(The FASEB Journal. 2001;15:1694-1703.)
© 2001 FASEB
A view through the clouds of imprinting
JASON L. BURNS*,
DEAN A. JACKSON
and
A. BASSIM HASSAN*1
* Department of Zoology, University of Oxford, Oxford, UK OX1 3PS; and
Department of Biomolecular Sciences, University of Manchester Institute for Science and Technology, Manchester, UK M60 1QD
1Correspondence: Department of Zoology, University of Oxford, South Parks Road, Oxford, UK. OX1 3PS. U.K. E-mail: bass.hassan{at}zoo.ox.ac.uk
 |
ABSTRACT
|
|---|
The purpose of this review is to examine whether our current knowledge
of the higher order control of gene expression and nuclear organization
can help us understand the mechanisms of genomic imprinting. Imprinting
involves the inheritance of a silenced allele of a gene through either
a paternal or maternal germline. We have approached the problem of
imprinting using a model based on the dynamic attachment of chromatin
loops to immobilized RNA polymerases and control elements. We have
combined the information from different experimental approaches,
examining primarily the IGF2-H19 locus, in an attempt to
simplify the complexity of the imprinting data that has accumulated. It
is hoped that a unified model may generate predictions amenable to
experimental testing and contribute to the interpretation of future
experiments.—Burns, J. L., Jackson, D. A., Hassan, A. B. A view through the clouds of imprinting.
Key Words: transcription cycles • transcription factories • insulin-like growth factor II • chromatin • boundary
|
INTRODUCTION
|
|---|
ALTHOUGH THE PROTEIN complexes (holoenzymes) that form
during transcription initiation, elongation, and RNA
processing are being extensively characterized in vitro, our
understanding of the long-range interactions that determine the overall
efficiency of transcription are less well understood in vivo. Key
determinants of transcriptional initiation and reinitiation include
promoters, enhancers, locus control regions (LCRs), and those involved
in the organization of chromosomal and nuclear context. Position effect
variegation is perhaps the best-known phenomena, where the chromosomal
position of a gene and promoter element can influence its overall
expression (1
, 2)
. One reason why chromosomal context can
have such a dominant effect relates to the position of the element in
relation to the competing effects of ‘open’, transcriptionally active
chromatin with ‘condensed’ transcriptional silent chromatin. Within
the context of the nucleus of the living cell, the degree of
condensation of chromatin can be thought of as different ‘cloud’-like
densities of chromatin loops depending on the degree of compaction.
Even though the mechanisms and interactions that govern the processes
of gene expression now appear more complex, the theme emerging
from experimental investigation is that of a dynamic regulated system.
|
CHROMATIN LOOPS, IMMOBILE POLYMERASES, AND CHROMOSOMAL CONTEXT
|
|---|
The higher order structure of chromatin is critically important
because it determines the accessibility of DNA to the processes of
replication, recombination, transcription, and repair. There are
multiple superfamilies of chromatin proteins that modify gene
expression and chromatin condensation, the details of which are beyond
the scope of this discussion (see refs 3
4
5
). Within the
spatial confines of the nucleus, chromatin is organized into loops
attached to the core structure of the chromosome, which can then
unravel into 30 nm fibers during interphase (6)
. The loops
are attached at their base to a protein substructure termed either the
nucleoskeleton, nuclear scaffold, or nuclear matrix (7)
.
Although interactions between chromatin and the substructure may arise
during experimental manipulation, attachments identified in vitro are
generally of functional importance. Such regions have been termed
matrix attachment regions, scaffold attachment regions, or base of
loops (8
, 9)
. We will refer to these collectively as loop
attachment regions (LARs). The current evidence from investigation of
base of loop libraries and transient episomal reporter gene expression
is that the LAR sequences map mainly to transcription units (7
, 8
, 10
, 11)
.
Transcription and RNA processing are performed by very large protein
complexes that are likely to be immobile structures within the gel-like
nucleoplasm (12)
. If the RNA polymerase tracks along the
template DNA dragging its RNA behind, as depicted in most models of
transcription, then the RNA molecule will be wound around DNA every 10
bp. If RNA secondary structures form cotranscriptionally, then the
complex is likely to result in a knot of protein and nucleic acid. For
genes several kilobases long, the energy required to untangle RNA from
DNA would appear to be a costly topological puzzle. The entwining
problem for the cell can be sidestepped if the polymerase is made
immobile and topoisomerases remove supercoils generated in the template
DNA. The experimental evidence for immobile polymerases comes from a
variety of observations (reviewed by Cook, ref 13
).
Perhaps the most compelling is the observation that active RNA
polymerases appear to concentrate in sites, termed factories, the
largest factory being the nucleolus (12
13
14
15)
. Labeling of
nascent RNA in nuclei produces multiple foci when visualized with light
microscopy. However, foci represent collections of nascent RNA around
multiple polymerases (Fig. 1
).
View larger version (27K):
[in this window]
[in a new window]
|
Figure 1. Chromatin clouds within mammalian nuclei. Mammalian chromosomes occupy
discrete ‘territories’ that can be visualized by fluorescent in situ
hybridization with chromosome-specific probes or indirect
immunofluorescence of territories labeled with Br-deoxyuridine
(A; ref 94
). In mammals, such territories
are typically  2.5 µm across and occupy a volume of 8
µm3; diploid mammalian nuclei are roughly 10 µm across
and 500 µm3 in volume. Each territory is composed of
100 generally 400–800 nm chromatin ‘superbeads’ strung together.
These subchromosomal structures are believed to correspond to
replication foci that represent fundamental subunits of chromosome
structure (95)
. A model (cross section) of the
architecture of one such focus is shown (B). This
transcriptionally active example contains 1 Mbp DNA, which is
replicated early in S-phase. Transcription is performed at specialized
transcription centers (‘factories’) found at the interface between
the euchromatin and an interstitial network of channels that contain
little DNA but are rich in RNA. The active transcription centers
contain all the components required to establish the transcription
complex, generate the nascent transcript, and process this into mature
mRNA. Within each factory (five are shown), distinct but overlapping
zones containing the transcription machinery (50 nm, represented by
green ovals) and nascent RNA (50 nm, represented by red circles)
have been described. The anatomy of an individual factory is also shown
(C). In this example, the path of DNA in the associated
chromatin ‘clouds’ is depicted (black line shows the volume occupied
by the euchromatin from a typical gene domain of 50 kb) to highlight
the spatial constraints that might influence the efficiency of gene
expression. Chromatin domains and loops are generated by any contact
between chromatin and the active center. Many sequence motifs (see
text) are capable of generating such attachments and establishing a
domain configuration that is permissive for gene expression.
Furthermore, the stability of any attachments, and hence levels of
expression, will depend on extremely complex interactions that reflect
the affinity of individual proteins for DNA and the local
concentrations of these proteins. Attachments within the active center
will be changing continuously; although some configurations of the
chromatin cloud will inevitably predominate, turbulence generated by
chromatin movement during transcription ensures that these are locally
dynamic. The compartmentalization of different chromatin remodeling
machines might allow different zones of individual chromatin clouds to
establish subtly different chromatin states that also influence their
dynamic properties and ability to compete efficiently for nuclear sites
that are permissive for gene expression. Though most interactions
necessary to establish gene expression involve the transcription
factory, certain chromatin sequences may interact with proteins
restricted to nuclear compartments that are nonpermissive for gene
expression (black box). In this way, a highly complex combination of
genetic and epigenetic factors could contribute to the allele-specific
variations in expression that arise from genomic
imprinting.
|
|
If active genes assemble in active nuclear compartments where RNA
polymerases and splicing factors act (13)
, then the
nuclear position of a gene can have a significant effect on its
expression. Position effect variegation, telomeric and centromeric
position effects, and mating type silencing can result in normal gene
expression in some cells and silencing in others (reviewed in ref
1
). Transcription factors that bind to enhancers appear to
maintain genes in an active state by preventing the formation of
repressive chromatin (16)
. One mechanism that may account
for the effect of an enhancer is to localize genes in an active
euchromatin nuclear compartment away from centromeres (17
, 18)
. The transcriptional activity of a gene can also determine
whether it is replicated either early or late in S-phase. With few
exceptions, most transcriptionally active genes are replicated early
during S-phase and transcriptionally silent genes are replicated late.
This is also true for the active and silenced imprinted alleles
(19
, 20)
. The factors that may determine the accessibility
of replication origins include both the condensation state of
euchromatin and location to nuclear sites of transcription
(21)
. Consequently, CpG islands, which are methyl
cytosine-deficient regions around gene promoters, often map close to
origins of replication (22)
.
LCRs are another higher order control element that were thought to
contribute to the position-independent expression of transgenes by
opening chromatin into a more accessible state, but may also target
genes into the appropriate nuclear compartment. The presence of LCRs
was initially identified from human mutations of the ß-globin locus
and by their hypersensitivity to nuclease digestion. However,
subsequent deletion of the ß-globin LCR showed it was not necessary
for the maintenance of a nuclease-sensitive chromatin state within the
whole region (100 kb), but did regulate the overall level of
transcription (reviewed by Bulger and Groudine, ref 23
).
Furthermore, the developmental switch from fetal forms of globin to
adult forms did not depend on the LCR, even though the level of
expression of each gene did (24)
. Further results suggest
that the transcriptional competence of the ß-globin domain and the
degree of histone acetylation correlate with a location away from
heterochromatin (centromeric) regions in the nucleus (18)
.
Further acetylation of histone H3 at the globin promoter influences
whether the LCR is able to act, but only if the domain is
transcriptionally competent to start with. Thus, LCRs appear to act as
powerful enhancers rather than having an independent function.
|
TRANSCRIPTION CYCLES AND ‘FLIP/FLOP’
|
|---|
If we can accept the evidence for immobilized polymerases, then
the ‘transcription cycle’ for a gene would involve the attachment of
gene promoters within a chromatin loop to the immobile polymerase
complex in a factory, followed by transcription and processing of RNA
(13
, 25
; Fig. 1
). Once transcription is terminated, the
chromatin loop dissociates from the polymerase. Reinitiation of
transcription would require rebinding of the immobile RNA polymerase to
the gene promoter buried in the loop or ‘cloud’. The factors likely
to determine the rate of reinitiation are those that determine the
proximity and cooperative stability of multiple protein–DNA (e.g.,
TATA binding proteins) and protein–protein (e.g., TAFs) interactions
between the holoenzyme and the promoter (12)
. If the gene
is present in a large chromatin domain, then attachment of other
cis elements (e.g., enhancers) close to the gene may
cooperate to activate transcription, establishing a dynamic equilibrium
that positions the DNA protein initiation complex closer to the
transcription site. By having a greater attaching effect on the
chromatin loop containing the gene, either an LCR or an enhancer may
function by enhancing the probability of these key transcription cycle
interactions. The catalysis of promoter engagement to the polymerase
would be independent of whether the control element lies 5' or 3'
relative to the position of the promoter and may not depend on the
direct interaction with the promoter. The accessibility of binding
sites in DNA to transcription factors would be determined by the degree
of condensation and compaction of chromatin and, consequently, the form
and composition of proteins such as histones.
The shift between the two extreme states of attachment and detachment,
previously referred to as the flip/flop of the transcription cycle
(26
27
28
29)
, can be explained by cycles of
attachment/detachment of genes and local transcription factories.
Hence, bouts of efficient expression will occur when a gene promoter is
held in close proximity to the transcription machinery, whereas dynamic
changes in this spatial organization allow prolonged periods of
transcriptional silence. Indeed, it is important to realize that in
eukaryotic cells most genes are inactive for long periods
(30)
. The fluctuations present within one cell,
particularly of short-lived RNA, may be smoothed out when considering a
population of cells. A dynamic transcription cycle model may predict
heterogeneity within a population of cells: some cells would be either
transcribing a gene from one allele, both alleles, or neither allele.
Evidence derived from the examination of globin gene expression now
supports this prediction by heterogeneity of expression of genes in a
given cell population (29
, 31)
.
|
GENOMIC IMPRINTING
|
|---|
Autosomal genes, including immunoglobulin, T cell receptor,
olfactory receptor, and X-inactivated genes, may undergo monoallelic
expression with random parental allele expression. Imprinted genes are
unusual autosomal genes because they are expressed predominantly from
either the paternal or maternal inherited allele in somatic cells.
Approximately 30 genes with specific parent of origin allele expression
have now been identified, with estimates that imprinted genes account
for 0.1% of genes in the mammalian genome (50–100 genes). Genes
that are not expressed from the maternal allele are said to be
‘maternally’ imprinted. Genes that are not expressed from the
paternal allele are said to be ‘paternally’ imprinted. Thus,
imprinting refers operationally to the allele that is marked for
reduced expression; however, there may be ‘marking’ of the expressed
allele, and so the term ‘imprint’ does not necessarily refer to a
specific epigenetic modification. The choice of the expressed allele
does not vary with successive generations and is determined exclusively
by the parent of origin (32)
. The fact that imprinting
persists in inbred genetically identical mice suggests that the imprint
is solely an epigenetic phenomena. Pronuclear transplantation
experiments in mice have shown that androgenetic and gynogenetic
embryos are not viable, with growth abnormalities of the fetus and
placenta, respectively (33
, 34)
. Therefore, the apparent
allele-specific expression of imprinted genes accounts for the abnormal
development observed during parthenogenesis and confirms the
nonequivalence of the two parental genomes. Imprinted genes play roles
in embryonic and postnatal growth, but are also implicated in several
diseases, including cancer (35)
.
The imprinted allele is not transcriptionally silent in imprinted
tissues, as RNA can usually be detected from the silenced allele. For
example, mouse maternal Igf2 mRNA levels are
up to 10% of the paternal allele in embryos and tissues, except in the
leptomeninges of the brain where expression is normally biallelic
(32
, 36
, 37)
. There must be a cutoff that determines
whether a gene has sufficient parent-of-origin differences in the
expression from each allele to be termed ‘imprinted’ or not. Often
this is arbitrarily determined, with nonimprinted to imprinted allele
RNA ratios less than 3:1 taken as reflecting significant silencing of
the imprinted allele. Transcriptional repression has also been examined
by nuclear run-on assays; for example, H19, a noncoding RNA
situated 90 kb 3' to the Igf2 gene,
transcripts double after duplication of the maternal allele
(38)
.
Extensive analysis of imprinted genes and their genomic regions has
resulted in several common features. First, the majority of imprinted
genes tend to be clustered in chromosomal regions (35
, 39
, 40)
. Within each cluster may be genes that are either
reciprocally imprinted or expressed from both alleles. We shall
concentrate on the best-characterized cluster containing
Ins2, Igf2, H19,
p57kip2, and Kvlqt1 on the distal tip of mouse
chromosome 7 (Fig. 2
). Second, epigenetic modification of specific sequences during
gametogenesis marks the allele to be either silenced or expressed
(reviewed in ref 41
). Thus, the modification that defines
the imprint occurs when the genomes are separate. Finally, evidence now
suggests that imprinting control occurs at the level of transcription.
View larger version (9K):
[in this window]
[in a new window]
|
Figure 2. The imprinting domain on the distal tip of mouse chromosome 7.
Imprinting appears to cluster around chromosome domains. This is a map
of the distal tip of mouse chromosome 7 close to the telomere.
Paternally expressed genes are shown in boldface and distances are in
kilobases (kb).
|
|
|
EPIGENETIC MODIFICATION OF IMPRINTED DOMAINS
|
|---|
A candidate modification with a critical role in transmitting
imprinting signals is the methylation of cytosine bases by DNA
methyltransferases specifically at CpG dinucleotides.
Differentially methylated regions (DMRs) are thought to be
elements that are resistant to the wave of demethylation that occurs
during preimplantation development, and are therefore strong candidates
as gametic marks (42
43
44)
. After implantation, a wave of
de novo methylation establishes a new embryonic methylation pattern.
The majority of the maintenance methylation activity is catalyzed by
DNA methyltransferase 1 (Dnmt 1), but several other enzymes are also
identified in the mouse (Dnmt2, 3a, and 3b). Imprinting can be reversed
either by modifying genetic elements within the domains or by
alteration of epigenetic factors such as methylation (40
, 41
, 45
, 46)
.
There are differences in methylation patterns between alleles of
imprinted genes, which has been taken to indicate that methylation
marks an allele for silencing. For H19, a CpG region 6 kb
upstream and 0.6 kb downstream of the gene appears to be heavily
methylated on the paternal allele. The distribution of methylation
inversely correlates with chromatin accessibility to enzymes (DNAseI)
at the promoter of the maternal expressed allele relative to the
methylated and silenced paternal allele (38
, 46)
. For
Igf2, the promoter regions on both alleles
are equally sensitive to restriction enzymes, indicating they are in a
relatively open transcriptionally active chromatin conformation
(36)
. A hypermethylated region is present upstream of the
first Igf2 promoter on the paternal allele,
known as DMR1, and a second in the 3' part of the gene (DMR2)
(37
, 47)
. However, this methylation pattern is absent in
the gametes. Even though they appear equally accessible to nuclease
digestion, the more transcriptionally active paternal allele of
Igf2 tends to be more heavily methylated
than the maternal allele on mouse chromosome 7
(48)
.
In contrast, the Igf2r (receptor) gene on
mouse chromosome 17 appears to be more complicated and suggests that
different imprinting controls may exist. A region (region 1) involving
the promoter is hypermethylated on the paternal unexpressed
allele, but this is established late in the development of embryos.
Another region on the paternal allele (region 2) is present within
intron 2 of the gene. Here CpGs are methylated in oocytes and remain
methylated throughout development (49)
. This second region
would constitute the requirements of an ‘imprinting center’, defined
as the gametic mark for the parent of origin and facilitating
expression from the maternal allele. The methylation of
Igf2r region 2 appears to depend on Dnmt 1
whereas the DMR2 region in Igf2 may depend
on Dnmt 3a and 3b.
All of the DMRs that have been identified are G + C-rich and therefore
are similar to the CpG islands that occur around gene promoters. They
probably bind several types of proteins that may produce a noncanonical
distribution of nucleosomes across the region and subsequent
hypersensitivity to nuclease digestion. Although altered chromatin
conformation may be present at DMRs, they differ from true CpG islands
because they appear to maintain allele-specific methylation during
waves of de novo demethylation. Thus, once established during
gametogenesis, the DMR methylation pattern appears stable. However,
with age and in disease states, relaxation of these patterns may occur.
The experimental evidence defining a direct role of methylation in
imprinting comes from mice with hypomorphic and null mutations of the
DNA methyltransferase enzyme (MTase=Dnmt1) (45)
. Although
hypomorphic mutations result in embryonic lethality before E12.5
(Mtasen), inactivating mutations result in
an earlier more severe phenotype
(Mtases). The former mutation
results in biallelic H19 expression associated with reduced
Igf2. As described later, changing the
methylation status of the paternal allele H19 region can
alter the interaction with a boundary element, which disrupts the
cis interaction of the H19 enhancers with
Igf2, resulting in biallelic H19
expression. Compared with Igf2,
Igf2r expression was only reduced on a
Mtases background
(45)
. Thus, not every locus is equally sensitive to the
effect of DNA methyltransferase disruption.
A further connection between methylation of DNA and transcriptional
repression appears to be mediated via histone acetylation. The binding
of MeCP2 to methyl CpGs forms a complex with histone deacetylase
(50
, 51)
and targets transcriptional repression. Thus, the
function of MeCp2 and other methyl-CpG binding proteins includes
actions as corepressors (52)
.
Methylation may also influence long-range transcriptional control and
nuclear context. Investigation of LAR region associated with the µ
enhancer of the heavy chain immunoglobulin locus has shown that
methylation impairs distance independence of enhancer function
(53)
. Additional attachment regions appear to overcome the
inhibitory effect of methylation and facilitate the long-range action
of the enhancer (53)
. The effect may be mediated by
long-range modification of histone acetylation, with methylation
resulting in recruitment of histone deacetylase by corepressor methyl
CpG binding proteins (51
52
53)
. This is an unusual
situation, as the methylated allele is not imprinted. Methylation of
DMRs close to the Igf2 gene may be one way
to limit the dynamic range of expression of the
Igf2 gene and to increase dependence on
enhancer activity from either the H19 endodermal enhancers
or enhancers from other regions acting in different cell lineages
(36
, 37
, 52
, 54)
. If the location of methylation generates
more of an enhancer-driven dependence on the level of gene expression,
it suggests that the reduced Igf2 expression
from the maternal allele will be greatly affected by enhancer blocking.
The fact that the maternal allele does not methylate the
Igf2 DMRs when it is expressed in liver
(endoderm) also points to the methylation of DMRs as having a modifying
effect only (55)
. This does not exclude the possibility
that mesoderm-specific enhancers may interact differently with DMRs of
Igf2 on each allele. Whether the gene is
expressed or not depends ultimately on which promoter elements are
active, and these may act either with or without tissue-specific
enhancer effects to regulate the dynamic expression of a gene. The
epigenetic modifications we have described will influence the degree of
chromatin accessibility and condensation, effectively controlling the
density of chromatin clouds and influencing the probability of
molecular interactions with the transcription sites.
|
PREVIOUS MODELS OF IMPRINTING
|
|---|
The enhancer competition model was first described in 1993 by
Bartolomei and colleagues (46)
. The model concerns the
reciprocal expression of Igf2 and
H19 on each allele. Both genes occur in the same chromosomal
region and are closely linked (Fig. 2)
. They share an enhancer
downstream of H19, often termed the H19
endodermal enhancer (56)
. The enhancer is not methylated
and is situated in an open chromatin conformation. The model proposes
that the enhancer activates transcription of the nearest gene located
in cis. For the maternal allele, this would be
H19; for the paternal allele it would be
Igf2. However, even though there may be
distance constraints between the enhancers and genes, the hypothesis
falls down in the case of maternal allele expression. Significant
expression of the maternal Igf2 gene may
still be predicted unless there is a boundary between H19
and Igf2. One would also predict that
expression of H19 would be an absolute requirement for the
silencing of Igf2, and vice versa. Three
sources of evidence reveal that these predictions do not hold for all
mammals and tissues. First, several species maintain
Igf2 imprinting yet have biallelic
expression of H19 (57)
. Second, biallelic
expression of Igf2, with H19
expression confined mainly from the maternal allele, occurs in the
leptomeninges (58
, 59)
. Third, specific expression of
H19 RNA is not required for imprinting
(60
61
62)
. Thus, no simple reciprocal relationship occurs
between Igf2 and H19 expression
that can account for all circumstances. Furthermore, it is still
unknown what determines expression of Igf2
in the case of the ‘natural’ relaxation of imprinting in the
leptomeninges of the brain. Overall, such results do suggest some
degree of plasticity in this system, with expression of both
H19 and Igf2 able to switch on
and off independently (38
, 58
, 59)
.
Efstratiadis (63)
suggested a variant of the enhancer
competition model. In this model, an imprinting box (an allele- and
gamete-specific epigenetic mark) is situated between H19 and
Igf2 (Fig. 2)
. Attachment of the box to the
nuclear scaffold (matrix or nucleoskeleton) via an adapter or reader
confines the action of the H19 enhancer to the maternal
allele. On the paternal allele, the box is methylated and does not bind
the scaffold, allowing free access of the enhancer to regulate
Igf2 expression. Methylation of the promoter
and upstream region of H19 might preclude such an
interaction (63)
. This prediction may now turn out to be
correct to some extent if the imprinting box, now termed the
‘imprinting control center’ (ICR), also acts a chromatin boundary
element (see Fig. 3
).
View larger version (38K):
[in this window]
[in a new window]
|
Figure 3. A transcription cycle model of genomic imprinting. The
Igf2/H19 locus is visualized in relation
to immobilized RNA polymerases located on the surface of transcription
factories. On the paternal allele, methylation of the H19 upstream
region and promoter (DMR; see text) renders the DNA inaccessible to
interaction with either polymerase or boundary element binding proteins
(e.g., CTCF). The interaction of the chromatin loop (black line) with
the enhancer can have direct cis acting affects on the
interaction and on/off kinetics of the
Igf2 gene (purple lines). The
consequence is the increased expression of
Igf2 transcripts (arrows) and relative
lack of expression of H19. On the maternal allele, there
is little methylation of the H19 upstream region and the chromatin
configuration becomes accessible. Binding to a structurally remote site
tethers the chromatin loop and is only one mechanism that interrupts
direct cis acting effects of the enhancers on
Igf2. The enhancer action is then
confined to H19 expression (green arrows), with
Igf2 expression falling to a basal level
that is highly promoter dependent. Duplication of the enhancer region
on the maternal allele restores direct cis acting
interactions to the same structural collection of RNA polymerases, so
that there is high-level expression of both
Igf2 and H19. Deletion of
H19 and the upstream region results in no interactions
with either the boundary element or the polymerase, but leaves the
enhancer to influence Igf2 initiation
rates. Moving the enhancer between the boundary element and
Igf2, results in increased expression of
Igf2 but down-regulation of
H19, as the enhancer is prevented from activating
H19 expression because of the active boundary element
tether, resulting in expression levels dependent on the
H19 promoter.
|
|
A revision of the enhancer competition model followed, with
incorporation of a boundary element and shared enhancers
(61)
. Subsequent work has located the position of the
boundary element in relation to the H19 gene. Replacement of
13 kb of the H19 locus and its promoter with a neomycin
resistance gene resulted in mice with relaxation of imprinting of
Igf2 and an allele ratio of 
2:1
(64)
. Deletion of 3 kb of the H19 promoter with
a similar resistance gene produced only modest relaxation, with allele
ratios no greater than 4:1 (65)
. Finally, specific
deletion of the mouse upstream boundary element (1.6 kb) alone results
in complete relaxation of imprinting of both
Igf2 and H19, with both genes
coexpressed at lower levels (66
; Fig. 2
). Confirmation of
the boundary function of this region has also been obtained from
Drosophila using a mini-white gene reporter separated by the
H19 upstream DMR and the H19/Igf2
downstream enhancers (67)
. The removal of the boundary
element (1.2 kb) promoted mini-white expression driven by the enhancer.
On the maternal allele, the boundary model holds that enhancer
interaction with Igf2 is blocked, giving
preferential enhancer interaction with the H19 gene. The
characteristic sensitivity of boundary elements to nucleases has also
been confirmed for the maternal allele, but not the paternal allele
(68)
. Only when the boundary is deleted is there complete
relaxation of imprinting (64
, 66)
, unlike the situation
when it is left intact and imprinting persists (61
, 65)
.
One prediction from the evidence for a boundary element upstream of
H19 is that moving the enhancers to the same side of the
boundary as Igf2 is likely to lead to
increased Igf2 expression (see Fig. 3
).
Removal of the H19 endodermal enhancer to a region
equidistant from H19 and Igf2 has
been shown to result in equal expression of
Igf2 from the paternal and maternal alleles
(55)
. Duplication of the enhancer also results in
increased maternal allele expression, but only to 30% of the paternal
allele, as assessed with an RNase protection assay. A similar effect on
the paternal H19 gene was not seen, so presumably the
expression of H19 is strongly influenced by paternal allele
methylation (55)
.
If boundary elements act to form a LAR, they may do so via a stable
protein DNA interaction or via polymerases. A proposal by Geyer
(69)
suggests that boundary elements may be acting as
pseudopromoters by engaging RNA polymerases to produce short-lived
transcripts. There are several precedents for this idea. First, RNA
transcripts have been detected from LCR elements upstream of the
ß-globin locus (70)
. Second, one of the
best-characterized boundary elements/insulators in
Drosophila, the scs' element, can also act as a gene
promoter (71
, 72)
. Third, RT-PCR may detect RNA from the
region located upstream of the H19 gene
(61)
.
Recent experiments have identified CTCF, a multiple zinc finger protein
that binds ICR elements in a methylation-dependent manner. CTCF binds
to a 42 bp consensus sequence hypersensitive site 4 (HS-4) that
normally acts as a boundary element of the ß-globin locus
(73)
. Examination of the boundary element upstream of
H19 has revealed four footprints on the maternal allele
consistent with CTCF binding sites (74
75
76)
. Footprints
are absent from the methylated paternal allele, which also shows
disruption of transcription factor footprint patterns across the
H19 promoter (74
75
76)
. It remains unknown
whether CTCF binding to the ICR occurs in the gametes and protects the
region from subsequent methylation. However, the data from tissue
culture enhancer-blocking assays and transgenesis now support the idea
of a regulated boundary element (74
75
76)
. Further
characterization of CTCF and boundary function is likely to be
complemented by evidence from nonmammalian experimental systems
(71
, 77
, 78)
.
One further imprinting model concerns ‘repressor competition’. This
arose from observations in mice when a single copy
Igf2 transgene was expressed in addition to
the endogenous gene in chimeras (79)
. The overgrowth
phenotype that resulted from chimeras with excess
Igf2 expression had similarities to
Beckwith-Wiedeman syndrome, a human overgrowth syndrome with
predisposition to cancer. Surprisingly, Igf2
RNA expression from the transgene appeared to be progressively
inhibited in chimeras with increasing contribution from the transgene
line. The overgrowth phenotype appeared to derive from increasing
expression from the endogenous gene, mainly from the maternal allele.
If a repressor normally controls the expression of the imprinted
allele, then its action may be diluted by the presence of competition
from a similar gene—in this case, leading to increased expression from
the normally repressed maternal Igf2 allele
(79)
. These results further support the reversible nature
of imprinted gene expression to account for the titratable growth
effect observed. An allele discriminating protein would be a candidate
repressor, as has been suggested for the proteins that might bind the
Igf2r intronic DMR or imprinting box
(80)
. Identification of the DNA sequences that target
methylation (de novo methylation) and allele discrimination within a
DMR control element indicates that the factors that bind specifically
may soon be identified (43)
.
|
EVIDENCE FOR TRANSCRIPTION CYCLES AND IMPRINTING
|
|---|
As for the globin genes, RNA fluorescent in situ hybridization of
mouse liver and hemopoietic cells from embryonic liver (E13.5) has
revealed multiple patterns of expression of imprinted genes, visualized
as nuclear dots (29
, 81)
. This type of experiment sheds
further light on the dynamic nature of imprinting control. Jouvenot at
al. made spreads of cell suspensions from livers of E13.5 mice
(81)
. After permeabilization, fixed cells were either
hybridized with H19 oligonucleotide probes and labeled with
digoxygenin or Igf2 probes were labeled with
biotin. Probe detection with fluorescence revealed at least nine
patterns of transcription sites in nuclei. The expected pattern of one
Igf2 signal and one H19 signal in
the same nucleus occurred in only 25% of nuclei examined. If one
H19 allele was expressing RNA whereas the
Igf2 allele was silent momentarily, or vice
versa, then the frequency that this occurred was 27% vs. 17%
respectively. Biallelic expression of H19 occurred at a frequency of
13% whereas biallelic expression of Igf2
occurred at a frequency of only 8%. The most common situation when
there were three or more signals was biallelic expression of
H19 in conjunction with monoallelic expression of
Igf2. Naturally, the interpretation of these
results would depend on the cell type and stage of the cell cycle, as
late S-phase and G2 cells may have replicated a transcriptionally
active allele, duplicating the active parental
Igf2 allele and doubling the number of dots.
However, only liver tissue was examined, and cell cycle parameters were
not taken into account in this case. Even so, the implication from this
data is that the accumulation of transcripts from a particular allele
of an imprinted gene depends on the dynamic behavior of its
transcription cycle. Hence, for the expressed paternal allele of
Igf2, the gene will be transcribed for a
significantly greater proportion of a given period than the imprinted
maternal allele. As a consequence, complete silencing of the imprinted
allele in a particular cell is not seen. This analysis also
demonstrated that the paternal H19 RNA appears to be less
stable than the maternally derived transcript. Even for the noncoding
RNA, this implies that nuclear context can influence different aspects
of RNA metabolism that may also be important in imprinting
(81)
.
A potentially related observation concerns the homologous association
of oppositely imprinted chromosomal domains in the nuclei of cells.
Close association of nuclear foci from chromosome 15q11–13 in situ
hybridization probes was observed for both alleles of the imprinted
domain. This effect occurred in late S-phase in human lymphocytes and
was lost in cells derived from patients with Prader-Willi and Angelman
syndromes, human syndromes with imprinted phenotypes. This implies that
homologous association may be rather transient. The association of
homologous chromosomes can also occur in nonmammalian cells and can
result in the process of transvection. This has been studied most in
Drosophila, where enhancers of a gene can function in
trans to affect the expression of the homologue (82
, 83)
.
A transcription cycle model of imprinting would help to explain the
dynamic and reversible nature of imprinted gene expression. The model
proposes that tissue-specific enhancers, promoters, and insulators act
as LARs that combine to activate the desired levels of gene expression.
However, as the numerous interactions needed to establish gene
expression are dynamic, the local organization of chromatin domains
will be changing continuously. During this process, structures will
arise that position genes in such a way that their promoters are able
to interact productively with nuclear sites that perform RNA synthesis
(Fig. 3)
. Such changes represent a transcription cycle, where levels of
gene expression will depend on a combination of promoter ‘strength’
and the proportion of time the promoter occupies the active
configuration. For imprinted genes, differences (perhaps subtle) in
chromatin structure of the respective alleles must perturb the balance
of this transcription cycle, allowing one to be expressed more
efficiently than the other.
Such a dynamic model may also explain imprinting of other genes. We
suggest that the Igf2r DMR in intron 2 of
the gene may form a LAR via transcription units. One explanation is
that DMR methylation on the expressed maternal allele favors sense
transcripts of the gene, whereas antisense transcripts from intron 2
reduce the overall probability of gene transcription in the sense
direction (84
, 85)
. However, in marsupials (opossum),
Igf2r imprinting is detected, but the DMR in
intron 2 lacks the CpG island and instead has multiple repeat
sequences, suggesting that the antisense transcript detected from this
region may not be the full story to imprinting of the receptor
(86)
. In Prader-Willi syndrome, deletions of the SNRPN
promoter occur on the paternal allele, resulting in reduced expression
of SNRPN and genes in a region extending at least 150 kb downstream and
1 Mb upstream (reviewed in ref 40
; see also ref
87
). This suggests that the promoter region is a critical
LAR that sets the transcriptional competence of the whole domain.
Angelman syndrome (AS) involves deletions slightly further upstream, a
region that produces noncoding RNA (BD exons) which are spliced to
downstream exons on the maternal allele. Expression of the noncoding
RNA appears to be essential to the expression of UBE3A, a
ubiquitin protein ligase situated downstream and thought to account for
the phenotype of AS. Again, the noncoding region could be acting as a
LAR to influence expression of UBE3A. Deletion of the
promoter region of SNRPN results is reversion to the maternal allele
pattern, and deletion of the maternal AS region results in
down-regulation of the UBE3A gene (88)
.
Finally, the similarities of genomic organization between the
Dlk2 and Gtl2 genes and the
H19/Igf2 locus suggest that the latter locus
is not unique; interpretation of available data on this locus will have
implications for imprinting control in general (89
, 90)
.
Boundary elements or silencing elements are believed to act as
uncoupling agents if placed between enhancers and gene promoters. One
possibility is that boundary elements operate to antagonize enhancer
function by disrupting the chromatin configuration that is essential to
establishing the productive phase of the transcription cycle.
Stimulating interactions that tend to drive critical sequences toward
inert nuclear compartments—and hence remote from the active
sites—would be one way of achieving this (Figs. 1
and 3)
. In the case
of imprinting and the Igf2-H19 locus, the
methylation-dependent H19 boundary element interaction
relies on the accessibility of protein binding to the boundary element.
In the case of the paternal allele, methylation and chromatin
compaction reduces the probability of interaction to both RNA
polymerase and the boundary attachment site, but does not exclude it.
Thus, reversible effects can occur and may account for lack of
Igf2 expression in DNA methyltransferase
mutants. On the maternal allele, the accessibility of the chromatin
allows the boundary element to bind and interrupt the interactions of
enhancers with Igf2. This effect is also
determined by the relative probability of either RNA polymerase
enhancer-driven interaction or boundary element interaction. Thus, it
is still possible to generate H19 RNA expression, although
the effect is weighted toward boundary element binding. It is possible
that the boundary element will contain two competing functional parts:
one that binds the polymerase and one that binds the boundary. However,
we await further evidence that investigates this aspect of the model.
|
TRANSCRIPTION CYCLES DURING ESTABLISHMENT AND MAINTENANCE OF
IMPRINTING
|
|---|
Even though a transcription cycle model may explain the mechanism
of differential allele expression, we have not addressed the question
of how the imprinted genes are inherited. Clearly, differences in the
marking and modification of alleles must occur during gametogenesis in
the male and female germline (44
, 91
, 92)
. Once a mark is
made, we would argue that this establishes a defined interaction of
chromatin with a base of loop attachment via either a polymerase or a
boundary/silencer protein complex. The chromatin arrangement, the
distribution of histones, and the accessibility to methylation may all
be modified by such an interaction. What is key is that this
interaction should be stable during mitosis and through the waves of
demethylation and methylation during development. Once established, the
somatic maintenance of the interaction may not be stable, and
modification of allele RNA expression may occur under certain
circumstances, such as during alterations in methylation and
differentiation. Recent evidence using loxP/Cre deletion of the
H19 imprinting control region confirms that this element
controls both H19 expression and
Igf2 expression (93)
. Deletion
of the H19 ICR in the early embryo confirms that the element
is required during the establishment of the imprint, as well as later
during development and in differentiated muscle. Paternal allele
methylation of the ICR establishes two functions: the first is to
determine the expression of Igf2 from the
paternal allele and second to inhibit H19 promoter activity
and down-regulate H19 expression in differentiated tissues.
|
CONCLUSIONS
|
|---|
Extensive evidence suggests that imprinting is controlled at the
level of transcription. We propose a transcription cycle model of
control that emphasizes the probability of the interaction of control
elements within DNA and chromatin with the transcription and RNA
processing machinery within the nucleus.
|
ACKNOWLEDGMENTS
|
|---|
We thank Chris Graham for advice and discussion. J.L.B. is a BBSRC
Student. A.B.H. is a CRC Senior Research Fellow. We acknowledge the
support of The Cancer Research Campaign, UK and the Royal Society.
|
REFERENCES
|
|---|
-
Henikoff, S. (1992) Position effect and related phenomena. Curr. Opin. Genet. Dev. 2,907-912[Medline]
-
Karpen, G. H. (1994) Position-effect variegation and the new biology of heterochromatin. Curr. Opin. Genet. Dev. 4,281-291x[Medline]
-
Tyler, J. K., Kadonaga, J. T. (1999) The ‘dark side’ of chromatin remodeling: repressive effects on transcription. Cell 99,443-446[Medline]
-
Knoepfler, P. S., Eisenman, R. N. (1999) Sin meets NuRD and other tails of repression. Cell 99,447-450[Medline]
-
Maldonado, E., Hampsey, M., Reinberg, D. (1999) Repression: targeting the heart of the matter. Cell 99,455-458[Medline]
-
Cook, P. R. (1995) A chromomeric model for nuclear and chromosome structure. J. Cell. Sci. 108,2927-2935[Abstract]
-
Jackson, D. A., Dickinson, P., Cook, P. R. (1990) The size of chromatin loops in HeLa cells. EMBO J 9,567-571[Medline]
-
Jackson, D. A., Bartlett, J., Cook, P. R. (1996) Sequences attaching loops of nuclear and mitochondrial DNA to underlying structures in human cells: the role of transcription units. Nucleic Acids Res 24,1212-1219[Abstract/Free Full Text]
-
Stief, A., Winter, D., Stratling, W. H., Sippel, A. E. (1989) A nuclear DNA attachment element mediates elevated and position independent gene activity. Nature (London) 341,343-345[Medline]
-
Jackson, D. A., Cook, P. R. (1993) Transcriptionally active minichromosomes are attached transiently in nuclei through transcription units. J. Cell. Sci. 105,1143-1150[Abstract]
-
Jackson, D. A., Iborra, F. J., Manders, E. M., Cook, P. R. (1998) Numbers and organization of RNA polymerases, nascent transcripts, and transcription units in HeLa nuclei. Mol. Biol. Cell 9,1523-1536[Abstract/Free Full Text]
-
Kimura, H., Tao, Y., Roeder, R. G., Cook, P. R. (1999) Quantitation of RNA polymerase II and its transcription factors in an HeLa cell: little soluble holoenzyme but significant amounts of polymerases attached to the nuclear substructure. Mol. Cell. Biol. 19,5383-5392[Abstract/Free Full Text]
-
Cook, P. R. (1999) The organization of replication and transcription. Science 284,1790-1795[Abstract/Free Full Text]
-
Jackson, D. A., Hassan, A. B., Errington, R. J., Cook, P. R. (1993) Visualization of focal sites of transcription within human nuclei. EMBO J 12,1059-1065[Medline]
-
Pombo, A., Jackson, D. A., Hollinshead, M., Wang, Z., Roeder, R. G., Cook, P. R. (1999) Regional specialization in human nuclei: visualization of discrete sites of transcription by RNA polymerase III. EMBO J 18,2241-2253[Medline]
-
Walters, M. C., Magis, W., Fiering, S., Eidemiller, J., Scalzo, D., Groudine, M., Martin, D. I. (1996) Transcriptional enhancers act in cis to suppress position effect variegation. Genes Dev 10,185-195[Abstract/Free Full Text]
-
Francastel, C., Walters, M. C., Groudine, M., Martin, D. I. K. (1999) A functional enhancer suppresses silencing of a transgene and prevents its localizaton close to centromeric heterochromatin. Cell 99,259-269[Medline]
-
Schubeler, D., Francastel, C., Cimbora, D., Reik, A., Martin, D. I. K., Groudine, M. (2000) Nuclear localization and histone acetylation: a pathway for chromatin opening and transcriptional activation of the human ß-globin locus. Genes Dev 14,940-950[Abstract/Free Full Text]
-
Kitsberg, D., Selig, S., Brandeis, M., Simon, I., Keshet, I., Driscoll, D. J., Nicholls, R. D., Cedar, H. (1993) Allele-specific replication timing of imprinted gene regions. Nature (London) 364,459-463[Medline]
-
Knoll, J. H. M., Cheng, S. D., Lalande, M. (1994) Allele specificity of DNA replication timing in the Angelman/Prader-Willi syndrome imprinted chromosomal region. Nat. Genet. 6,41-46[Medline]
-
Hassan, A. B., Cook, P. R. (1994) Does transcription by RNA polymerase play a direct role in the initiation of replication?. J. Cell. Sci. 107,1381-1387[Abstract]
-
Delgado, S., Gomez, M., Bird, A., Antequera, F. (1998) Initiation of DNA replication at CpG islands in mammalian chromosomes. EMBO J 17,2426-2435[Medline]
-
Bulger, M., Groudine, M. (1999) Looping versus linking: toward a model for long distance gene activation. Genes Dev 13,2465-2477[Free Full Text]
-
Epner, E., Reik, A., Cimbora, D., Telling, A., Bender, S., Fiering, S., Enver, T., Martin, D. I. K., Kennedy, M., Keller, G., Groudine, M. (1998) The b-globin LCR is not necessary for an open chromatin structure or developmentally regulated transcription of the native mouse ß-globin locus. Mol. Cell 2,447-455[Medline]
-
Cook, P. R. (1994) RNA polymerase: structural determinant of the chromatin loop and the chromosome. Bioessays 16,425-430[Medline]
-
Trimborn, T., Gribnau, J., Grosveld, F., Fraser, P. (1999) Mechanisms of developmental control of transcription in the murine alpha- and beta-globin loci. Genes Dev 13,112-124[Abstract/Free Full Text]
-
Gribnau, J., de Boer, E., Trimborn, T., Wijgerde, M., Milot, E., Grosveld, F., Fraser, P. (1998) Chromatin interaction mechanism of transcriptional control in vivo. EMBO J 17,6020-6027[Medline]
-
Fraser, P., Grosveld, F. (1998) Locus control regions, chromatin activation and transcription. Curr. Opin. Cell Biol. 10,361-365[Medline]
-
Wijgerde, M., Grosveld, F., Fraser, P. (1995) Transcription complex stability and chromatin dynamics in vivo. Nature (London) 377,209-213[Medline]
-
Jackson, D. A., Pombo, A., Iborra, F. (2000) The balance sheet for transcription: an analysis of nuclear RNA metabolism in mammalian cells. FASEB J 14,242-254[Abstract/Free Full Text]
-
Robertson, G., Garrick, D., Wu, W., Kearns, M., Martin, D., Whitelaw, E. (1995) Position-dependent variegation of globin transgene expression in mice. Proc. Natl. Acad. Sci. USA 92,5371-5375[Abstract/Free Full Text]
-
DeChiara, T. M., Robertson, E. J., Efstratiadis, A. (1991) Parental imprinting of the mouse insulin-like growth factor II gene. Cell 64,849-859[Medline]
-
McGrath, J., Solter, D. (1984) Completion of mouse embryogenesis requires both the maternal and paternal genomes. Cell 37,179-183[Medline]
-
Barton, S. C., Surani, M. A., Norris, M. L. (1984) Role of paternal and maternal genomes in mouse development. Nature (London) 311,374-376[Medline]
-
Graham, C. F., Lund, G., Zaina, S. (1998) Growth and the distal tip of mouse chromosome 7. Genet. Res. 72,247-253[Medline]
-
Sasaki, H., Jones, P. A., Chaillet, J. R., Ferguson Smith, A. C., Barton, S. C., Reik, W., Surani, M. A. (1992) Parental imprinting: potentially active chromatin of the repressed maternal allele of the mouse insulin-like growth factor II (Igf2) gene. Genes Dev 6,1843-1856[Abstract/Free Full Text]
-
Feil, R., Walter, J., Allen, N. D., Reik, W. (1994) Developmental control of allelic methylation in the imprinted mouse Igf2 and H19 genes. Development 120,2933-2943[Abstract]
-
Ferguson-Smith, A. C., Sasaki, H., Cattanach, B. M., Surani, M. A. (1993) Parental-origin-specific epigenetic modification of the mouse H19 gene. Nature (London) 362,751-755[Medline]
-
Cattanach, B. M., Kirk, M. (1985) Differential activity of maternally and paternally derived chromosome regions in mice. Nature (London) 315,496-498[Medline]
-
Bartolomei, M. S., Tilghman, S. M. (1997) Genomic imprinting in mammals. Annu. Rev. Genet. 31,493-525[Medline]
-
Tilghman, S. M. (1999) The sins of the fathers and mothers: genomic imprinting in mammalian development. Cell 96,185-193[Medline]
-
Jaenisch, R. (1997) DNA methylation and imprinting: why bother?. Trends Genet 13,323-329[Medline]
-
Feil, R., Khosla, S. (1999) Genomic imprinting in mammals, an interplay between chromatin and DNA methylation. Trends Genet 15,431-435[Medline]
-
Brandeis, M., Frank, D., Keshet, I., Siegfried, Z., Mendelsohn, M., Nemes, A., Temper, V., Razin, A., Cedar, H. (1994) Sp1 elements protect a CpG island from de novo methylation. Nature (London) 371,435-438[Medline]
-
Li, E., Beard, C., Jaenisch, R. (1993) Role of DNA methylation in genomic imprinting. Nature (London) 366,362-365[Medline]
-
Bartolomei, M. S., Webber, A. L., Brunkow, M. E., Tilghman, S. M. (1993) Epigenetic mechanisms underlying the imprinting of the mouse H19 gene. Genes Dev 7,1663-1673[Abstract/Free Full Text]
-
Brandeis, M., Kafri, T., Ariel, M., Chaillet, J. R., McCarrey, J., Razin, A., Cedar, H. (1993) The ontogeny of allele-specific methylation associated with imprinted genes in the mouse. EMBO J 12,3669-3677[Medline]
-
Forne, T., Oswald, J., Dean, W., Saam, J. R., Bailleul, B., Dandolo, L., Tilghman, S. M., Walter, J., Reik, W. (1997) Loss of the maternal H19 gene induces changes in Igf2 methylation in both cis and trans [published erratum appears in Proc. Natl. Acad. Sci. USA, 1997, vol. 94, p. 14211]. Proc. Natl. Acad. Sci. USA 94,10243-10248[Abstract/Free Full Text]
-
Stoger, R., Kubicka, P., Liu, C. G., Kafri, T., Razin, A., Cedar, H., Barlow, D. P. (1993) Maternal-specific methylation of the imprinted mouse Igf2r locus identifies the expressed locus as carrying the imprinting signal. Cell 73,61-71[Medline]
-
Bird, A. (1999) DNA methylation de novo. Science 286,2287-2288[Free Full Text]
-
Nan, X., Ng, H. H., Johnson, C. A., Laherty, C. D., Turner, B. M., Eisenman, R. N., Bird, A. (1998) Transcriptional repression by the methyl-CpG-binding protein MeCP2 involves a histone deacetylase complex [see comments]. Nature (London) 393,386-389[Medline]
-
Bird, A. P., Wolffe, A. P. (1999) Methylation-induced repression—belt, braces, and chromatin. Cell 99,451-454[Medline]
-
Forrester, W. C., Fernandez, L. A., Grosschedl, R. (1999) Nuclear matrix attachment regions antagonize methylation-dependent repression of long range enhancer-promoter interactions. Genes Dev 13,3003-3014[Abstract/Free Full Text]
-
Constancia, M., Pickard, B., Kelsey, G., Reik, W. (1998) Imprinting mechanisms. Genome Res 8,881-900[Abstract/Free Full Text]
-
Webber, A. L., Ingram, R. S., Levorse, J. M., Tilghman, S. M. (1998) Location of enhancers is ess
TOP
ABSTRACT
INTRODUCTION
