[13:22 16/12/2010 Bioinformatics-btq637.tex] Page: 147 147–152
BIOINFORMATICS ORIGINAL PAPER
Vol. 27 no. 2 2011, pages 147–152
doi:10.1093/bioinformatics/btq637
Genome analysis Advance Access publication November 11, 2010
Differential effects of chromatin regulators and transcription
factors on gene regulation: a nucleosomal perspective
Dong Dong1, Xiaojian Shao1,2 and Zhaolei Zhang1,3,4,∗
1Donnelly Centre for Cellular and Biomolecular Research, University of Toronto, 160 College Street, Toronto, ON,
M5S 3E1, Canada, 2College of Science, China Agricultural University, Beijing 100083, China, 3Department of
Molecular Genetics, University of Toronto, 1 King’s College Circle, Toronto, ON, M5S 1A8, Canada and 4Banting and
Best Department of Medical Research, University of Toronto, 112 College Street, Toronto, ON, M5G 1L6, Canada
Associate Editor: John Quackenbush
ABSTRACT
Motivation: Chromatin regulators (CR) and transcription factors (TF)
are important trans-acting factors regulating transcription process,
and many efforts have been devoted to understand their underlying
mechanisms in gene regulation. However, the inﬂuences of CR and
TF regulation effects on nucleosomes during transcription are still
minimally understood, and it remains to be determined the extent
to which CR and TF regulatory effect shape the organization of
nucleosomes in the genome. In this article we attempted to address
this problem and examine the patterns of CR and TF regulation
effects from the nucleosome perspective.
Results: Our results show that the CR and TF regulatory
effects exhibit different paradigms of transcriptional control in
Saccharomyces cerevisiae. We grouped yeast genes into two
categories, ‘CR-sensitive’ genes and ‘TF-sensitive’ genes, based
on how their expression proﬁles change upon deletion of CRs or
TFs. We found that genes in these two groups have very different
patterns of nucleosome organization. The promoters of CR-sensitive
genes tend to have higher nucleosome occupancy, whereas the
promoters of TF-sensitive genes are depleted of nucleosomes.
Furthermore, the nucleosome proﬁles of CR-sensitive genes tend to
show more dynamic characteristics than TF-sensitive genes. These
results reveal that the nucleosome organizations of yeast genes have
a strong impact on their mode of regulation, and there are differential
regulation effects on nucleosomes between CRs and TFs.
Availability: http://www.utoronto.ca/zhanglab/papers/bioinfo_2010/
Contact: zhaolei.zhang@utoronto.ca
Supplementary information: Supplementary data are available at
Bioinformatics online.
Received on August 14, 2010; revised on November 3, 2010;
accepted on November 5, 2010
1 INTRODUCTION
Transcriptional regulation in eukaryotic cells is a very important
and complicated process. In the model organism Saccharomyces
cerevisiae, this process involves a precise orchestration of complex
molecular events, and is mediated by a series of regulators to create
transcription factories, such as transcription factors (TFs), chromatin
regulators (CRs) and the general transcription machinery with
∗To whom correspondence should be addressed.
related regulators (such as RNA polymerase molecules) (Berger,
2000; Li et al., 2007; Orphanides and Reinberg, 2002; Pugh, 2000).
Many studies have attempted to investigate the mechanisms of
TFs and CRs and how they work together to regulate the gene
transcription process (Buck and Lieb, 2006; Field et al., 2008;
Hartley and Madhani, 2009; Kim and O’Shea, 2008; Yarragudi
et al., 2004). In yeast, the activation of TFs is usually affected by
environmental stimuli through signaling pathways, which further
determines the level of mRNA abundance (Harbison et al., 2004).
One common strategy is to use regulatory feedback loops, in
which the TFs are embedded in the feedback loops and tune gene
expression with the corresponding needs (Amit et al., 2007; Segal
et al., 2003). CRs and TFs are often coupled together and jointly
regulate the transcriptional state of speciﬁc genes (Li et al., 2007).
Some of the CRs can utilize the energy of ATP hydrolysis to affect
the interaction of histone and DNA as a result of nucleosome
delocalization, such as histone sliding, eviction or replacement
(Li et al., 2007; Rando and Chang, 2009). This in turn affects the
accessibility of TFs to their binding targets, leading to transcriptional
activation or repression.
Gene transcription in the cell need to be tightly regulated, not only
to optimize the expression level at different physiological conditions,
but also to provide the cells the capacity to quickly adjust genes’
expression level in response to external signals and perturbations.
It has been recognized that, in addition to regulation by TFs,
CRs also play important roles in regulating gene expression. The
arrangements of nucleosomes along the DNA sequences are pivotal
mechanism inﬂuencing gene regulation (Choi and Kim, 2009; Field
et al., 2009; Lam et al., 2008; Lee et al., 2006). A number of recent
genome-wide in vivo and in vitro experiments have revealed some
general principles that govern nucleosome organization (Kaplan
et al., 2009; Lee et al., 2007). Although it was considered that
DNA sequence itself largely determines the mode of binding by
nucleosomes, this concept is still under debate (Stein et al., 2009;
Zhang et al., 2009). For example, it has been reported that the
nucleosome positioning can be the consequence of the regulatory
activities of trans-acting factors (Badis et al., 2008; Buck and Lieb,
2006; Hartley and Madhani, 2009; Kim and O’Shea, 2008; Lam
et al., 2008; Li et al., 2007). Among these trans- acting factors, TFs
can compete with nucleosomes for accessing to the DNA sequences
(Komili and Silver, 2008), and CRs can help to conduct chromatin
assembly and organization, leading to nucleosome arrays of uniform
spacing (such as ISWI family) (Ito et al., 1997; Langst et al.,
© The Author 2010. Published by Oxford University Press. All rights reserved. For Permissions, please email: journals.permissions@oup.com 147
atMasarykUniversityonFebruary21,2011bioinformatics.oxfordjournals.orgDownloadedfrom
[13:22 16/12/2010 Bioinformatics-btq637.tex] Page: 148 147–152
D.Dong et al.
1999). For example, Whitehouse and Tsukiyama (2006) showed
that ATPase Isw2 acts to position nucleosomes over less favorable
sequence elements at the POT1 promoter. About 35% of Isw2
regulated genes (∼400 genes) were subject to detectable chromatin
remodeling compared with wild-type cells, and the transcription of
these genes can be initiated after Isw2 is deleted (Whitehouse et al.,
2007).
In this study, we analyzed the different effects of CR and TF
regulations on nucleosomes, e.g. nucleosome positioning, packing,
dynamics and histone activities. We also described how nucleosomal
effects can have an impact on gene expression. Our results indicated
that TFs and CRs might have distinct strategies on the gene
regulation by inﬂuencing nucleosome organization at the promoter
region.
2 MATERIALS AND METHODS
2.1 Gene expression data
In this study, transcription plasticity, mRNA abundance and transcription
frequency were regarded as effects of transcriptional regulation.
Transcription plasticity data were obtained from Tirosh and Barkai (2008),
which was calculated as expression variation among ∼1500 transcription
proﬁles. mRNA abundance and transcription rate were retrieved from
Holstege et al. (1998).
To study the CR and TF regulatory effects on transcriptional regulation,
we compiled two previously assembled expression compendiums in
S.cerevisiae, in which various CR and TF genes were either deleted or
mutated (Hu et al., 2007; Steinfeld et al., 2007). (i) The CR effect (CRE):
we removed the expression proﬁles under histone perturbations, such
as the deletion of histone tails (H3 and H4), histone methyltransferase,
acetyltransferase, deacetyltransferase, etc., and only the deletions of CR
proteins were chosen. The ﬁnal reﬁned CR dataset contains only104
expression proﬁles for CR gene perturbation. (ii) The TF regulation effect
(TRE): expression compendium with individual deletions of 269 TFs was
compiled from Hu et al. (2007). We further checked the percentage of missing
expression values for each gene, and genes with >10% missing expression
values were excluded in the ﬁnal reﬁned data. Following previously
described procedures (Choi and Kim, 2008), we normalized the reﬁned
expression data under each perturbation condition, and then calculated CR
and TF regulation effects as the average of absolute values of logarithm
of the expression changes across these trans-acting factors perturbations,
respectively.
To detect the inﬂuence of CR and TF regulation effects on nucleosomes,
we determined two cohorts of genes that are sensitive to the perturbations
of CRs and TFs (the top 20% genes by ranking CRE and TRE, respectively)
based on the assumption that signiﬁcantly differentially expressed genes tend
to be the targets of CRs and TFs. To more speciﬁcally reveal the distinct
regulatory mechanism between CR and TF, genes categorized into both
cohorts were not included in any group. These remaining genes are referred
to as ‘CR-sensitive genes’ and ‘TF-sensitive genes’, respectively.
2.2 Nucleosome data
2.2.1 Nucleosome occupancy High-throughput nucleosome occupancy
data measured in vivo and in vitro were retrieved from Kaplan et al.
(2009); average nucleosome occupancy at the promoter region [500 bp
upstream to 100 bp downstream of the transcription start site (TSS)] was
calculated. To compare the differences in nucleosome occupancy between
genes, we redeﬁned the nucleosome occupancy over the promoter region as
the lowest average nucleosome occupancy across any 100 bp window in the
200 bp region upstream of the TSS. This metric can best reﬂect the ‘open’
and ‘closed’ state of the nucleosome positioning in the promoter regions
(Supplementary Table S4).
2.2.2 Nucleosome dynamics We compared several published nucleosome
occupancy datasets to reveal the nucleosome dynamic characteristics
(Supplementary Table S1). Here, we mainly focused on the ‘−1 nucleosome’
(the ﬁrst one upstream the TSS) and ‘+1 nucleosome’ (the ﬁrst one
downstream the TSS) of each gene. (i) Nucleosome occupancy in the normal
and heat-shock conditions were obtained from Shivawamy et al. (2008).
They measured the nucleosome occupancy data of yeast S288c strain using
deep-sequencing method. To compare the nucleosome occupancy patterns
in the normal and heat-shock conditions, we calculated the percentage of
nucleosomes in the promoter region that had dramatic changes (>45 bp)
before and after heat shock; (ii) With more and more nucleosome occupancy
data becoming available, it offers new opportunity to study nucleosome
dynamics. Two recently published deep-sequencing-based nucleosome
occupancy data were taken from Mavrich et al. (2008) and Jiang and Pugh
(2009), respectively. We chose these data because they were performed on
the same yeast strains (BY4741) and under normal conditions in the same
laboratory, and mapped the observed discrepancies of nucleosome occupancy
between two datasets. Then, we calculated the fraction of inconsistent
nucleosomes (changes >30 bp); (iii) Kaplan et al. (2009) measured in
vivo nucleosome organization from the cells grown in different growth
medium (in YPD, galactose and ethanol). We calculated the differences in
nucleosome occupancy signals in the promoter region (500 bp upstream to
100 bp downstream of TSS) using analysis of variance (ANOVA) method,
and genes with signiﬁcant nucleosome packing differences (P<0.001) were
selected.
2.2.3 Nucleosome fuzziness (i) Lee et al. (2007) determined a genomewide
nucleosome map and deﬁned nucleosome occupancy as three
states: unoccupied linker region, fuzzily positioned nucleosomes and wellpositioned
nucleosomes. We used the deﬁned fuzzily positioned nucleosomes
in this study, and calculated the fraction of genes with fuzziness region
>50 bp. (ii) Genome-wide nucleosome fuzziness data were taken from
Mavrich et al. (2008), which measured the nucleosome phasing and
quantiﬁed nucleosome fuzziness by counting the standard deviation of tag
distances from consensus position. To examine the nucleosome fuzziness
relative to the TSSs, the data were plotted as a moving average of 500
nucleosomes along the focal genes (500 bp upstream to 1000 bp downstream
to TSS).
2.3 Epigenetic marks
H2A.Z data were obtained from Albert et al. (2007). We restricted the
10% highest H2A.Z occupancy as H2A.Z containing, and calculated the
percentage of genes containing H2A.Z in the promoter region. Histone H3
turnover rates were taken from Dion et al. (2007).
2.4 TF binding sites and TATA box containing genes
TF binding sites were obtained from MacIsaac et al. (2006), which reanalyzed
the previously published genome-wide ChIP-chip data (Harbison
et al., 2004). We only selected those binding sites that have a P-value of
0.001 or better and are conserved in two related yeast species. TATA box
containing genes were taken from Basehoar et al. (2004).
2.5 Partial correlation
In this work, a non-parametric Spearman’s rank correlation with
corresponding statistical test was used since it has no prior assumption on
data distribution. It was always used to uncover linear association between
two ranked variables. To perform the joint variable analysis, non-parametric
Spearman’s partial correlation and the corresponding signiﬁcance test were
used to detect which measurements are the most inﬂuential features. Partial
correlation can measure the degree of association between two variables with
the effect of controlling variables removed. For example, when examining
(Yuan et al., 2006) which of the two measurements y or z has stronger
correlation with x, we can compare the values of ρx,y|z and ρx,y|z. The bigger
148
atMasarykUniversityonFebruary21,2011bioinformatics.oxfordjournals.orgDownloadedfrom
[13:22 16/12/2010 Bioinformatics-btq637.tex] Page: 149 147–152
Differential effects of CRs and TFs on gene regulation
one means that it has stronger association with x. ρx,y|z indicates the partial
correlation between x and y given a set of n controlling variables z. It is
deﬁned as:
ρx,y|z =
ρx,y −ρx,zρy,z
(1−ρ2
x,z)(1−ρ2
y,z)
where ρx,y is the correlation between x and y.
3 RESULTS
3.1 CR and TF have different regulatory effects on
gene transcription
As described in a previous work (Choi and Kim, 2009), CRE
and TRE were calculated from two expression compendiums, in
which individual CRs and TFs were perturbed and genome-wide
expression changes were quantiﬁed, respectively. To determine how
CRE and TRE inﬂuence transcriptional regulation, we collected
data for transcriptional plasticity (ﬂuctuation in gene expression
under various conditions) (Ihmels et al., 2002; Tirosh and Barkai,
2008), mRNA abundance and transcription frequency (Holstege
et al., 1998) and regarded them as proxies of different aspects
of transcription. Table 1 shows that both CRE and TRE are
signiﬁcantly associated with transcriptional plasticity, whereas CRE
has higher correlation with transcriptional plasticity than TRE
(ρ=0.68, P-value ≈0 for CRE, and ρ=0.28, P-value <1e-20 for
TRE, respectively). It is in agreement with previous ﬁndings that
CRs have greater effects on gene expression variation than TFs
(Choi and Kim, 2009), and suggested that chromatin regulation
plays a pivotal role in expression ﬂuctuations caused by changes
in chromatin organization.
It is possible that, even though this observation is very signiﬁcant,
it is only conﬁned to a few gene categories such as TATA box
containing genes (Basehoar et al., 2004). To control for potential
biases, partial correlation coefﬁcients were calculated to measure the
association strength between transcriptional plasticity and CRE/TRE
after controlling TATAbox containing genes, respectively. The result
showed that ρtranscriptionalplasticity,CRE|TATAbox =0.61 (P ≈ 0) and
ρtranscriptionalplasticity,TRE|TATAbox =0.24 (P < 1e-20). The partial
correlation coefﬁcients when TATA box is controlled are similar
to the corresponding correlations without controlling TATA box,
which indicates that the results are independent of TATA box
presences. When comparing the inﬂuences of CRE and TRE on
Table 1. Comparison of the inﬂuences of CR and TF regulation effect on
transcription plasticity, mRNA abundance and transcription frequency in
S.cerevisiae
Transcription mRNA Transcription
plasticity abundance rate
CRE
TRE uncontrolled 0.68 (<1e-20) 0.05 (0.18) 0.06 (0.13)
TRE controlled 0.61 (<1e-20) −0.01 (0.88) −0.01 (0.86)
TRE
CRE uncontrolled 0.28 (<1e-20) 0.27 (1e-14) 0.25 (1e-13)
CRE controlled 0.24 (<1e-20) 0.26 (1e-14) 0.23 (1e-13)
Note: when TRE and CRE were uncontrolled, Spearman’s correlation coefﬁcients were
shown; when CRE or TRE was controlled, partial correlation coefﬁcients were shown;
the numbers in parentheses are P-values measured based on the null hypothesis that
there is no signiﬁcant relationships.
mRNA abundance and transcription frequency, we found that TRE
is more signiﬁcantly correlated with mRNA abundance (ρ=0.05,
P=0.18 for CRE, and ρ=0.27, P = 1e-14 forTRE, respectively) and
transcription frequency (ρ=0.06, P = 0.13 for CRE, and ρ=0.25,
P = 1e-13 for TRE, respectively) than that of CRE (Table 1),
indicating that transcript levels are largely determined by the binding
of TFs to DNA target.
We next asked whether CRE and TRE can collectively explain
the transcription regulation program in yeast. We carried out the
joint effect analysis of CRE and TRE on transcription regulation,
i.e. we calculated the partial correlations by controlling CRE or
TRE, respectively. The result clearly showed that CRE plays more
important roles in inﬂuencing transcription plasticity, whereas TRE
is more dominant in controlling mRNA abundance and transcription
frequency (Table 1). These results further implicated that TF
and CR have distinct roles in transcriptional regulation. CRs are
more responsible to expression changes by regulating chromatin
organization, and have more dynamic characteristics, whereas
TFs are responsible for initiating gene transcription and properly
regulating mRNA synthesis.
3.2 CR and TF regulatory effects on nucleosome
organization
We next asked whether these two classes of regulators also have
different effect on nucleosome organization. The rationale for this
study is motivated by the fact that the nucleosome structure can
have a profound impact on gene expression (Lee et al., 2007; Tirosh
and Barkai, 2008). To detect the different inﬂuences of CRE and
TRE on the nucleosome organization, we determined two cohorts
of genes that are sensitive to the perturbations of CRs and TFs,
respectively (see Section 2, Supplementary Table S2). We next
examined nucleosome organizations in the promoter region for
the genes in these two groups, as recent studies have shown that
nucleosome positioning plays an important role in gene expression
activation (Lee et al., 2007; Tirosh and Barkai, 2008). After plotting
the average nucleosome occupancy measured by experimental
method in vivo (Kaplan et al., 2009), we observed a clear open
state (nucleosome depleted) for TF-sensitive genes and a closed
state (nucleosome occupied) for CR-sensitive genes, respectively
(Fig. 1A). To quantify such difference in nucleosome occupancy,
we devised a metric, the Lowest Average Nucleosome Occupancy
(termed as ‘LANO’), which was calculated over 100 bp sliding
windows in the 200 bp region upstream of the transcriptional start
site. As shown in Figure 1A, indeed we found the TF-sensitive
genes have stronger nucleosome depletion (lower LANO value)
in the promoter region than CR-sensitive genes (Wilcoxon rank
sum test, P = 4.1e-3). As genes in both cohorts have relatively
higher level of transcription plasticity, our result seems to contradict
with previous ﬁndings that genes with higher level of nucleosome
occupancy tend to have dramatic transcriptional plasticity (Lee et al.,
2007; Tirosh and Barkai, 2008). With the notion that genes with
open nucleosome organization at the proximal region of TSS are
constitutively expressed, we dissectedTF-sensitive genes into highly
expressed and lowly expressed groups by ranking their mRNA
abundances. By comparing the ‘LANO’ scores between these two
groups of genes, we found that there was no signiﬁcant difference
between them (Wilcoxon rank sum test, P=0.59). It indicated
that nucleosome depletion in promoters cannot guarantee higher
149
atMasarykUniversityonFebruary21,2011bioinformatics.oxfordjournals.orgDownloadedfrom
[13:22 16/12/2010 Bioinformatics-btq637.tex] Page: 150 147–152
D.Dong et al.
Fig. 1. Different strategies of CR and TF regulation effect on nucleosome
organization. (A) Nucleosome occupancy of the promoter regions (500 bp
upstream, ∼100 bp downstream relative to TSS) in TF- and CR-sensitive
genes. The promoter regions of TF-sensitive genes show a nucleosomefree
region, whereas the promoter regions of CR-sensitive genes
reﬂect nucleosome occupied organization. The dash curve represents the
nucleosome occupancy of all yeast genes. (B) Pearson’s correlation of
promoter regions and nucleosome-free regions (NFR, 200 bp upstream,
∼50 bp upstream relative to TSS) of nucleosome occupancy measured in
vivo and in vitro. (C) TF binding sites of CR- and TF-sensitive genes under
the promoter regions and nucleosome-free regions. The white bars represent
the results for all yeast genes, and the black and gray bars represent the
results for CR and TFsensitive genes, respectively.
expression level of genes, and transcription plasticity does not mean
higher level of nucleosome occupancy.
Until now, the concept of nucleosome positions being encoded by
DNA sequences is still under debate. we were interested to examine
the role of DNA sequence in determining nucleosome positioning
for CR- and TF-sensitive genes. In this work, we compared the
relationship between in vitro and in vivo maps of nucleosome
organization at promoters, and found no difference between TF- and
CR-sensitive genes (ρ=0.70 and 0.69, respectively). Intriguingly,
signiﬁcant difference was found at the nucleosome depleted regions
(Fig. 1B, ρ=0.6 and 0.5, Wilcoxon rank sum test, P-value ≈0).
TF-sensitive genes have a relative weaker effect on the positioning
of nucleosomes than CR-sensitive genes, which indicated that TFs
dramatically inﬂuence the resulting nucleosome organization in vivo.
We next tested whether this difference is caused by the competition
between TFs and nucleosomes to access DNA as we examined
the difference of TF binding sites in promoters between these
two cohorts of genes. We found TF binding sites are more highly
enriched in the promoters of CR-sensitive genes than TF-sensitive
gene promoters. Interestingly, when focusing on the nucleosomefree
region, TF-sensitive genes have signiﬁcantly more TF binding
sites than CR-sensitive genes (Fig. 1C). This result is consistent with
previous ﬁndings that TFs can inﬂuence chromatin organization by
competing with nucleosomes for accessing to DNA (Komili and
Silver, 2008; Segal and Widom, 2009).
Fig. 2. Dynamic characteristics (DC) of nucleosome organization. (A) DC
of nucleosome organization in the promoter regions. They were measured
by nucleosome positioning before and after heat shock (DC1); nucleosome
positioning between different cross-platform datasets (DC2) and nucleosome
occupancy among cells grown at different conditions (DC3), respectively. (B)
Nucleosome fuzziness relative to TSS. Fuzziness is reported as the standard
deviation of nucleosome locations for each individual reference nucleosome.
The distribution of nucleosome fuzziness is plotted by binning nucleosomes
together moving along the genes. The inset ﬁgure shows the fraction of
genes with linker region longer than 50 bp in the promoter region (500 bp
upstream to 1000 bp downstream to TSS). All error bars were calculated by
1000 bootstrapping.
3.3 Dynamic characteristics of nucleosome
organization between CR- and TF-sensitive genes
In order to fully describe the extents and patterns of CRs and TFs
in driving dynamic characteristics of nucleosome positioning, we
next show it from different lines of evidences. (i) We mapped
and compared the locations of nucleosomes measured under the
normal and heat-shock conditions (Shivaswamy et al., 2008). By
comparing nucleosome positions within 45 bp of their loci in both
two conditions, the results indicate that ∼66% of nucleosomes
retain their nucleosome patterns after heat shock. Nucleosomes
with dramatic positional changes are more highly enriched in CRsensitive
genes than in TF-sensitive genes (Fig. 2A). (ii) With
more and more genome-wide maps of nucleosome position being
available, we compiled two genome-wide maps of nucleosome
positions with high resolutions in the same strain (BY4714) and
compared their nucleosome locations (see Section 2, Supplementary
Table S3). After mapping the inconsistent nucleosome positions,
the result shows that CR-sensitive genes have a more dynamic
nucleosome pattern than TF-sensitive genes (Fig. 2A). (iii) We
compared the discrepancies of nucleosome occupancy derived from
three different growth conditions, and also found that CRsensitive
genes are likely to have dynamic nucleosome characteristics. These
results suggested that the inﬂuence of CRE was markedly higher
150
atMasarykUniversityonFebruary21,2011bioinformatics.oxfordjournals.orgDownloadedfrom
[13:22 16/12/2010 Bioinformatics-btq637.tex] Page: 151 147–152
Differential effects of CRs and TFs on gene regulation
than that of TRE on the dynamic characteristics of nucleosome
organization.
Most nucleosomes are ‘well-positioned’ at the same location in
the cells, whereas some of them tend to spread out their positions
and show a delocalized status, and they have been deﬁned as ‘fuzzily
positioned’ nucleosomes (Mavrich et al., 2008). To evaluate the
extent of nucleosomes having different binding locations between
CR- and TF-sensitive genes, we calculated the percentage of DNA
sequences occupied by fuzzily positioned nucleosomes (Lee et al.,
2007) within the promoter regions. Figure 2B (inset ﬁgure) shows
that it was signiﬁcantly higher in CR-sensitive genes than in TFsensitive
genes (Wilcoxon rank sum test, P = 4.3e-5), which suggests
that CRs are more responsible for the nucleosome dynamic patterns
than TFs. Recently, Mavrich et al. (2008) measured ‘fuzziness score’
using standard deviation of all number of reads of nucleosome
location. To further validate our result, we plotted the ‘fuzziness
score’ within the focal genes. The result revealed that nucleosome
fuzziness increased from the promoters to the 3 end of the genes in
both CR-and TF-sensitive genes (Fig. 2B), and nucleosomes located
in CR-sensitive genes show signiﬁcantly higher level of fuzziness
(Wilcoxon rank sum test, P = 1.7e-47).
3.4 Histone modiﬁcations induced by CR and TF
regulation
Histone proteins are subject to numerous modiﬁcations that function
as important transcriptional regulatory mechanisms (Li et al., 2007).
To detect the histone activities induced by CR and TF regulation,
we compiled the histone tails perturbation data and examined the
general relationship between transcriptional regulation by TFs and
CRs and histone activities. We analyzed genome-wide expression
data of yeast strains after deleting the H3 (H3 1-28) and H4
(H4 2-26) amino termini (Sabet et al., 2003), and found that
genes affected in the absence of the amino termini of histone
H3 and H4 were extremely regulated by TFs and CRs. These
genes might be highly regulated by histone tails through activating
or repressing gene expression. Moreover, genes de-repressed by
the deletion of histone tails are dramatically sensitive to CRs
(showing extremely the right-side tails, Fig. 3A), which indicated
that these genes require histone amino termini for proper regulation,
particularly repression. For example, genes regulated by histone
modiﬁcation which is associated with repressed transcription will
show upregulated expression after histone depletion (right side tails
in Fig. 3), and the expression of these genes should be highly
depended on CR to activate gene expression. This result is also in
line with previous ﬁndings that chromatin remodeler ISW2 interacts
with histone amino acid termini and represses the expression of
many genes, especially genes involved in meiosis (Georgel et al.,
1997; Goldmark et al., 2000). This scenario suggested that histone
H3 and H4 are regulated by both TFs and CRs, and genes requiring
histone tails depend on speciﬁc effects of chromatin.
We also examined the histone H3 turnover rate (Dion et al., 2007)
and histone-variant H2A.Z occupancy (Albert et al., 2007) within
these two cohorts of genes. As expected, histone-variant H2A.Z is
highly enriched in the cohort of TF-sensitive genes (60%, relative
to CR-sensitive genes 48%, P=0.002, chi-square test), which is
in agreement with the ﬁndings that H2A.Z is a general feature
of promoter chromatin architecture characterized by nucleosome
depleted region (Tirosh and Barkai, 2008).In contrast, higher histone
Fig. 3. Impact of CR and TF regulation effect on the activity of histones.
(A) CR and (B) TF regulation effect with the varying sensitivity to histone
regulation. Genes were ordered by expression changes resulting from histone
H3 (H3 1-28) and H4 (H4 2-26) amino terminus depletion. Both the
average values of CR and TF regulation effect were obtained by a sliding
window of 200 ordered genes.
turnover rates were observed in the promoters of CR-sensitive
genes (0.29) than that in TF-sensitive genes (0.21) (P=0.03, 1000
permutation test).
4 DISCUSSION AND CONCLUSION
Recent works elucidating nucleosome positioning in yeast have
revealed that the nucleosome organization at the promoter region
is characterized by a nucleosome depleted region located at
∼100–200 bp upstream of TSS (Lee et al., 2007; Yuan et al., 2005).
This depleted region can be occupied by nucleosomes to adjust
expression levels under particular environmental perturbations,
whereas promoters of active genes are depleted of nucleosomes
to permit TF-binding events required for gene expression activity.
Here, we showed that nucleosome organization has distinct patterns
in the promoter region under the regulation of CRs and TFs. Notably,
our result revealed that depleted nucleosome occupancy does not
always cause higher gene expression level in the cohort of TFsensitive
genes. This phenomenon could be largely explained by
the competition of TFs with nucleosomes to access speciﬁc genomic
location. This competition might result in reduced binding efﬁciency
of the cognate TF, and thus tune the gene expression to adapt the
varying environmental conditions. It was documented that CRs are
capable of remodeling chromatin structures through nucleosome
movement and eviction in vivo (Li et al., 2007).
Previous studies suggested that an ATP-dependent chromatin
remodeling enzyme, the SWI/SNF complex, is involved in
nucleosome eviction (Narlikar et al., 2002; Saha et al., 2006), and
the Iswi2 enzyme is involved in shifting nucleosomes to a position
with an energetically unfavorable DNA sequence (Whitehouse and
Tsukiyama, 2006). These works support our ﬁnding that dynamic
characteristics of nucleosome positioning are more likely to be
mediated by CRs than TFs.
Taken together, we performed a comprehensive analysis to detect
the differences between CR and TF regulation effects on gene
regulation, and found the two different underlying patterns of
nucleosome organization at nucleosome-free region. It suggested
that CR-sensitive genes tend to have dynamic characteristics of
151
atMasarykUniversityonFebruary21,2011bioinformatics.oxfordjournals.orgDownloadedfrom
[13:22 16/12/2010 Bioinformatics-btq637.tex] Page: 152 147–152
D.Dong et al.
nucleosomes than TF-sensitive genes. In spite of the different
mechanism, however, CR and TF by no means work independently,
and they are likely to work jointly during the transcription process.
Nevertheless, our results provide a better understanding of how TRE
and CRE determine nucleosome positioning and how they affect
transcription process.
ACKNOWLEDGEMENTS
We thank Dr Martin Kupiec for providing us the chromatin modiﬁer
perturbation compendiums. We would like to thank the Associate
Editor and three referees for their constructive comments that have
signiﬁcantly improved the quality of this article.
Funding: Team Grant from the Canadian Institutes of Health
Research (CIHR MOP#82940); Discovery Grant from Natural
Sciences and Engineering Research Council of Canada (NSERC).
Conﬂict of Interest: none declared.
REFERENCES
Albert,I. et al. (2007) Translational and rotational settings of H2A.Z nucleosomes across
the Saccharomyces cerevisiae genome. Nature, 446, 572–576.
Amit,I. et al. (2007) A module of negative feedback regulators deﬁnes growth factor
signaling. Nat. Genet., 39, 503–512.
Badis,G. et al. (2008) A library of yeast transcription factor motifs reveals a widespread
function for Rsc3 in targeting nucleosome exclusion at promoters. Mol. Cell, 32,
878–887.
Basehoar,A.D. et al. (2004) Identiﬁcation and distinct regulation of yeast TATA boxcontaining
genes. Cell, 116, 699–709.
Berger,S.L. (2000) Gene regulation. Local or global? Nature, 408, 412–413, 415.
Buck,M.J. and Lieb,J.D. (2006) A chromatin-mediated mechanism for speciﬁcation of
conditional transcription factor targets. Nat. Genet., 38, 1446–1451.
Choi,J.K. and Kim,Y.J. (2008) Epigenetic regulation and the variability of gene
expression. Nat. Genet., 40, 141–147.
Choi,J.K. and Kim,Y.J. (2009) Intrinsic variability of gene expression encoded in
nucleosome positioning sequences. Nat. Genet., 41, 498–503.
Dion,M.F. et al. (2007) Dynamics of replication-independent histone turnover in
budding yeast. Science, 315, 1405–1408.
Field,Y. et al. (2008) Distinct modes of regulation by chromatin encoded through
nucleosome positioning signals. Plos Comput. Biol., 4.
Field,Y. et al. (2009) Gene expression divergence in yeast is coupled to evolution of
DNA-encoded nucleosome organization. Nat. Genet., 41, 438–445.
Georgel,P.T. et al. (1997) Role of histone tails in nucleosome remodeling by Drosophila
NURF. EMBO J., 16, 4717–4726.
Goldmark,J.P. et al. (2000) The Isw2 chromatin remodeling complex represses early
meiotic genes upon recruitment by Ume6p. Cell, 103, 423–433.
Harbison,C.T. et al. (2004) Transcriptional regulatory code of a eukaryotic genome.
Nature, 431, 99–104.
Hartley,P.D. and Madhani,H.D. (2009) Mechanisms that specify promoter nucleosome
location and identity. Cell, 137, 445–458.
Holstege,F.C. et al. (1998) Dissecting the regulatory circuitry of a eukaryotic genome.
Cell, 95, 717–728.
Hu,Z. et al. (2007) Genetic reconstruction of a functional transcriptional regulatory
network. Nat. Genet., 39, 683–687.
Ihmels,J. et al. (2002) Revealing modular organization in the yeast transcriptional
network. Nat. Genet., 31, 370–377.
Ito,T. et al. (1997) ACF, an ISWI-containing and ATP-utilizing chromatin assembly and
remodeling factor. Cell, 90, 145–155.
Jiang,C. and Pugh,B.F. (2009) A compiled and systematic reference map of nucleosome
positions across the Saccharomyces cerevisiae genome. Genome Biol., 10, R109.
Kaplan,N. et al. (2009) The DNA-encoded nucleosome organization of a eukaryotic
genome. Nature, 458, 362–U129.
Kim,H.D. and O’Shea,E.K. (2008)Aquantitative model of transcription factor-activated
gene expression. Nat. Struct. Mol. Biol., 15, 1192–1198.
Komili,S. and Silver,P.A. (2008) Coupling and coordination in gene expression
processes: a systems biology view. Nat. Rev. Genet., 9, 38–48.
Lam,F.H. et al. (2008) Chromatin decouples promoter threshold from dynamic range.
Nature, 453, 246–U216.
Langst,G. et al. (1999) Nucleosome movement by CHRAC and ISWI without disruption
or trans-displacement of the histone octamer. Cell, 97, 843–852.
Lee,S.I. et al. (2006) Identifying regulatory mechanisms using individual variation
reveals key role for chromatin modiﬁcation. Proc. Natl Acad. Sci. USA, 103,
14062–14067.
Lee,W. et al. (2007) A high- resolution atlas of nucleosome occupancy in yeast. Nat.
Genet., 39, 1235–1244.
Li,B. et al. (2007) The role of chromatin during transcription. Cell, 128, 707–719.
MacIsaac,K.D. et al. (2006) An improved map of conserved regulatory sites for
Saccharomyces cerevisiae. BMC Bioinformatics, 7, 113.
Mavrich,T.N. et al. (2008) A barrier nucleosome model for statistical positioning of
nucleosomes throughout the yeast genome. Genome Res., 18, 1073–1083.
Narlikar,G.J. et al. (2002) Cooperation between complexes that regulate chromatin
structure and transcription. Cell, 108, 475–487.
Orphanides,G. and Reinberg,D. (2002) A uniﬁed theory of gene expression. Cell, 108,
439–451.
Pugh,B.F. (2000) Control of gene expression through regulation of the TATA-binding
protein. Gene, 255, 1–14.
Rando,O.J. and Chang,H.Y. (2009) Genome-wide views of chromatin structure. Ann.
Rev. Biochem., 78, 245–271.
Sabet,N. et al. (2003) Global and speciﬁc transcriptional repression by the histone H3
amino terminus in yeast. Proc. Natl Acad. Sci. USA, 100, 4084–4089.
Saha,A. et al. (2006) Chromatin remodelling: the industrial revolution of DNA around
histones. Nat. Rev., 7, 437–447.
Segal,E. and Widom,J. (2009) What controls nucleosome positions? Trends Genet., 25,
335–343.
Segal,E. et al. (2003) Module networks: identifying regulatory modules and their
condition-speciﬁc regulators from gene expression data. Nat. Genet., 34, 166–176.
Shivaswamy,S. et al. (2008) Dynamic remodeling of individual nucleosomes across a
eukaryotic genome in response to transcriptional perturbation. PLoS Biol., 6, e65.
Stein,A. et al. (2009)Are nucleosome positions in vivo primarily determined by histoneDNA
sequence preferences? Nucleic Acids Res., 38, 709—719.
Steinfeld,I. et al. (2007) A genome-wide analysis in Saccharomyces cerevisiae
demonstrates the inﬂuence of chromatin modiﬁers on transcription. Nat. Genet.,
39, 303–309.
Tirosh,I. and Barkai,N. (2008) Two strategies for gene regulation by promoter
nucleosomes. Genome Res., 18, 1084–1091.
Whitehouse,I. and Tsukiyama,T. (2006) Antagonistic forces that position nucleosomes
in vivo. Nat. Struct Mol. Biol., 13, 633–640.
Whitehouse,I. et al. (2007) Chromatin remodelling at promoters suppresses antisense
transcription. Nature, 450, 1031–U1033.
Yarragudi,A. et al. (2004) Comparison of ABF1 and RAP1 in chromatin opening and
transactivator potentiation in the budding yeast Saccharomyces cerevisiae. Mol.
Cell. Biol., 24, 9152–9164.
Yuan,G.C. et al. (2005) Genome-scale identiﬁcation of nucleosome positions in
S.cerevisiae. Science, 309, 626–630.
Yuan,G.C. et al. (2006) Statistical assessment of the global regulatory role of histone
acetylation in Saccharomyces cerevisiae. Genome Biol., 7, R70.
Zhang,Y. et al. (2009) Intrinsic histone-DNA interactions are not the major determinant
of nucleosome positions in vivo. Nat. Struct. Mol. Biol., 16, 847–852.
152
atMasarykUniversityonFebruary21,2011bioinformatics.oxfordjournals.orgDownloadedfrom