The role of nucleosome positioning in the evolution of gene regulation.

The Role of Nucleosome Positioning in the Evolution of
Gene Regulation
Alexander M. Tsankov1,2, Dawn Anne Thompson1, Amanda Socha1, Aviv Regev1,3,4*., Oliver J. Rando5*.
1 Broad Institute of MIT and Harvard, Cambridge, Massachusetts, United States of America, 2Department of Electrical Engineering and Computer Science, Massachusetts
Institute of Technology, Cambridge, Massachusetts, United States of America, 3Department of Biology, Massachusetts Institute of Technology, Cambridge, Massachusetts,
United States of America, 4Howard Hughes Medical Institute, Cambridge, Massachusetts, United States of America, 5Department of Biochemistry and Molecular
Pharmacology, University of Massachusetts Medical School, Worcester, Massachusetts, United States of America
Abstract
Chromatin organization plays a major role in gene regulation and can affect the function and evolution of new
transcriptional programs. However, it can be difficult to decipher the basis of changes in chromatin organization and their
functional effect on gene expression. Here, we present a large-scale comparative genomic analysis of the relationship
between chromatin organization and gene expression, by measuring mRNA abundance and nucleosome positions
genome-wide in 12 Hemiascomycota yeast species. We found substantial conservation of global and functional chromatin
organization in all species, including prominent nucleosome-free regions (NFRs) at gene promoters, and distinct chromatin
architecture in growth and stress genes. Chromatin organization has also substantially diverged in both global quantitative
features, such as spacing between adjacent nucleosomes, and in functional groups of genes. Expression levels, intrinsic anti-
nucleosomal sequences, and trans-acting chromatin modifiers all play important, complementary, and evolvable roles in
determining NFRs. We identify five mechanisms that couple chromatin organization to evolution of gene regulation and
have contributed to the evolution of respiro-fermentation and other key systems, including (1) compensatory evolution of
alternative modifiers associated with conserved chromatin organization, (2) a gradual transition from constitutive to trans-
regulated NFRs, (3) a loss of intrinsic anti-nucleosomal sequences accompanying changes in chromatin organization and
gene expression, (4) re-positioning of motifs from NFRs to nucleosome-occluded regions, and (5) the expanded use of NFRs
by paralogous activator-repressor pairs. Our study sheds light on the molecular basis of chromatin organization, and on the
role of chromatin organization in the evolution of gene regulation.
Citation: Tsankov AM, Thompson DA, Socha A, Regev A, Rando OJ (2010) The Role of Nucleosome Positioning in the Evolution of Gene Regulation. PLoS Biol 8(7):
e1000414. doi:10.1371/journal.pbio.1000414
Academic Editor: Peter B. Becker, Adolf Butenandt Institute, Germany
Received February 18, 2010; Accepted May 27, 2010; Published July 6, 2010
Copyright: � 2010 Tsankov et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: AT was supported by the National Science Foundation Graduate Research Fellowship. Work was supported by the Human Frontiers Science Program,
the Howard Hughes Medical Institute, a Career Award at the Scientific Interface from the Burroughs Wellcome Fund, a National Institutes of Health PIONEER
award, the Broad Institute, a Sloan Fellowship, National Cancer Institute grant R01 CA119176-01 (AR), by NIGMS grant GM079205, HFSP, and the Burroughs
Wellcome Fund (OJR). The funders had no rule in study design, data collection and analysis, or preparation of the manuscript.
Competing Interests: The authors have declared that no competing interests exist.
Abbreviations: BYA, billion years ago; cRPs, cytoplasmic ribosomal proteins; GRFs, general regulatory factors; K-S, Kolmogorov-Smirnov; LCA, last common
ancestor; mRP, mitochondrial ribosomal protein; MYA, million years ago; NFRs, nucleosome free regions; PSSM, position specific scoring matrix; TF, transcription
factor; TSSs, transcription start sites; UTR, untranslated region
* E-mail: Oliver.Rando@umassmed.edu (OJR); aregev@broad.mit.edu (AR)
. These authors contributed equally to this work.
Introduction
Regulatory differences affecting gene expression can play a
major role in species evolution [1] and can help elucidate the
functional mechanisms that control gene regulation [2,3].
Although specific examples of regulatory divergence are known
in bacteria [4], fungi [5,6,7,8], flies [9], and mammals [10], a
general understanding of the evolution of gene regulation is still
lacking. The recent availability of many sequenced genomes and
accessibility of genomic profiling approaches open the way for
comparisons of gene regulation across multiple species.
Among eukaryotes, the Hemiascomycota yeasts (Figure 1A), which
span over,250 million years of evolution, are particularly suitable
for studying evolution of gene regulation. This is due to the genetic
tractability of yeasts, the wealth of knowledge about the model
organism Saccharomyces cerevisiae, the large number of sequenced
genomes, and the diversity of yeast lifestyles [3]. Notably,
Hemiascomycota yeasts diverged before and after a whole genome
duplication event (WGD, Figure 1A) [11], which marked a shift
from using respiration for energy production in pre-WGD species
to primarily using fermentation in post-WGD species [12].
Nucleosomes modulate eukaryotic gene regulation by affecting
the accessibility of other proteins to the DNA, which can impact
gene activation and repression [13]. In particular, many genes
have nucleosome-depleted ‘‘Nucleosome Free Regions’’ (NFRs) in
their proximal promoters (Figure 1B, top), providing access to
sequence specific transcription factors (TFs) and to the basal
transcription machinery [14,15,16,17]. Three major determinants
have been proposed to impact nucleosome depletion at NFRs: (1)
active transcription by RNA polymerase II results in eviction of
the 21 nucleosome [18,19], (2) intrinsic ‘‘anti-nucleosomal’’ DNA
sequences such as Poly(dA:dT) bind histones with low affinity and
PLoS Biology | www.plosbiology.org 1 July 2010 | Volume 8 | Issue 7 | e1000414

can ‘‘program’’ NFRs constitutively [20,21,22,23,24], and (3) trans-
acting proteins can move nucleosomes away from their thermo-
dynamically preferred locations [25,26].
Recent studies in yeast suggest a broad role for chromatin
organization in regulatory evolution. Most regulatory divergence
between closely related S. cerevisiae strains is associated with
divergence in unlinked (trans) chromatin remodelers [27,28].
Conversely, many transcriptional differences between S. cerevisiae
and S. paradoxus (Last Common Ancestor (LCA) ,2 million years
ago (MYA)) are due to linked cis polymorphisms predicted to affect
nucleosome occupancy [29,30]. Furthermore, a recent study
suggested that changes in the regulation of mitochondrial
ribosomal protein (mRP) genes between the distant species C.
albicans and S. cerevisiae (LCA ,200 MYA) were associated with a
change in nucleosome organization [31,32]. In particular, the
higher expression of mitochondrial genes in respiratory C. albicans
is accompanied by enrichment for the PolyA-like ‘‘RGE’’ binding
site in the mRP gene promoters [31], which appears to ‘‘program’’
the constitutive presence of wider, more open NFRs at these genes
[32]. All of these are absent from the promoters of mRPs in the
fermentative S. cerevisiae. Finally, a recent study [33] compared
genome-wide nucleosome positioning in S. cerevisiae and S. pombe
(LCA ,300M–1 BYa), finding changes in global nucleosome
spacing and in the apparent sequences that intrinsically contribute
to nucleosome positioning in vivo.
While these examples are intriguing, they are limited in their
phylogenetic coverage (a pair of species) or their functional scope
(one regulon). Thus, we understand little about the evolutionary
interplay between gene expression, regulatory sequence elements,
and chromatin organization. How does chromatin organization
change over evolutionary time scales? Are the mechanisms
underlying chromatin packaging of functional gene modules
conserved? If not, how do they evolve and what is the role of
different factors in this divergence? Are changes in chromatin
organization related to changes in gene regulation? Can
phylogenetic comparisons shed light on the distinct mechanisms
that help establish chromatin organization?
Here, we present the first large-scale experimental and
computational study of chromatin organization across a eukaryotic
phylogeny. We measured genome-wide nucleosome locations and
mRNA abundance in 12 Hemiascomycota yeast species, spanning
over 250 million years of evolution (Figure 1A). We developed an
analysis framework that integrates the experimental data with
genome sequences, functional gene sets, and TF binding sites
across the 12 species.
Our analysis uncovers several major principles that govern the
evolutionary and functional relationship between chromatin
organization and gene regulation in this phylogeny. (1) While
qualitative features of chromatin organization are conserved in all
species, quantitative features such as nucleosome packing, NFR
length, and NFR to ATG distance have substantially diverged; (2)
promoter chromatin organization and gene expression levels of
‘‘growth’’ and ‘‘stress’’ genes follow distinct patterns, and this
dichotomy is conserved in all species; (3) evolutionary divergence
in gene expression is often accompanied by transition of chromatin
organization from a ‘‘growth’’ to a ‘‘stress’’ pattern; (4) changes in
transcription levels, gain/loss of anti-nucleosomal sequences, and
gain/loss of binding sites for ‘‘general regulatory factors’’ (GRFs)
all play substantial and complementary roles in divergence of
chromatin organization; (5) the loss of anti-nucleosomal sequences
and parallel gain of binding sites for GRFs drive shifts from
intrinsic to trans-regulated chromatin organization; (6) regulatory
divergence can also occur by re-positioning of binding sites relative
to nucleosome positions or by expanding the use of accessible sites
by paralogous TFs. These mechanisms played a role in the
evolution of respiro-fermentation, as well as in the evolution of
regulation of other key regulons at different phylogenetic points,
including mating, meiosis, RNA polymerase subunits, proteaso-
mal, and splicing genes. Together, they uncover novel insights into
the general roles for chromatin in regulating genomic access and
in the evolution of regulatory programs, and provide a rich
resource for future investigation.
Results
A Chromatin Map for 12 Hemiascomycota Species
We mapped nucleosome positions genome-wide in 12 Hemi-
ascomycota species (Figure 1A) [34] by Illumina sequencing of
mononucleosomal DNA [19,21,35] isolated from mid-log cultures
(Materials and Methods, Figures 1A and S1). To minimize
condition- and stress-related differences, we grew all species in the
same rich medium, where the growth rate of each species was at
least ,80% of its maximal measured rate in any of over 40 tested
media formulations. In order to compare our data to transcrip-
tional output, we also used species-specific microarrays to measure
mRNA abundance in all species in the same mid-log cultures used
for nucleosome mapping (Table S2, Materials and Methods).
Aligning nucleosome reads to each genome and averaging over
all genes showed remarkably similar profiles in all species studied
(Figures 1A, S2, S3). All gene-averaged profiles are dominated by
a pronounced depression upstream of the ATG that corresponds
to the NFR [14,15,16,17,36]. To quantitatively compare chro-
matin structure between various genes, we first called nucleosome
positions, identified 59 and 39 NFRs, and measured a number of
nonredundant features that describe the chromatin organization at
each gene (Materials and Methods, Figures 1B and S4). Below, we
will study each feature at three levels: (1) globally, averaged across
all genes in a genome; (2) functionally, averaged across all genes in a
functional category; and (3) locally, at a single gene.
Packaging of Coding Regions Is Qualitatively Conserved,
but Quantitative Features Such as Nucleosome Spacing
and NFR Width Have Diverged between Species
Several qualitative chromatin features have previously been
identified in all eukaryotes studied [14], and these are conserved
Author Summary
Divergence in gene regulation plays a major role in
organismal evolution. Evidence suggests that changes in
the packaging of eukaryotic genomes into chromatin can
underlie the evolution of divergent gene expression
patterns. Here, we explore the role of chromatin structure
in regulatory evolution by whole-genome measurements
of nucleosome positions and mRNA levels in 12 yeast
species spanning ,250 million years of evolution. We find
several distinct ways in which changes in chromatin
structure are associated with changes in gene expression.
These include changes in promoter accessibility, changes
in promoter chromatin architecture, and changes in the
accessibility of specific transcription factor binding sites. In
many cases, changes in chromatin architecture are
coupled to physiological diversity, including the evolution
of a respiration- or fermentation-based lifestyle, mating
behavior, salt tolerance, and broad aspects of genomic
structure. Together, our data will provide a rich resource
for future investigations into the interplay between
chromatin structure, gene regulation, and evolution.
Fungal Nucleosome Positioning
PLoS Biology | www.plosbiology.org 2 July 2010 | Volume 8 | Issue 7 | e1000414

across all 12 species (Figures 1A, S2, and S3). These include an
abundant 59NFR, a common 39NFR, a well-positioned +1
nucleosome (Nuc+1), and increasing nucleosome fuzziness over
the body of genes (Figures S2 and S3, Table S3), which is
consistent with statistical positioning of nucleosomes [23,37,38].
In contrast, quantitative global features were often variable
between species (Figures 1C–F and S5, Table S3). Our measure-
ments recapitulated previous predictions or bulk assays in the few
cases where these were available, thus validating our dataset and
analytical methods. For example, nucleosome spacing in coding
regions was variable between species (Figure 1C,D), consistent
with observed nucleosome laddering on gels [39,40]. This leads to
variation in the specific coding sequences exposed in linker DNA
and could affect patterns of sequence variation [41,42,43] and
higher-order packaging into the 30 nm fiber [44]. The distance
between the NFR and a gene’s start codon (Figures 1E,F and S5) is
also variable between species, consistent with prior computational
predictions [45].
Other evolutionary variations in global features were not
previously described, showing that additional major aspects of
chromatin architecture can substantially diverge. Most notably,
the median NFR width was highly variable between species (Table
S3), ranging from 109 to 155 nucleotides. This likely reflects the
variation in the length and abundance of anti-nucleosomal
Poly(dA:dT) tracts between species (discussed below). Shorter
NFRs may constrain regulatory information into more compact
promoters.
A Conserved Dichotomy in Chromatin Organization of
‘‘Stress’’ and ‘‘Growth’’ Genes
We next explored possible functional implications of chromatin
organization in specific sets of genes with related function. Prior
studies in S. cerevisiae and C. albicans have shown that in both
species, ‘‘growth’’ genes, defined by their co-expression with
cytoplasmic ribosomal proteins (cRPs), have a more open
chromatin organization on average [32]. Conversely, ‘‘stress’’
Figure 1. Global chromatin organization in 12 Hemiascomycota fungi. (A) Phylogeny of species included in this study (adapted from [34]). Right,
gene-averaged nucleosome sequencing data from 4 of the 12 species, aligned by Nuc+1. (Data for all species are in Figure S2.) (B) Chromatin features.
Shown is a schematic of a gene (green box), its promoter (black line) and associated nucleosomes (yellow), along with nucleosome sequencing data
(dark blue curve), and several extracted features, as indicated. (C) Global variation between species in nucleosome spacing in coding regions. Shown are
the median nucleosome-to-nucleosome distances over coding regions, averaged over all genes in each species. Values are arranged from low to high
rather than by phylogeny to emphasize the range of variability. Species names are colored by their relation to WGD as in (A). (D) Spacing differences
between two Kluyveromyces species. Shown are 59 NFR-aligned averaged data for K. lactis (red) and K. waltii (blue), showing differences in coding region
spacing. (E) Global variation in NFR to ATG distance (D59NFR-ATG). Shown are median distances from the 59 NFR to start codon for all genes in each species,
sorted from low to high values. (F) Distribution of NFR to ATG distances (D59NFR-ATG) in S. kluyverii (blue) and C. glabrata (red).
doi:10.1371/journal.pbio.1000414.g001
Fungal Nucleosome Positioning
PLoS Biology | www.plosbiology.org 3 July 2010 | Volume 8 | Issue 7 | e1000414

genes, whose expression is anti-correlated to that of growth genes,
have a more closed chromatin organization in both species.
To assess the generality of this observation, and identify
additional trends, we tested in each species thousands of functional
gene sets for enrichment of each of 22 distinct chromatin
parameters. We used gene orthology [34] to project functional
gene sets defined in S. cerevisiae across species (Materials and
Methods). For a given gene set in each species we calculated
whether its constituent genes tended to have high or low values of
each of the chromatin features (Figure 1B), relative to the
background of that feature’s overall distribution in that species
(Kolmogorov-Smirnov (K-S) test, Figure 2A,B). This provides a
comprehensive overview of chromatin organization at 59 promo-
ters and 39 ends for each functional gene set across the 12 species
(Figure 2C–J, middle panels, Figures S6 and S7, and Tables S4–
S5). In order to compare chromatin changes to gene expression
levels, we also calculated the enrichment for high or low mRNA
expression in all gene sets for each species (K-S test, Figure 2C–J,
left panels).
We confirm a strong dichotomy in the promoter chromatin
architecture of most ‘‘stress’’ and ‘‘growth’’ genes in S. cerevisiae
[19,46,47,48,49] and C. albicans [32] and find that it is conserved
across all 12 species (Figures 2C,D and S6–S8). Promoters of
‘‘growth’’ genes (e.g., ribosomal, proteasomal, and nuclear pore
proteins, Figure 2C,E,G) exhibit long and deep (low occupancy)
59NFRs. Conversely, those of ‘‘stress’’ genes (e.g., toxin-response
genes, integral membrane proteins, Figure 2D) exhibit a more
variable chromatin architecture, with shallower (higher occu-
pancy) and narrower 59NFRs. A host of other chromatin features
also distinguish between the two functional groups (Figure S6).
Thus, the separation of the ‘‘growth’’ and ‘‘stress’’ axes is a
hallmark of Hemiascomycota gene regulation [2,3] and imposes
strong constraints at all levels from evolution of gene content [34]
to chromatin organization. There are, however, several exceptions
to this rule. Most notably, several key ‘‘growth’’ genes, including
glycolysis genes and endoplasmic reticulum genes, are highly
expressed, yet do not exhibit deep NFRs in any species (Figure 2F).
We identify a range of additional conserved patterns of
chromatin architecture associated with other specific functions,
which were not previously reported. For example, a number of
gene sets (e.g., reproduction, cell wall, inositol phosphate,
benzoate, and nicotinamide metabolism genes) have conserved
long NFR to ATG distances (Figure S6), but have few other
hallmarks of stress genes, and are expressed at average levels. In S.
cerevisiae, these genes have long 59 untranslated regions (59UTRs)
[50], suggesting that relatively long 59UTRs are conserved at their
orthologs in all 12 species. This may indicate a conserved role for
translational control in the regulation of these functions [51].
Coherent Changes in Chromatin Organization
Accompanied the Evolutionary Divergence of Gene
Regulation in Mitochondrial, Splicing, and Cytoskeleton
Genes
On this backdrop of conservation, we find that coordinated
changes have occurred in chromatin organization of specific
functional gene sets, consistent with major phenotypic changes.
Most notably, respiration and mitochondrial genes have switched
from a ‘‘growth’’-like chromatin pattern in pre-WGD species
(where they are highly expressed) to a more ‘‘stress’’-like pattern
post-WGD (Figures 2H and S6). We confirm the previously
reported change between S. cerevisiae and C. albicans for genes
involved in respiratory metabolism [32]. We further extend these
results across the full phylogenetic scope and to several other gene
sets of related function (Figures 2H and S6). This change
corresponds to a major change in lifestyle from respiration to
respiro-fermentation after the WGD [12,31,32,52]. We also
discover the converse evolutionary pattern (Figure 2I): a number
of gene sets involved in cytoskeletal organization are packaged into
deeper NFRs in post-WGD species than in pre-WGD species.
Surprisingly, the expression level of these genes has not
substantially changed with this transition.
Changes in chromatin organization have also occurred at other
phylogenetic points of phenotypic evolution, suggesting a general
evolutionary mechanism. For example, we discovered that in
Yarrowia lipolytica spliceosome genes are associated with long and
deep NFRs, but in all other species they are enriched for short and
shallow NFRs (Figure 2J, middle panel). This switch between deep
and shallow NFRs is accompanied by a decrease in expression of
these genes (Figure 2J, left panel) and is consistent with the much
larger number of introns in Yarrowia lipolytica genes [53] and with
the loss of introns and reduction of splicing in the subsequently
diverged species.
Differences in Expression and Intrinsic Anti-Nucleosomal
Sequences Only Account for Some of the Changes in
Chromatin Organization Within and Between Species
We next asked what mechanisms contribute to conservation and
variation in chromatin organization across species. Three
determinants have been previously implicated in establishing
NFRs in S. cerevisiae [14]: (1) the expression level of the gene, as
RNA polymerase recruitment affects NFR width; (2) the presence
of intrinsic anti-nucleosomal sequences such as Poly(dA:dT) tracts
in the gene’s promoter; and (3) the binding of proteins such as
chromatin remodelers that actively evict or move nucleosomes.
We first consider these three determinants independently, and
then assess their relative contributions.
In some cases, variation in chromatin organization in a gene set,
both within and between species, correlates with gene expression
level. Within each species, many highly expressed ‘‘growth’’ genes
(e.g., RP genes) are packaged with wide and deep NFRs, while
many poorly expressed stress genes have shorter, occupied NFRs
(Figures 2C,D, S6). Between species, evolutionary shifts from high
to low expression levels were sometimes accompanied by
corresponding changes in chromatin organization (e.g., mitochon-
drial RP and splicing genes, Figure 2H,J).
However, transcription level is insufficient to solely explain the
NFR occupancy measured across the 12 species. Globally,
expression level alone explains only 1.7%–13.1% of the variation
in NFR occupancy in each of the 12 species (Lowess fit, Figure
S9A,C,E, Materials and Methods). Furthermore, when we use
Lowess subtraction to correct for the relationship between mRNA
level and each chromatin feature, the enrichments of most gene
sets for high or low values of chromatin features were maintained
(Figure S10, Materials and Methods). Within species, the
discrepancy is prominent in some of the gene sets (e.g., glycolysis,
gluconeogenesis) that are highly expressed in all species but do not
exhibit the expected deep NFRs (Figure 2F). Between species,
cytoskeleton and nuclease-related gene sets have shifted from
shallow to deep NFRs at the WGD, often without a concomittant
change in expression levels (Figure 2I). The failure of transcript
levels to fully explain NFR width and depth is consistent with
recent experimental results in S. cerevisiae, where the distinctive
chromatin organization of growth and stress genes was largely
maintained even after genetically inactivating RNA Pol II [19].
We next tested an alternative hypothesis that chromatin
organization at the NFR is determined by intrinsic ‘‘anti-
nucleosomal’’ sequences with low affinity for the histone octamer,
Fungal Nucleosome Positioning
PLoS Biology | www.plosbiology.org 4 July 2010 | Volume 8 | Issue 7 | e1000414

Figure 2. Conservation and variation in chromatin structure of functional gene sets. (A,B). Strategy for associating chromatin features with
gene sets. (A) Shown is Nuc+1-aligned nucleosome data for all genes (blue) and ribosomal protein genes (red) in S. cerevisiae, demonstrating that
ribosomal protein genes are associated with wider NFRs. (B) Cumulative distribution plot of NFR occupancy in all genes (blue) versus ribosomal
protein genes (red). y-axis shows fraction of promoters with NFR occupancy below a given value, with NFR occupancy values on the x-axis. Wide
separation between curves (light blue vertical line) is captured by a significant K-S statistic, indicating that ribosomal genes have significantly low
occupancy, or ‘‘deep’’ NFRs. K-S P values are converted to color scale (right panel): blue, significantly low feature values; yellow, significantly high
Fungal Nucleosome Positioning
PLoS Biology | www.plosbiology.org 5 July 2010 | Volume 8 | Issue 7 | e1000414