It was discovered that a histone acetylation is enriched in active genes [4], b active genes adopt a more accessible chromatin conformation [5–7] and c gene regulatory elements are assoc
Trang 1Structure and function of active chromatin
and DNase I hypersensitive sites
Peter N Cockerill
Experimental Haematology, Leeds Institute of Molecular Medicine, University of Leeds, UK
Introduction
Our current understanding of chromatin structure
really began in the 1970s when it was demonstrated
that chromatin was built up from nucleosomes [1,2]
and it was found that histones could be acetylated [3]
In the late 1970s and early 1980s it was then
recog-nized that chromatin structure was likely to play a
sig-nificant role in gene regulation It was discovered that
(a) histone acetylation is enriched in active genes [4],
(b) active genes adopt a more accessible chromatin
conformation [5–7] and (c) gene regulatory elements
are associated with nucleosome-free regions that came
to be known as DNase I hypersensitive sites (DHSs)
[7–10] This remained a relatively obscure field of
research until the mid-1990s when the current intenseinterest in chromatin modifications was prompted bythe discovery that transcription factors recruit histonemodifying enzymes [11] and chromatin remodellingcomplexes [12,13] Since then there has been an explo-sion of papers on the multitude of chromatin modifica-tions and the factors that can either create orrecognize them We now have a very detailed picture
of the chromatin modifications normally associatedwith transcription units Hence, we know that promot-ers, gene bodies, termination regions and even intro-
n⁄ exon boundaries have very characteristic signatures
of histone modifications, histone replacements and
Keywords
chromatin; DNase I hypersensitive; gene
regulation; nucleosome; transcription
Correspondence
P N Cockerill, Experimental Haematology,
Leeds Institute of Molecular Medicine,
University of Leeds, Wellcome Trust
Brenner Building, St James’s University
Hospital, Leeds LS9 7TF, UK
Fax: +44 113 343 8502
Tel: +44 113 343 8639
E-mail: p.n.cockerill@leeds.ac.uk
(Received 18 December 2010, revised 10
February 2011, accepted 5 April 2011)
doi:10.1111/j.1742-4658.2011.08128.x
Chromatin is by its very nature a repressive environment which restricts therecruitment of transcription factors and acts as a barrier to polymerases.Therefore the complex process of gene activation must operate at two levels
In the first instance, localized chromatin decondensation and nucleosomedisplacement is required to make DNA accessible Second, sequence-specifictranscription factors need to recruit chromatin modifiers and remodellers tocreate a chromatin environment that permits the passage of polymerases Inthis review I will discuss the chromatin structural changes that occur atactive gene loci and at regulatory elements that exist as DNase I hypersensi-tive sites
Abbreviations
BE, boundary element; ChIP, chromatin immunoprecipitation; CTD, C-terminal domain; DHS, DNase I hypersensitive site; DNMT, DNA methyltransferase; EM, electron microscopy; GM-CSF, granulocyte macrophage colony-stimulating factor; HAT, histone acetyltransferase; HDAC, histone deacetylase; Hsp70, heat shock protein 70; IL-4, interleukin-4; LCR, locus control region; MAR, matrix attachment region; MBD, methyl binding domain; MMTV, mouse mammary tumour virus; MNase, micrococcal nuclease; ncRNA, non-coding RNA; NF1, nuclear factor 1; NFAT, nuclear factor of activated T cells; PARP, poly(ADP-ribose) polymerase; PEV, position effect variegation;
TCR-a, T cell receptor a; TFIIH, transcription factor II H.
Trang 2nucleosome positions [14–16] However, these advances
have been accompanied by a relative decrease in the
number of studies aimed at gaining an understanding
of the structural conformation of chromatin, and the
changes in chromatin structure that accompany gene
activation Furthermore, it has now become
common-place for chromatin immunoprecipitation (ChIP)
assays to be used as a surrogate for true structural
studies However, these studies cannot by themselves
give a detailed understanding of the relationships
between specific chromatin modifications and
chroma-tin architecture It is also important to recognize that
the principal function of many modifications is to
embed a specific recognizable code within chromatin
[14,17] as opposed to directly altering chromatin
con-formation per se
To understand the basis of the fundamental
mecha-nisms that lead to gene activation it is necessary to
appreciate that chromatin is by its very nature
repressed by nucleosomes and highly inaccessible The
normal process of gene activation involves the ordered
recruitment of factors that assemble on DNA in a
highly cooperative manner The key point of control
in this process is the restriction of accessibility to the
DNA sequence One obvious consequence of this is
the fact that the genome encompasses many cryptic
binding sites for transcription factors that are not
uti-lized because they do not exist in the correct context
In this review I will therefore focus primarily on the
actual chromatin structure of active genes, with regard
to nucleosomal organization and higher order
struc-ture, and the chromatin structure changes that occur
during locus activation I will discuss the nature of
transcription factor interactions with chromatin, which
can lead to localized nucleosome displacement at
DHSs within regulatory elements, as well as long
range changes in the organization and accessibility of
nucleosomes within chromatin During the course of
these discussions I will draw upon our own
experi-ences using the highly inducible human
granulocyte-macrophage colony-stimulating factor (GM-CSF) gene
as a model system that undergoes extensive
remodel-ling It is beyond the scope of this review to enter
into an extensive discussion of the role of all the
vari-ous specific histone modifications and the activities of
the different ATP-dependent chromatin remodelling
complexes There are many other reviews on these
subjects by the experts in these fields [18–27] I will
discuss in detail, however, the structural implications
of the cycle of histone acetylation and deacetylation
that accompanies cycles of transcription, and highlight
the special significance of histone H4 lysine 16
com-H3 H3
H4 H4
H2B
H2A
H2B H2A
Tetramer
Upper H2A/H2B dimer
Octamer + 146 bp DNA
H2B H2A
H3 H4
+
Fig 1 Composition of nucleosomes The assembly of the histone octamer on DNA is represented by this model which depicts the incorporation of two H3 ⁄ H4 dimers with an inner core of 60 bp
of DNA, followed by the loading of two H2A ⁄ H2B dimers onto the flanking DNA segments above and below the H3 ⁄ H4 tetramer Throughout the nucleosome, each DNA strand of the helix is con- tacted by histones at 10 bp intervals The lighter colour shades depict the bottom half of the nucleosome, and the exploded view below the octamer depicts the arrangement of the histones con- tacting 73 bp of DNA within each half Note that each H4 molecule actually bridges two turns of the DNA helix, by contacting the inner core DNA within one half of the nucleosome plus the DNA at the exit point of the opposite half.
Trang 3which is a greatly simplified version of the X-ray
crys-tal structure obtained at 2.8 A˚ resolution [32] but is
used here to convey the concept that an H3⁄ H4
tetra-mer making up the inner core is first loaded onto the
central 60 bp of DNA, followed by two H2A⁄ H2B
dimers which are loaded above and below the H3⁄ H4
tetramer onto the flanking DNA segments Most of
the genome exists in the form of regularly spaced
nucleosomes with a DNA repeat length of 180–
200 bp Most nucleosomes also recruit either histone
H1 or high mobility group (HMG) proteins
(some-times both) which bind to the outside of the
nucleo-some to form a particle known as the chromatonucleo-some,
which occupies 166 bp of DNA [29,33–35] Within
native chromatin, nucleosomes assemble into higher
order structures, and both the core histone tails and
linker histone H1 (or H5) play major roles in
main-taining higher order chromatin condensation [36–38]
However, even in the absence of histone H1, chains of
nucleosomes spontaneously assemble into a higher
order fibre 30 nm in diameter if physiological levels of
monovalent or divalent cations are present It requires
just 0.5 mm MgCl2, or 60 mm NaCl, to promote
coil-ing of 10-nm diameter fibres into 30-nm diameter
fibres [37,39] The 30-nm fibre represents the
predomi-nant type of chromatin structure observed in electron
microscopy (EM) studies of either ruptured interphase
nuclei [40] or metaphase chromosomes that have been
partially dissociated in 1 mm MgCl2 [39] The exact
nature of the structure of this fibre is still a subject of
intense debate [41], but it can potentially be
repre-sented either by a double helix with crossed linkers,
where the linkers zigzag across the centre of the fibre
[42,43], or alternatively as a simple solenoid made up
of six nucleosomes per coil [44], where the nucleosomes
interdigitate between adjacent coils [45]
Chromatin fibres are naturally highly condensed
in vivo
Under salt-free conditions, and in the absence of
his-tone H1, chains of nucleosomes can be visualized as
unfolded chains of regularly spaced 10-nm diameter
particles, giving rise to the popular ‘beads on a
string’ images Unfortunately, this textbook image
has led to the popular misconception that active
gene loci decondense completely into these unfolded
10-nm diameter fibres In reality, the eukaryotic
gen-ome is assembled in a much more condensed state
under physiological conditions, and exists in
confor-mations at least as complex as 30-nm diameter
fibres, within all but the most actively transcribed
genes [46,47]
Micrographic studies of interphase and prophasenuclei reveal that most of the genome is actuallyassembled at degrees of condensation much higherthan even the 30-nm fibre [47–49] By EM, chromatinfibres are typically seen to be 110–170 nm in diameterduring interphase [48] and 200–250 nm in diameterduring prophase [49] These high levels of chromatincondensation were also observed within active genesvia a different approach whereby megabase segments
of chromatin were fluorescently labelled inside livingcells [47] By this means it is possible to visualize genesaligned in a linear array both before and after induc-tion of transcription However, after transcription acti-vation, the level of compaction detected was still 10- to30-fold higher than the level of the 30-nm fibre [47].Similar results were obtained using fluorescencemicroscopy of arrays of steroid-inducible mouse mam-mary tumour virus (MMTV) DNA, where a DNAcompaction ratio of 50- to 1300-fold remained afterinduction of transcription [50] Hence, transcribedgenes can in some cases remain compacted to anextent far greater than the DNA packing ratio of30–40 predicted for a 30-nm fibre and 5–10 predictedfor a 10-nm fibre The exceptions to this are the highlytranscribed genes such as the ribosomal RNA geneswhich are so heavily loaded with polymerases thatmost of the nucleosomes are evicted and no conven-tional chromatin fibre remains
The concept of the 30-nm fibre as the universalbuilding block of chromatin in vivo has also been chal-lenged by an independent cryo-EM analysis of meta-phase chromosomes which depicted homogeneousgrainy images of chromatin sections with no evidencefor any discrete higher order fibre formation [51] Theinterpretation of these images was that chains of nucle-osomes within chromosomes exist primarily in a disor-dered interdigitated state, rather than conforming
to the well organized helical structures observed for
in vitroreconstituted chromatin fibres
The Balbiani rings observed in polytene somes in Chironomus tentans provide another represen-tation of very actively transcribed genes These arelooped out domains of highly decondensed chromatincontaining genes heavily loaded with polymerases Theelegant EM studies of Balbiani rings by Daneholt andco-workers [52,53] gave us one of our first glimpses ofthe true nature of transcribed chromatin In this modelsystem, sequences immediately upstream and down-stream of genes can be seen in most cases to remaincoiled as 30-nm fibres In the cases where the RNApolymerases are the most densely packed, the interven-ing DNA can be seen typically as either nucleosome-free or as a 10-nm fibre However, even in these highly
Trang 4chromo-transcribed structures there sometimes remain stretches
of condensed 30-nm fibres formed in between more
distantly spaced polymerases [52,53] This suggests that
chromatin can transiently exist as a decondensed
10-nm fibre during transcription, perhaps even
nucleo-some-free, but that coding regions return to a
conven-tional 30-nm diameter chromatin fibre once a
polymerase has passed
The challenge presented here, then, is to gain a
bet-ter understanding of the significance of the different
degrees of chromatin condensation and chromatin
modification that prevail in the nucleus, that enable
the appropriate activation of specific gene loci Clearly,
it is not sufficient to merely think in terms of
con-densed 30-nm chromatin fibres versus open 10-nm
chromatin fibres We also need to be able to define the
specific mechanisms that create a more dynamic
chro-matin structure in which nucleosomes and chrochro-matin
proteins are more mobile [15,54] For example, it is
accepted that active gene loci are less condensed and
more accessible than inactive loci, and that a passing
polymerase must at least transiently create openings in
the chromatin fibre However, in normal interphase
nuclei, it is likely that most sections of most active
genes will remain condensed to at least the level of
30-nm fibres The exceptions to this rule will be the actual
sites of ongoing transcription where individual
polyme-rases are bound and any genes which are so loaded
with polymerases that this does not permit the
reas-sembly of nucleosomes
Active chromatin domains
Evidence from a wide range of sources confirms that
active gene loci are associated with fundamental
changes in chromatin architecture across broad
domains spanning genes Electron micrographs of
interphase nuclei reveal areas of condensed
heterochro-matin and decondensed euchroheterochro-matin that are generally
assumed to represent inactive and active chromatin –
although this is now known to be somewhat of an
over-simplification, as some active genes reside within
heterochromatin Drosophila polytene chromosomes
offer one of the clearest examples of active chromatin
domains whereby active genes appear as highly
decon-densed ‘puffs’
Active chromatin domains are permissive for
transcription
It is generally accepted that active genes lie within
broad active chromatin domains that carry a variety of
modifications associated with active chromatin [18–23]
The significance of this was highlighted by a study thatfound that chromatin domains marked by H3 acetyla-tion and H3-K4 methylation were permissive for thestable expression of integrated transgenes, whereastransgenes integrated at other sites were prone tosilencing [55]
Active genes reside within extensivenuclease-sensitive domains
It was recognized in the 1970s and 1980s that tin domains encompassing active genes are at leasttwice as sensitive to DNase I digestion as non-tran-scribed genes [5–7,56–62] These studies used either Cotanalysis of DNA hybridization kinetics, slot-blot filterhybridization, or the disappearance of discrete restric-tion enzyme DNA fragments as a measure of the rate
chroma-of DNase I digestion In many cases it was found thatthese accessible domains exhibiting general DNase Isensitivity extended many kilobases upstream anddownstream of the transcription units they encom-passed For example, the chicken lysozyme activedomain extends for about 14 kb upstream and 6 kbdownstream of the gene, and is preferentially sensitive
in the oviduct which expresses lysozyme, but not inliver or erythrocytes which do not [59] In the chickenb-globin locus the DNase I sensitive domain extendsfrom 6 kb upstream to 8 kb downstream of the gene,although in this instance the coding sequences are evenmore sensitive than the immediate flanking sequences[7] In the mouse b-globin locus, the active adult b-glo-bin genes are in a more nuclease-sensitive domain thanthe inactive embryonic globin gene [58] However,increased nuclease accessibility does not mean that thechromatin fibre is completely decondensed Recentstudies suggest that active genes remain, for the mostpart, in a condensed state, with the linker regions pro-tected within the fibre and no more accessible toDNase I than the nucleosomes [63] This study alsosuggested that some of the reports of general nucleasesensitivity might in fact be attributable to the hyper-sensitivity at the DHSs within these active chromatindomains
It was once thought that one DNase I sensitivedomain would correspond to one gene plus its regula-tory elements However, this concept is now outdated,because regulatory elements can reside far from thegenes they control, sometimes existing within inactiveloci In the case of the lysozyme locus, which was ini-tially used to help establish the active domain model,
it was later found that its domain encompasses theubiquitously expressed Gas41 gene, even though thisdomain was thought to be sensitive in lysozyme-
Trang 5expressing cells only [64] In chicken embryo
erythro-cytes, the inactive lysozyme gene has almost the same
DNase I sensitivity as the active Gas41 gene [63] This
issue was also addressed by a genome-wide analysis
which found that open chromatin domains more
clo-sely correlated with gene density than gene activity,
because inactive genes can also be found within active
gene domains [65]
Demarcation of active chromatin domains
It was once proposed that active chromatin domains
would be demarcated by the rigid attachment of
nuclear matrix attachment regions (MARs or SARs)
to the nuclear skeleton [66–68] However, there is little
evidence for this [69], and MARs are often found
inside active domains or associated with enhancers
[66,70] In contrast, there are numerous examples
where the borders of active domains are defined by a
class of DNA elements termed boundary elements
(BEs) (or barrier elements) which block the spread of
repressive chromatin [71,72] In this regard, BEs may
function in a way that is not quite the same as another
class of elements termed insulators that block
enhan-cer–promoter communication but do not necessarily
demarcate active chromatin domains The terminology
here can be very confusing, however, because the two
terms are often used interchangeably, and some DNA
elements have both BE and insulator activity [71,72]
BEs were first identified in Drosophila, where they were
found to block position effect variegation (PEV) of
expression of mobile integrated transgenes containing
transposons One of the best studied such examples
exists in the Drosophila 87A7 heat shock protein 70
(Hsp70) locus where two BEs termed SCS and SCS’
directly flank an inducible active chromatin domain
spanning 12 kb These BEs function both as
enhancer-blocking insulators [69,73] and as active chromatin
domain boundaries [74,75] that block PEV [76] The
SCS and SCS’ elements are the prototypes of one of
the major classes of BE in Drosophila, which bind a
protein complex termed BEAF [77] This complex is
associated with about half of the interbands in
poly-tene chromosomes, and in many cases is present at the
borders of active genes within polytene chromosome
puffs [78]
One of the proposed mechanisms of BE function
involves the recruitment of chromatin modifying
com-plexes that create islands of active chromatin which
counteract the repressive complexes that mediate
het-erochromatin spreading [71,72] Many BEs are known
to have promoter activity and to recruit chromatin
activators, and in yeast some BEs are in fact tRNA
genes [71,72] This model of BE function is furthersupported by the fact that many components of repres-sive chromatin complexes, such as the histone H3-K9methyltransferase SUV39H1 [Su(var)3-9 in Drosophila],were themselves initially identified via mutations thatblocked PEV [79,80] These proteins are typicallyinvolved in heterochromatin spreading mediated byHP1 [71,80] Conversely, enhancer-blocking insulatorscan function by an alternative mechanism Vertebrateinsulators invariably recruit CTCF which in turnrecruits the cohesin chromosomal cohesion complex[71,81] This leads to a model whereby CTCF controlschromatin looping [82] and defines independent func-tional DNA domains within which enhancers and pro-moters can cooperate, as opposed to demarcatingactive chromatin domains
Active loci undergo extensive nucleosomemobilization
Classical models of chromatin depict chains of larly spaced nucleosomes that fold up into a helix ashighly ordered chromatin fibres However, this image
regu-is really only representative of inactive loci that tute the bulk of chromatin in the nucleus The highlyregular ordering of nucleosomes is more closely associ-ated with gene silencing, and with decreased sensitivity
consti-to DNase I [83]
Although it is well known that gene activationinduces alterations in chromatin, there are still rela-tively few studies which have assessed the organization
as opposed to the modification status of active matin Significantly, those studies which haveaddressed this issue have typically found that geneactivation is associated with extensive nucleosomemobilization which results in the formation of a highlydisorganized nucleosome array incapable of conform-ing to any of the current models of the higher orderchromatin fibre It is even possible that this highly dis-organized form of chromatin includes some nucleo-somes fused together, as there is evidence that adjacentnucleosomes can in some cases merge to form a singlefused particle [84] This type of information is difficult
chro-to gather from genome-wide studies that have definedthe average nucleosome positions, because thisapproach does not necessarily provide a meaningfulpicture of how individual nucleosomes are packagedwithin chromatin relative to each other in any one cell.Nucleosome mobilization is best visualized by elec-trophoretic size fractionation and southern blothybridization of chromatin digested with micrococcalnuclease (MNase), which cuts primarily in linkerregions This type of analysis typically reveals ladders
Trang 6of regularly spaced discrete oligo-nucleosome bands
for bulk chromatin, but a smeared pattern for active
chromatin An example of the phenomenon is
pre-sented in Fig 2A, which shows MNase digestion data
for the human GM-CSF locus in T cells [85,86] In
unstimulated T cells, where the gene is completely
silent, MNase generates very uniform ladders of evenly
spaced nucleosomes with an average repeat length of
about 190 bp throughout the GM-CSF locus [85,86]
Parallel mapping of nucleosome positions by indirect
end-labelling [85] shows that nucleosomes are
posi-tioned at 200 bp intervals at highly specific locations
throughout at least 6 kb of the locus (Fig 3A)
How-ever, after gene activation by stimulation of calcium
and kinase signalling pathways, nucleosomes out this 6 kb region adopt a highly disorganized struc-ture with nucleosomes redistributed to randompositions in both T cells and mast cells Interestingly,the degree of nucleosome position randomization is farmore extreme within the first few kilobases of the non-transcribed upstream region than within the gene itself(Fig 2A) This could mean that each cycle of tran-scription resets the normal spacing of nucleosomes.Furthermore, for genes undergoing moderate levels oftranscription, it is thought that RNA polymerase II(Pol II) proceeds via a mechanism that actually pre-vents nucleosome translocation [87] However, the situ-ation may be very different at highly transcribedgenes, where closely spaced Pol II molecules can dis-place the entire histone octamer [88]
through-As will be discussed in more detail below, there iswidespread evidence for both nucleosome repositioningand increased chromatin accessibility in the neighbour-hood of regulatory elements For example, in mastcells, GATA factors are able to bind to an accessiblenucleosome-free linker region within the GM-CSFenhancer, leading to lineage-specific repositioning ofthe flanking nucleosomes (Fig 3B) This involves therelocation of the upstream nucleosome N0 to a newposition 100 bp further upstream and the down-stream nucleosomes about 20–30 bp further down-stream A similar finding was obtained in studies ofthe MMTV long terminal repeat where Oct1 andnuclear factor 1 (NF1) were sufficient to direct nucleo-some repositioning [89] A further consequence ofGATA factor recruitment at the GM-CSF enhancer isincreased accessibility of the linker regions flanking thetwo nucleosomes located immediately downstream ofthe GATA sites (Fig 3A) [85] This appears to repre-sent a primed active state that precedes the disruption
of these same two nucleosomes upon subsequentinducible binding of nuclear factor of activated T cells(NFAT) and AP-1 (to be discussed in more detailbelow) A similar situation may exist in the humaninterleukin-4 (IL-4) locus, where a total of six nucleo-some linker regions at the 5¢ end of the gene aremore accessible specifically in type 2 T helper cells thatexpress IL-4 [90]
Nucleosome mobilization in the 3 kb region betweenthe GM-CSF enhancer and promoter is dependentupon this upstream enhancer [85] In the absence ofthe enhancer, inducible nucleosome mobilization in theupstream region is completely abolished (Fig 2B).These findings suggest that one important aspect ofenhancer function is to direct localized nucleosomemobilization within an active chromatin domain Thisimplies that enhancers can function both by recruiting
GM-CSF Enhancer
Gene probe
Non-stimulated Stimulated
Fig 2 Nucleosome mobilization within the activated GM-CSF
locus Southern blot analysis of oligo-nucleosome fragments
pro-duced by increasing amounts of MNase digestion and probed
directly with specific GM-CSF locus probes In this analysis
chroma-tin fragments were prepared from T cells before or after stimulation
of TCR signalling pathways that induce NFAT and AP-1 [85,86].
Nucleosome mobilization is characterized by a smear of random
products at early digestion points, and by the small proportion of
very close packed nucleosomes that are more resistant to MNase
and remain after increased digestion In this analysis, nucleosomes
have an average repeat length of 190 bp before mobilization,
whereas the closed packed nucleosomes have a repeat length of
150 bp after mobilization The densitometric traces of the middle
lanes are shown below each panel and reveal that the predominant
pattern is essentially random after mobilization (A) Analysis of the
intact GM-CSF locus (B) Analysis of the GM-CSF locus with a
spe-cific deletion of the 0.7 kb enhancer.
Trang 7remodellers that can act within a few kilobases and by
looping to function over larger distances GM-CSF
enhancer activation is mediated by the inducible
tran-scription factors NFAT and AP-1 which direct the
for-mation of a DHS (Fig 4, discussed in more detail
below) NFAT⁄ AP-1 complexes are thought to recruit
CBP⁄ P300 family histone acetyltransferases (HATs) as
well as SWI⁄ SNF family chromatin remodelling
com-plexes which may well account for the observed
nucle-osome mobilization [91]
Within the region of nucleosome mobilization
upstream of the GM-CSF gene, it can also be seen
that a fraction of the nucleosomes end up as fragments
of close packed nucleosomes with a repeat length of
just 150 bp which resist digestion (Fig 2A) It is
inconceivable that such a close packed arrangement
could either accommodate histone H1 or assemble into
a 30-nm chromatin fibre I have attempted to depictthis chromatin structure transition in Fig 4A, whereby
a well organized inactive chromatin fibre compacted
by histone H1 is converted to a disorganized activechromatin fibre that is probably depleted of histoneH1 Because it is so disorganized, active chromatinmay have an intrinsic resistance to folding into a rigidcompacted structure
Similar nucleosome mobilization within active locihas been observed in many model systems (which Ihave summarized previously [85]) and is not justrestricted to transcribed regions For example, in thechicken oviduct, a 2.5 kb region of chromatin justupstream of the ovalbumin gene undergoes extensivenucleosome randomization, whereby some chromatinfragments contract to a nucleosome repeat length ofabout 150 bp [92] This is also observed in mouse B
Nucleosome positions and functional binding sites in the human GM-CSF enhancer
717 Bgl II
T cells stim.
Mast cells non-stim.
T cells non-stim.
T cells (blue) and mast cells (red) before and after stimulation with 4b-phorbol 12-myri- state 13-acetate and calcium ionophore The graphs of MNase cleavage represent the ratio of the level of MNase digestion in chromatin divided by the level of cleavage for purified genomic DNA [85] The scale represents position relative to the transcrip- tion start site (B) A map showing the posi- tions of regulatory elements required for function in either T cells or mast cells Shown below are the positions that nucleo- somes and GATA-2 occupy in unstimulated
T cells and mast cells [85].
Trang 8cells expressing Igj within the coding sequences of the
Igj gene, where nucleosome mobilization extends to
just beyond the end of the transcription unit [93,94]
The role of histone H1 in chromatin accessibility
The molecular basis of the general DNase I sensitivity
observed both within and around genes is likely to be
highly complex At the simplest level, loss of histone
H1 is sufficient to reduce the level of compaction of
the chromatin fibre, and at active genes the amount of
histone H1 is reduced compared with inactive genes
[95–97] Conversely, addition of histone H1 to active
chromatin results in gene repression [98] Although
sig-nificant levels of histone H1 do remain at active loci,
the ratio of histone H1:nucleosomes is less than the
1 : 1 predicted for inactive loci, and this may be
suffi-cient to trigger a breakdown of chromatin compaction
[95] Furthermore, chromatin within nuclei stripped of
histone H1 is about two- to three-fold more sensitive
to DNase I [99], consistent with the increased level ofDNase I sensitivity typically observed at active geneloci Histone H1 is also implicated as a factor thatmaintains the differential DNase I sensitivity of themouse adult and embryonic b-globin genes [58] How-ever, it is probably safe to assume that general DNase
I sensitivity arises from the concerted effects of many
of the chromatin modifications associated with activegenes, plus the act of transcription itself For example,
a recent study found that both acetylation of H4-K16and eviction of histone H1 were required for thedecompaction of the 30-nm fibre in vitro [100]
Genetic analyses have found that histone H1 is not
as essential for correct gene regulation as previouslythought [97] Histone H1 can be eliminated from uni-cellular organisms without much impact, and reduction
of histone H1 levels in mouse stem cells to 50% ofnormal levels results in a global reduction in averagenucleosome linker length but not much effect on geneexpression [97,101] Although this reduction in H1
A Anatomy of the inducible DNaseI hypersensitive site in the GM-CSF enhancer in T cells
DNase I and MNase
Active chromatin with DHS and mobilised nucleosomes with less histone H1
Promoter
Condensed chromatin + histone H1
NFAT + AP-1
Fig 4 DHS formation and nucleosome
mobilization at the human GM-CSF locus.
(A) Model of the DHS within the human
GM-CSF enhancer induced by activation of
TCR signalling pathways that induce NFAT
and AP-1 [85,86] Prior to activation, the
locus exists as an array of regularly spaced
nucleosomes assembled as condensed
chromatin The induction of the DHS is
accompanied by the eviction of two
posi-tioned nucleosomes that otherwise occupy
two discrete sets of factor binding sites and
block binding of the constitutively expressed
factors Sp1 and Runx1 Upon activation,
NFAT and AP-1 bind cooperatively to
com-posite NFAT ⁄ AP-1 elements within each
nucleosome, and are predicted to support
the formation of enhanceosome-like
com-plexes including co-factors such as CBP and
SWI ⁄ SNF [85] In vivo footprinting
con-firmed inducible binding of NFAT, AP-1, Sp1
and Runx1 [86,233] Nuclease digestion
studies have determined that the
nucleo-somes normally occupy 150 bp of DNA
before stimulation, and are replaced by
complexes that protect 50 bp of DNA.
(B) High resolution DHS mapping of the
GM-CSF enhancer in activated T cells and
mast cells by indirect end-labelling [85] The
protected regions between zones of DNase
I hypersensitivity (arrowed) indicate the
potential presence of enhanceosomes.
Trang 9levels is tolerated by stem cells in vitro, the defect is
embryonic-lethal in mice and reveals a need for H1 in
embryonic development
Histone replacement at active gene loci
It is established that the act of transcription involves
at least partial transient displacement of histones from
nucleosomes, as well as the substitution of some of the
canonical histones with histone variants It has been
known since 1983 that active genes are enriched in
nu-cleosomes lacking one molecule of each of histones
H2A and H2B, and that these partially disassembled
nucleosomes are preferentially bound by Pol II in vitro
[102] As depicted in Fig 5, RNA polymerase can
recruit facilitator of active transcription (FACT) which
displaces one H2A⁄ H2B dimer as each nucleosome is
transcribed [103] Once the polymerase has passed, the
H2A⁄ H2B dimer is replaced There is also evidence for
more substantial histone core displacement during
transcription because histone H3.3 is highly enriched
within transcribed or recently transcribed genes [104–
106] H3.3 is synthesized during interphase whereas
H3.1 and H3.2 are synthesized during S phase This
may be one reason why H3.3 is found enriched at
active genes It was once assumed that the presence of
H3.3 in active genes was of little structural
signifi-cance, because H3 variants are structurally very similar
to each other However, it is now believed that
H3.3-containing nucleosomes are much less stable than
H3.1-containing nucleosomes [107] Furthermore, H3.3
may suppress histone H1 mediated chromatin
compac-tion, because H3.3-containing nucleosomes appear to
be unable to recruit histone H1 [108]
Regulation of chromatin structure bypoly(ADP-ribose) polymerase (PARP)Studies in Drosophila and mammals have revealed thatPARP-1, the enzyme that directs modification ofhistones by poly ADP ribosylation, can direct eithergene activation or repression [75,109,110] Theseopposing actions appear to work by distinct mecha-nisms At repressed loci, PARP-1 can function as astructural protein whereby it binds to nucleosomes at a
1 : 1 molar ratio in place of histone H1 and, like H1,
it promotes chromatin condensation [110] In this text, PARP-1 does not PARylate chromatin, and acti-vation of its enzymatic activity actually relievessilencing [110] PARP-1 binds to chromatin by engag-ing each of the two strands of DNA at the point atwhich they exit from the nucleosome, thereby opposingthe actions of transcriptional activators that mobilize
con-or disassemble nucleosomes [110]
If the enzymatic functions of PARP-1 are activated
in the presence of NAD+ it mediates the PARylation
of both histones and PARP-1 itself, and thereby motes decondensation of higher order chromatin struc-ture [75,109] However, in studies of condensedchromatin assembled in vitro in the presence of PARP-
pro-1, it was found that chromatin decondensation can beinduced by activation of PARP-1 without PARylation
of the underlying core histones and without disruption
of nucleosomes [110] In this model system, chromatindecondensation occurred primarily via auto-PARyla-tion and loss of binding of PARylated PARP-1 tochromatin
PARP-1 was also found to contribute to extensiveremodelling of nucleosomes across the DrosophilaHsp70 in response to heat shock [75] This study madethe surprising observation that nucleosomes through-out the Hsp70A locus were rendered MNase sensitiveafter just 1 or 2 min of heat shock This extensive dis-ruption or modification of nucleosomes spanned theentire region defined by the SCS and SCS’ boundaryelements, was independent of transcription, and wassuppressed by RNAi-depletion of PARP-1 [75]
Active genes partition differentially duringchromatin fractionation
Active chromatin has very different physical propertiesfrom inactive chromatin For example, minichromo-somes assembled in Xenopus oocytes partition intoinactive soluble chromatin and insoluble active chro-matin [111] Early attempts to fractionate native chro-matin into functionally distinct fractions wereperformed by digestion of nuclei with MNase followed
Ac
Ac
HDACs
Histone hexamer
H2A/H2B dimer
RNA
Fig 5 Model of the chromatin structure in the vicinity of an
elon-gating Pol II complex Histone acetylation in advance of
polymeras-es is likely to create an open chromatin structure The advancing
polymerase recruits FACT which partially disassembles the
nucleo-some, allowing Pol II to pass this barrier Once Pol II has passed,
HDACs such as Rpd3S can be recruited via dimethylated or
trime-thylated H3-K36 and act to return chromatin to the condensed
state.
Trang 10by separation based on solubility under different ionic
conditions [93,112] This involved progressively
fractionating chromatin into (i) highly soluble small
chromatin fragments readily released from nuclei
dur-ing digestion (fraction S1), (ii) the bulk of the
remain-ing chromatin which could be subsequently solubilized
from digested nuclei by extraction in 2 mm EDTA
(fraction S2), and (iii) the residue comprising highly
insoluble chromatin (fraction P) These studies found
that fraction S1 was predominantly mono-nucleosomal
and was highly enriched in transcribed genes, but was
depleted of inactive genes and histone H1; fraction S2
contained classically organized oligo-nucleosomes
depleted of transcribed genes but retaining most of the
nuclear histone H1; fraction P was composed of
disor-ganized chromatin fragments that were also enriched
in active genes [93] Hence, the S1 fraction represented
the highly accessible and extensively modified active
gene fraction containing highly acetylated
nucleo-somes, which were more soluble and could be released
from highly remodelled chromatin segments that were
tightly associated with the transcription apparatus
[113]
At first it appears paradoxical that the more
accessi-ble active genes should be split between the most and
the least soluble chromatin fractions However, the
explanation for this observation lies in the fact that
active genes are tightly associated with
multi-compo-nent transcription factor and polymerase complexes at
sites that have been termed transcription factories
[114–116] The residual insoluble fraction is in essence
equivalent to the ‘nuclear matrix’ fraction that was
shown to be enriched in active genes [117–119] While
the ‘nuclear matrix’ was originally proposed to be a
true nuclear skeleton organizing the functions of the
nucleus, it may in reality represent an aggregate of all
the active sites in the nucleus, such as transcription
factories, that remain when the inactive chromatin
fraction is removed These may be the sites bound by
MARs and may explain why MARs often exist
along-side enhancers
Chromatin structure regulation by
histone acetylation
The role of histone acetylation
Histone modifications help to create a more accessible
and dynamic chromatin environment and thereby play
a major role in making chromatin permissive for
tran-scription [54] Acetylation of lysines leads to
neutraliza-tion of the positively charged nitrogen atoms that
mediate contacts between histone tails and DNA,
ren-dering individual nucleosomes more unstable andmobile These histone tail contacts occur primarily withthe linker DNA rather than the nucleosomal DNA [21]
In contrast, other non-neutralizing modifications such
as methylation may have a less direct impact on ture, but serve as docking sites for regulatory moleculessuch as chromatin remodelling factors
struc-Acetylation of histone H4-K16 suppresseschromatin condensation within active genes
In a study of chromatin fibre dynamics, it was revealedthat acetylation of lysine 16 on histone H4 (H4-K16)was the only modification that was able to destabilizehigher order chromatin structure [120] In sedimenta-tion velocity analyses, acetylation of this one aminoacid led to a degree of chromatin fibre decompactionequivalent to loss of the entire histone H4 tail [120].The reason for this may be because H4-K16 mediatesinteractions with adjacent stacks of nucleosomes withinthe 30-nm fibre and its acetylation disrupts H4 tailsecondary structure and salt bridging [121,122] Subse-quent EM studies confirmed that acetylation of H4-K16 led to a breakdown of 30-nm compacted fibres[100] A more recent chromatin sedimentation studyalso found that H4-K16 acetylation is sufficient togreatly reduce chromatin folding, whereas combinedacetylation of H4-K5, K8 and K12 had a much moremodest effect [123]
Acetylation of H4-K16 does appear to have specialsignificance in vivo [21] Unlike AcH3-K9, which ismainly confined to promoters, AcH4-K16 is also pres-ent at elevated levels throughout the transcribedregions of active genes in human T cells [124] In astudy in yeast, mutations were introduced alone or incombination in lysines 5, 8, 12 and 16 in the gene forhistone H4 [125] Of these, the only mutation that had
a specific effect on patterns of yeast gene expressionwas the mutation in H4-K16 In Drosophila, specificacetylation of H4-K16 is an integral feature of dosagecompensation that results in a global two-fold increase
in gene activity [126] Interestingly, AcH4-K16 plays
an additional role in countering the repressive effects
of chromatin because it reduces the ability of the ISWIremodelling complex to reset active chromatin as com-pacted chromatin [127]
The HAT primarily responsible for the bulk ofAcH4-K16 in vivo is likely to be MOF in mammalsand Drosophila, and its homologue Sas2 in yeast.MOF is H4-K16 specific and was originally identified
in Drosophila as a component of the dosage tion complex [126] in association with MSL1, MSL2and MSL3 [128,129] MSL3 specifically binds to
Trang 11compensa-methylated H3-K36, which promotes recruitment of
MOF to recently transcribed regions, especially at the
3¢ ends of genes where H3-K36me3 is enriched [128]
In mammalian cells MOF also exists as part of a
sepa-rate MOF–MSLv1 complex that co-purifies with
MLL1 and WDR5 which binds to dimethylated and
trimethylated H3-K4, and this is found preferentially
bound at active promoters [128,129] In genome-wide
analyses in human cells, the enrichment of AcH4-K16
within transcribed genes is closely correlated with
binding of both MOF and Tip60 (equivalent to Esa1
in yeast NuA4) [124,130] However, in human cells
depletion of MOF, but not Tip60, results in reduced
global levels of AcH4-K16 and defective DNA damage
response [131] In mouse embryos, MOF is essential
for AcH4-K16, and loss of MOF results in
embryonic-lethal chromatin condensation [132–134] In yeast,
acetylation of H4-K16 is also required to suppress the
spread of heterochromatin, and Sas2 mutations lead to
spreading of heterochromatin mediated by repressive
Sir protein complexes [135] Conversely, the
hetero-chromatin protein Sir2 directs deacetylation of
H4-K16 and promotes heterochromatin spreading by
allowing Sir3 to bind to non-acetylated H4-K16 [21]
Although not required for acetylation of H4-K16,
the NuA4 group of HATs are essential for H4-K5, K8
and K12 acetylation [136–138] In yeast, this group is
made up of NuA4 and Piccolo NuA4 which both
uti-lize Esa1 as the HAT, and Esa1 was found to be
essen-tial for H4-K5, K8 and K12 acetylation [137] In
mammalian cells, this group is composed of two
dis-tinct complexes which employ different HATs: Tip60
which forms a NuA4-like complex, and HBO1 which
more closely resembles yeast Piccolo NuA4 [138] In
mammals, it is HBO1 and not Tip60 which is
responsi-ble for the bulk of the global H4-K5, K8 and K12
acetylation [136] Each of these HATs exists in
com-plexes that include PHD domains that interact with
methylated histone H3-K4 and⁄ or K36 [138–140]
Transcription directs transient histone
acetylation
The regulated process of transcription is accompanied
by an ordered sequence of transient histone
modifica-tions that directly impact upon chromatin structure
across transcribed genes There is also evidence that
transcription initiation is a cyclical process [141,142],
involving alternate assembly and disassembly of an
open chromatin structure at promoters [143–145], as is
described in more detail in another review paper in this
issue [146] This cyclical process is accompanied by
transient sequential histone acetylation and
deacetyla-tion, and transient recruitment of remodellers andtranscription factors
A cycle of transcription commences with the ment of transcription factors and co-factors bound atthe promoter, which modify the local chromatin struc-ture and enable the assembly of the pre-initiation com-plex In yeast, transcription factors typically recruitHATs such as SAGA and NuA3, which mainly acety-late histone H3, and NuA4 which acetylates histone H4
recruit-on K5, K8 and K12 This cascade of events leads torecruitment of transcription factor II H (TFIIH) whichphosphorylates Pol II at the serines at position 5 (Ser5)within the heptapeptide repeats of the C-terminaldomain (CTD) of Pol II [147] This modification pro-motes the recruitment of histone H3-K4 histone meth-yltransferases (HMTs) such as Set1 and MLL1,typically as part of the COMPASS complex This class
of HMTs introduces the H3-K4me3 mark, which ispredominantly found at the 5¢ ends of active or recentlytranscribed genes [148] In mammals, Set1 and MLLexist in stable association with both WDR5, a proteinthat specifically interacts with dimethylated and trime-thylated H3-K4 [22,149,150], and MOF [149] This pro-vides a mechanism to both amplify H3-K4 methylationand decondense chromatin by introducing AcH4-K16.H3-K4me3 also recruits the Isw1 chromatin remodel-ling ATPase to the promoter to prevent premature ini-tiation of transcription elongation [137]
The initiation and elongating phases of the scription cycle are characterized by distinct sets of his-tone and Pol II modifications [19,151] Followingpromoter clearance, the elongating phase of transcrip-tion is associated with phosphorylation of CTD hepta-peptide repeat Ser2 by P-TEFb⁄ cdk9 This phase oftranscription can also be regulated by specific tran-scription factors because P-TEFb can be recruited bynuclear factor jB (NF-jB), c-Myc, MyoD andGATA-1 [147] Furthermore, recruitment of P-TEFb
tran-by c-Myc is instrumental in releasing proximal pausedPol II in many mammalian genes [152] During theelongation phase, the Pol II CTD Ser2 phosphatemodification plays a direct role in recruiting the HMTSet2 which marks regions downstream of the promoterwith H3-K36me3 Histone phosphorylation can alsoact as a trigger driving the onset of transcription elon-gation in mammalian cells [153] Phosphorylation ofhistone H3S10 by Pim1 kinase enables the recruitment
of both MOF and P-TEFb via interactions involvingthe adaptor protein 14-3-3 and the bromodomain pro-tein BRD4 which is recruited via AcH4-K16 and phos-pho H3 [153]
In yeast it is apparent that the histone acetylationassociated with transcription elongation is only a very
Trang 12transient event, whereby the H3-K36me3 (or me2)
modification plays a vital role in returning the
chroma-tin structure of a gene to the deacetylated state
follow-ing a cycle of transcription [19,21,54,130,154–158]
Hence, it is possible that RNA polymerase travels
within a moving window of decondensed active
chro-matin, with the chromatin structure returning to the
condensed state once the polymerase has passed This
cycle is summarized in Fig 5 The K36me3 or K36me2
H3 modifications, which are introduced during
tran-scription, can function as a docking site for Eaf3,
which is a component of the yeast Rpd3S histone
deacetylase (HDAC) complex [156] This complex
directs histone deacetylation in the wake of the
tran-scribing polymerase and serves the important function
of suppressing spurious transcription from cryptic
pro-moters within genes [19,21,154–157,159] Because this
process maintains the body of active genes in a
con-densed state most of the time it serves to suppress
cryptic transcription initiation This is one reason why
active genes are not routinely observed as unfolded
10-nm diameter chromatin fibres Curiously, a similar
mechanism also operates immediately downstream of
the promoter in yeast genes, where H3-K4me2 engages
Set3 to recruit an HDAC complex containing the
Rpd3-like protein Hos2 [160] This is thought to limit
the spread of nucleosome modification emanating from
promoter-associated factors Hence, the maintenance
of a condensed chromatin structure throughout the
entire body of a gene may rely on the combined
actions of Set3 at the 5¢ end and Set2 at the 3¢ end
In yeast the cycle of transient co-transcriptional
chromatin opening may be driven by factors carried
by the polymerase itself The recruitment of SAGA
to transcribed coding regions is also supported by
Pol II CTD Ser5 phosphorylation by TFIIH
[161,162] SAGA acetylates nucleosomes on histone
H3 and aids their eviction during transcription, and
is required for efficient transcription [161,162] The
phosphorylation of Pol II CTD Ser5, plus the histone
H3 methylation by Set1 or Set2, also function to
promote co-transcriptional recruitment of NuA4, and
the subsequent acetylation of H4 promotes
recruit-ment of RSC which destabilizes nucleosomes [163]
However, the specific function of CTD Ser5
phos-phorylation during transcription is unclear, because
TFIIH kinase activity can be suppressed without
major defects in the function of the elongating
poly-merase and its role may be more closely related to
mRNA capping [164]
The acetylation of H4-K16 by other HATs may
represent an additional important aspect of this cycle
that enables transcription elongation, because this
modification is known to be sufficient to triggerunfolding of the condensed 30-nm diameter fibre.However, whether AcH4-K16 is the key target forRpd3S within transcribed genes in yeast is notentirely clear because, in contrast to human T cells[124], AcH4-K16 is not typically enriched withinactive gene coding sequences in yeast [165], andRpd3S is required for the deacetylation of all sitesexcept for H4-K16 within heterochromatin [166] Nev-ertheless, in Drosophila there is a direct relationshipbetween H4-K16 acetylation and H3-K36 methyla-tion, whereby a reduction in H3-K36me3 leads to anembryonic-lethal accumulation of acetylation specifi-cally at H4-K16 [159] In this study it was found thatdimethylated and trimethylated H3-K36 had opposingeffects Hence, it was suggested that H3-K36me2might function first to recruit an H4-K16 HAT such
as MOF to enable transcription elongation, followed
by conversion of the dimethyl to a trimethyl stateand the recruitment of an HDAC to reform arepressed state The principal role of H3-K36 methyl-ation in mammalian cells remains far from clear Thismodification can recruit the HATs MOF and HBO1,and both factors are found enriched within activegenes [124,130,140] H3-K36me3 is recognized byboth JADE1, which is associated with HBO1 [140],and MSL3, which is associated with MOF [128].There is additional evidence from genome-wideanalyses suggesting that an acetylation⁄ deacetylationcycle also occurs during transcription in human Tcells [124,130] The levels of both HATs and HDACs,especially HDAC6, are higher within active genes,and the levels present are in each case directly pro-portional to gene activity [130] The level of HDACrecruitment also increases in direct proportion to thelevel of histone acetylation present In contrast, only
a minor proportion of HDACs are found associatedwith inactive genes, and when HDACs are presentthey are found in genes that are transcriptionallypoised by H3-K4 methylation [130] A similar situa-tion exists in yeast where the HDAC Hos2 was found
to be both preferentially associated with active genesand required for efficient transcription [167] Overall,these studies suggest that HDACs do indeed play avital role in controlling the cycle of gene transcrip-tion, and not just gene repression However, one dif-ference between yeast and humans is that, in yeast,HDACs are recruited to active genes via H3-K36me3,whereas in humans both Tip60 and HDAC6 appear
to be recruited to active genes directly by lated Pol II [130]
phosphory-Two key issues that remain to be fully resolved inthis model of a cycle of histone acetylation and deacet-
Trang 13ylation are (a) which specific histone modifications are
the most important to open up chromatin structure
ahead of Pol II and enable efficient transcription, and
(b) what specific mechanisms might acetylate active
genes in association with the elongating Pol II In
addition to yeast SAGA and NuA4, the mammalian
HAT HBO1 is also potentially able to be recruited to
and to acetylate H4 on K5, K8 and K12, as well as
H3 throughout active gene coding regions [139,140]
HBO1 may itself promote transcription elongation,
and it can be recruited to transcribed chromatin via its
close interaction with ING4 which binds H3-K4me3
and JADE1 which binds to H3-K36me3 [139,140]
JADE1 also has an additional PHD domain that
inter-acts with non-methylated histone H3, meaning that it
can direct recruitment of HBO1 both in advance of
and in the wake of a transcribing polymerase
The other most likely candidates driving histone
acetylation during transcription are the Elongator
complex, which has intrinsic HAT activity [168], and
COMPASS, which can promote histone acetylation
indirectly [22] Although COMPASS is mainly
associ-ated with promoters, both COMPASS and Elongator
can travel together with the elongating polymerase
COMPASS employs Set1 to introduce methylated
H3-K4 which can then recruit HAT complexes
H3-K4me3 is recognized by Chd1 within the SAGA
complex [22] and by Yng1 within the NuA3 complex
[169] The Elp3 component of the Elongator complex
is a Gcn5-like HAT which, like SAGA and NuA3,
preferentially targets histone H3 and so is not an
obvi-ous candidate driving H4-K16 acetylation [170] This
is more likely to require a histone H4 HAT such as
Sas2, Tip60 or MOF However, the use of
arginine-substituted histones in yeast indicated that H3-K14
and H4-K8 are both significant targets of the human
Elongator HAT Elp3 [171]
Paradoxically, another potential mechanism for
acet-ylation also involves H3-K36me3 Eaf3, which
recog-nizes H3-K36, is a component of both the HAT NuA4
and the HDAC Rpd3S [21,154,155] Furthermore,
NuA4-dependent acetylation of H4-K8 is decreased in
the absence of H3-K36 methylation [137] However,
this mechanism of NuA4 recruitment to sites of
tran-scription seems unlikely because yeast NuA4 does not
bind to nucleosomes containing methylated H3-K36
[172] Rpd3S recognizes this modification via the
com-bined actions of Eaf3 and the PHD domain protein
Rco1 [172] If the Nu4⁄ Yng2 PHD domain is replaced
by the Rpd3S⁄ Rco1 PHD domain, then this results in
mis-targeting to H3-K36me3 and also leads to
activa-tion of cryptic promoters within genes [172]
Further-more, mutation of Eaf3 results in an increase in
histone acetylation, indicating that it cannot be theprincipal factor recruiting HATs to sites of transcrip-tion [154,155]
Histone acetylation regulates not just the unfolding
of the 30-nm fibre but also influences the ability tounwrap DNA from the nucleosome Histone H3-K56plays a special role in this context This lysine is part
of the globular core region of histone H3, and islocated at the point where the DNA exits from thenucleosome Acetylation of H3-K56 weakens theinteractions with DNA at this critical position andthereby contributes to nucleosome disassembly at pro-moters [173,174] However, this acetylation event isnot thought to be coupled directly to the transcrip-tion cycle, but involves the replacement of disassem-bled nucleosomes at promoters with newlysynthesized acetylated histones [173] This means thatnucleosomes at promoters are likely to be highlydynamic once they have undergone a cycle of nucleo-some replacement The many additional lysine acety-lation events directed by transcription factors willfurther loosen contacts between histone tails andDNA to create a more open structure that is morereadily mobilized by remodellers and transcriptioncomplexes
Localized chromatin modifications within regulatory elements
Gene expression control by distal and proximalregulatory elements
It is typical for higher eukaryotic genes to be trolled not just by the proximal promoter but by one
con-or mcon-ore distal elements as well These include elementsthat have been defined as either enhancers or locuscontrol regions (LCRs), depending on the assay used
to identify them In some cases these elements arelocated far upstream or downstream, or inside genes,and even inside adjacent genes Hence, it is not alwaysobvious which elements control which genes, and theidentification of essential distal elements typically has
to be accomplished by experimental means usinggenetic manipulation
The first LCR to be formally identified was themammalian b-globin LCR, which exists as a cluster ofDHSs, including a classical enhancer, several kilobasesupstream of the b-globin gene cluster [175] This LCRwas first found to be required for correctly regulatedexpression of transgenes and for the expression of allglobin genes, but it was then found that it is not actu-ally essential for the maintenance of an open chroma-tin structure [176] In essence, an LCR is a region that
Trang 14includes an enhancer that is required for correct
expression of a transgene
The concept of a gene having a single essential LCR
is itself somewhat outdated, as others have established
that transgenes can also be correctly expressed by
inserting a full complement of cis regulatory elements,
and there is not necessarily one dominant control
ele-ment [177] Furthermore, gene locus activation
typi-cally involves direct interactions within chromatin that
bring multiple upstream and downstream elements
together with the promoter at sites that have been
termed active chromatin hubs [178–181] This process
involves chromatin looping to bring all these elements
together Furthermore, active chromatin hubs can
incorporate several co-expressed active genes [182],
plus their enhancers, and may in fact be no different
to the active sites termed transcription factories
[183,184] The co-localization of many active
promot-ers and enhancpromot-ers at one site is an efficient way of
sus-taining active transcription by mainsus-taining threshold
levels of polymerases and transcription factors
How-ever, there is also some evidence indicating that any
one specific enhancer can only activate one target gene
at any one moment in time [185]
Distal regulatory elements such as enhancers and
LCRs recruit many of the same factors as promoters,
and in many cases even support non-coding (nc) RNA
transcription [186,187] Their main mechanism of
action may therefore be simply to supply factors to the
promoter at the chromatin hub However, some distal
elements appear to function not by looping but by
directing ncRNA transcription in the direction of the
locus to be activated For example, in the case of the
pituitary-specific human growth hormone locus,
ncRNA transcription initiates within an LCR located
15 kb upstream of the gene and is required for efficient
expression, even though it does not reach the gene
itself [188] These ncRNA transcripts proceed towards
the gene, through a non-expressed B cell specific gene,
but terminate several kilobases before the actual
growth hormone gene In other cases, such as the yeast
fbp1 gene, ncRNAs initiate upstream of the gene,
tran-scribe through the gene, and progressively convert the
chromatin to an open configuration [189] Within the
b-globin locus, ncRNA transcription plays a role in
the development control of globin gene switching from
fetal e-globin to adult b-globin gene expression [190]
Genetic recombination in T cells and B cells is also
enhanced by ncRNA transcription In the T cell
recep-tor a (TCR-a) locus, ncRNA transcription from an
upstream element proceeds through the V regions, and
is required for efficient recombination, presumably
because it establishes an open chromatin structure
[191] The IgH locus intronic transcripts may serve thesame function in B cells [187]
Active promoters appear as nucleosome-freeregions together with variant histones
As summarized in a recent review paper [26], ers can be divided into (a) constitutively active openpromoters that are intrinsically depleted of nucleo-somes and (b) highly regulated covered promoterswhere specific transcription factors act to disrupt apositioned nucleosome and create a nucleosome-depleted region [26,192] Nucleosomes tend to natu-rally assemble at intrinsically defined positions,determined by the underlying sequence, and nucleo-some positioning evolves in parallel with the specificmode of regulation of a promoter [193] Open promot-ers often contain poly(dA):poly(dT) tracts which arenot easily assembled into nucleosomes Along the samelines, promoters can also be roughly divided into (a)TATA-containing promoters where the TATA isblocked by a nucleosome which can be disrupted byregulated factors which activate the gene, and (b)TATA-free constitutively active promoters that intrin-sically exist in nucleosome-free regions [26,194] Cov-ered promoters are naturally more dependent onchromatin remodelling complexes [26]
promot-Constitutively open promoters, as well as inducedactive promoters, are typically depleted of a singlenucleosome upstream of the transcription start site,and have positioned nucleosomes directly adjacentwhich contain the histone variant H2AZ [195–197].Regions immediately downstream of active or recentlyactive promoters are also enriched in the histone vari-ant H3.3 [104–106] One reason why histones H2AZand H3.3 are enriched at sites of transcription isbecause they are the predominant replacement histonesused by the genome during interphase This is alsoconsistent with studies showing that there is a highrate of turnover of nucleosomes at active promotersand their flanking regions [15,26] In yeast, H2AZ isfound at most promoters, not just active promoters[26,196] However, another major reason for the pro-moter-specific localization of H2AZ is the genome-wide INO80-directed removal of non-acetylated H2AZ
at sites other than promoters [198]
Although open promoter regions are often assumed
to be nucleosome-free, a recent study has found thatDHSs throughout the genome are occupied by highlyunstable nucleosomes containing both H2AZ and H3.3[107,199,200] This specific combination renders nucle-osomes so unstable that they typically either disassem-ble or are digested away during the assay process of