determine if there are global changes in chromatin structure
in the absence of DAXX. To address this, nuclei from wild type and DAXX null cells
were subject to micrococcal nuclease digestion and the isolated DNA was separated
by agarose gel electrophoresis (Fig. 5a). From the earliest time points, DNA from DAXX null cells was significantly more
sensitive to micrococcal nuclease compared to wild-type cells such that mononucleosomal
DNA was evident at 3.5 min in DAXX null cells, whereas in wild-type cells, this level
of digestion did not occur until 7.5 min. To quantify this observation, the rate of
loss of high molecular weight chromatin was calculated by measuring the signal ?1.5 kb
and the total signal in the lane for the 3.5–15 min time points 24]. Each ratio was normalized to the 1 min time point and plotted as a time course (Fig. 5b). The rate of digestion of high molecular weight chromatin was indeed higher in
the absence of DAXX, leading us to conclude that the loss of DAXX not only caused
local changes in pericentric heterochromatin and rDNA organization but increased global
chromatin sensitivity to micrococcal nuclease digestion.

Fig. 5. DAXX null cells are more sensitive to micrococcal nuclease digestion. a Micrococcal nuclease digestion of wild type and DAXX null cells. DNA was isolated
from extracted nuclei after digestion for the indicated time (T) in minutes and subjected
to agarose gel electrophoresis. b The rate of loss of high molecular weight chromatin