Patterns and Mechanisms of Ancestral Histone Protein
Inheritance in Budding Yeast
Marta Radman-Livaja1., Kitty F. Verzijlbergen2., Assaf Weiner3,4., Tibor van Welsem2, Nir Friedman3,4*,
Oliver J. Rando1*, Fred van Leeuwen2*
1Department of Biochemistry and Molecular Pharmacology, University of Massachusetts Medical School, Worcester, Massachusetts, United States of America, 2Division
of Gene Regulation, Netherlands Cancer Institute, and Netherlands Proteomics Center, Amsterdam, The Netherlands, 3School of Computer Science and Engineering, The
Hebrew University, Jerusalem, Israel, 4Alexander Silberman Institute of Life Sciences, The Hebrew University, Jerusalem, Israel
Replicating chromatin involves disruption of histone-DNA contacts and subsequent reassembly of maternal histones on the
new daughter genomes. In bulk, maternal histones are randomly segregated to the two daughters, but little is known about
the fine details of this process: do maternal histones re-assemble at preferred locations or close to their original loci? Here,
we use a recently developed method for swapping epitope tags to measure the disposition of ancestral histone H3 across
the yeast genome over six generations. We find that ancestral H3 is preferentially retained at the 59 ends of most genes,
with strongest retention at long, poorly transcribed genes. We recapitulate these observations with a quantitative model in
which the majority of maternal histones are reincorporated within 400 bp of their pre-replication locus during replication,
with replication-independent replacement and transcription-related retrograde nucleosome movement shaping the
resulting distributions of ancestral histones. We find a key role for Topoisomerase I in retrograde histone movement during
transcription, and we find that loss of Chromatin Assembly Factor-1 affects replication-independent turnover. Together,
these results show that specific loci are enriched for histone proteins first synthesized several generations beforehand, and
that maternal histones re-associate close to their original locations on daughter genomes after replication. Our findings
further suggest that accumulation of ancestral histones could play a role in shaping histone modification patterns.
Citation: Radman-Livaja M, Verzijlbergen KF, Weiner A, van Welsem T, Friedman N, et al. (2011) Patterns and Mechanisms of Ancestral Histone Protein Inheritance
in Budding Yeast. PLoS Biol 9(6): e1001075. doi:10.1371/journal.pbio.1001075
Academic Editor: Peter B. Becker, Adolf Butenandt Institute, Germany
Received September 23, 2010; Accepted April 22, 2011; Published June 7, 2011
Copyright: ? 2011 Radman-Livaja 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: OJR is supported in part by a Career Award in the Biomedical Sciences from the Burroughs Wellcome Fund. This research was supported by grants to
OJR and NF from the NIGMS (GM079205) and from the US-Israel Binational Foundation, and to FvL from the Netherlands Organisation for Scientific Research and
the Netherlands Genomics Initiative. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing Interests: The authors have declared that no competing interests exist.
* E-mail: email@example.com (NF); Oliver.Rando@umassmed.edu (OJR); firstname.lastname@example.org (FVL)
. These authors contributed equally to this work.
In addition to the information encoded in DNA sequence,
replicating cells can inherit epigenetic information, which refers to
variable phenotypes that are heritable without an underlying
change in DNA sequence. It is widely accepted that chromatin, the
nucleoprotein packaging state of eukaryotic genomes, provides one
potential carrier of epigenetic information. Although definitive
proof that chromatin per se carries epigenetic information during
replication exists in very few cases , genetic studies in numerous
organisms have identified key roles for chromatin regulators in
multiple epigenetic inheritance paradigms [2,3].
The idea that chromatin structure carries epigenetic informa-
tion poses a central mechanistic question—since chromosome
replication involves dramatic perturbations to chromatin structure
ranging from old histone displacement to widespread incorpora-
tion of newly synthesized histones, how can chromatin states be
stably maintained? To understand the mechanism by which
chromatin states could be inherited, it is necessary to understand
the unique challenges posed by histone protein dynamics during
replication [4–7]. First, histones must at least transiently dissociate
from the genome during passage of the replication fork—if old
histones carrying epigenetic information do not re-associate with
daughter genomes at the location from which they came, this
could lead to ‘‘epimutation,’’ analogous to DNA bases moving
relative to one another during genomic replication. Second, it is
unknown to what extent newly synthesized histones deposited at
different loci differ in their covalent modification patterns. Finally,
how old histones influence new histones, the basis for positive
feedback, can be considered analogous to asking what the
equivalent of base-pairing is during chromatin replication.
Classic radioactive pulse-chase studies demonstrated that, in
bulk, maternal histones segregate equally to the two daughter cells
[4,8–10]. It is unknown, however, whether maternal histones
remain close to the locus from which they were evicted by the
replication fork or whether maternal histones are incorporated at
preferred genomic loci in the two daughter genomes [5,7,11]. The
extent of maternal histone dispersal affects the stability of
epigenetic states in theoretical models of chromatin inheritance
, making experimental determination of this parameter a key
goal for epigenetics research.
To address these fundamental questions, we carried out a
genetic pulse-chase with epitope-tagged histone H3  to follow
ancestral H3 for several cell divisions after removal of the ancestral
PLoS Biology | www.plosbiology.org1June 2011 | Volume 9 | Issue 6 | e1001075
tag. We find that old histone proteins do not accumulate at
epigenetically regulated loci such as the subtelomeres but instead
accumulate at the 59 ends of long, poorly transcribed genes. As
expected, old histones do not accumulate at loci exhibiting rapid
histone turnover, but we also find that 39 to 59 movement of old
histones along coding regions and histone movement during
replication are required to explain the patterns of ancestral histone
retention we observe. We estimate that maternal histones stay
within ,400 bp of their original location during replication,
providing the first measure of this crucial parameter. Finally, we
identify a number of factors that affect old histone localization,
such as topoisomerase I and the H4 N-terminal tail, which both
affect the 59 bias in localization patterns. In contrast, CAF-1
mostly affects histone turnover at promoters. Together, these
results provide a detailed overview of the movement of ancestral
histones across multiple cell generations and identify a number of
mechanisms that play a role in shaping the landscape of ancestral
To follow the movement of old histone proteins over multiple cell
generations, we utilized a novel pulse-chase technique  to follow
ancestral epitope-tagged histone H3 for several cell divisions after
swapping epitope tags from H3-HA to H3-T7 (Figure 1A, B). We
have previously described use of this technique to assay replication-
independent H3 turnover in arrested cells and have shown that
prior to recombination all cells carry the H3-HA, and that
recombination is 98% efficient in cells that are not dividing due
to nutrient deprivation (Figure S1). Unlike inducible pGAL-based
systems for measuring replication-independent histone dynamics
[14–17], here the epitope-tagged histone is under the control of its
endogenous promoter, avoiding potential artifacts of H3/H4
misexpression  on histone dynamics throughout the cell cycle.
We used MNase-ChIP [15,19] for the HA and T7 tags after
recombination but before release into the cell cycle, and 3 and 6
generations after releasing yeast into the cell cycle . This
material was hybridized to tiling microarrays covering 4% of the
yeast genome , and HA/T7 ratios of normalized HA and T7
signals were computed for the 3 and 6 generation data (Figure 1C,
D). Since HA is eliminated via recombination leaving new H3-T7,
high HA/T7 ratios indicate loci enriched for ancestral histone H3.
Surprisingly, many of the highest HA/T7 levels were associated
with coding regions (discussed below).
Overall, HA/T7 patterns are consistent at 3 and 6 generations,
but the dynamic range of HA/T7 enrichment diminished from 3
to 6 generations (Figure 1D, Figure S2). This is an expected
consequence of the fact that ,1%–2% of cells do not recombine
the HA tag away (Figure S1)—since the amount of ancestral H3 is
decreasing by at least 2-fold in each generation, the relative
contribution of the ,2% of cells still expressing H3-HA will
increase over time, with this genomic background eventually
competing with the real signal from increasingly rare ancestral H3
(,2% of total H3 after 6 doublings).
Ancestral Histones Are Retained Over Long, Poorly
To extend our analyses to the entire genome, we carried out
deep sequencing of HA and T7 libraries. HA- and T7-tagged H3
were immunoprecipitated after the tag swap but before release
from arrest (0 generations), after release into a G2/M cell cycle
block, and at 1, 3, and 6 generations after release. Sequencing
reads were mapped to the yeast genome, normalized for read
count, and HA/T7 ratios were computed genome-wide. These
data correlated well with our microarray data, and we further
validated these measurements by q-PCR at SPA2 and BUD3, two
genes which both exhibit high and low HA/T7 ratios at their 59
and 39 ends, respectively (Figure S3).
In previous work, we and others [13,15–17,21–23] showed that
there is a partial correlation between transcription levels and
replication-independent histone dynamics. To understand how
transcription might affect multigenerational histone retention in
our system, we aligned all yeast genes by their transcription start
site (TSS) and clustered genes (K-means, K=5) based on the
pattern of the 3-generation HA/T7 ratios along the gene body
(Figures S4, S5, Table S1). We observed a striking enrichment of
H3-HA just downstream of the 59 ends of genes (typically peaking
around the +3 nucleosome). One exception to the 59 pattern
described is found in one cluster of short genes with uniformly low
H3-HA levels (Figure S4, Cluster 1), which is enriched for GO
categories (such as protein translation) related to high gene
expression levels. In contrast, long genes were generally associated
with higher levels of ancestral H3 (see for example Cluster 5).
To better visualize these trends, we sorted genes by the extent of
ancestral H3 retention after 3 generations (Figure 2A–B).
Retention of ancestral histones correlates both with low expression
levels and with longer genes (Figure 2C–D, Figure S6). While it is
the case that longer genes tend to be expressed at lower levels than
short genes (Figure 2E), these factors are partially independent
here—even when we focus on genes of 1–2 kb length, we still
observe the correlations between ancestral histone retention and
low expression (Figure 2E–F, and see below). Interestingly, in both
microarray and sequencing datasets we found that epigenetically
repressed loci such as the silent mating loci and subtelomeres
[24,25] did not preferentially accumulate ancestral histone
proteins (Figure 1C, Figure S7)—analysis of both unique and
repetitive subtelomeric genes showed similar H3-HA retention
patterns to euchromatic genes of similar length and expression.
This was not a consequence of silencing defects in our strains, as
they showed efficient mating (unpublished data).
It is widely believed that chromatin, the nucleoprotein
packaged state of eukaryotic genomes, can carry epige-
netic information and thus transmit gene expression
patterns to replicating cells. However, the inheritance of
genomic packaging status is subject to mechanistic
challenges that do not confront the inheritance of
genomic DNA sequence. Most notably, histone proteins
must at least transiently dissociate from the maternal
genome during replication, and it is unknown whether or
not maternal proteins re-associate with daughter genomes
near the sequence they originally occupied on the
maternal genome. Here, we use a novel method for
tracking old proteins to determine where histone proteins
accumulate after 1, 3, or 6 generations of growth in yeast.
To our surprise, ancestral histones accumulate near the 59
end of long, relatively inactive genes. Using a mathemat-
ical model, we show that our results can be explained by
the combined effects of histone replacement, histone
movement along genes from 39 towards 59 ends, and
histone spreading during replication. Our results show that
old histones do move but stay relatively close to their
original location (within around 400 base-pairs), which
places important constraints on how chromatin could
potentially carry epigenetic information. Our findings also
suggest that accumulation of the ancestral histones that
are inherited can influence histone modification patterns.
Transgenerational Histone Retention in Yeast
PLoS Biology | www.plosbiology.org2June 2011 | Volume 9 | Issue 6 | e1001075
What properties of short or highly transcribed genes might lead
to loss of ancestral histones? Replication-independent histone
replacement occurs most rapidly over intergenic regions and over
the coding regions of highly transcribed genes [15,17,21,23], the
converse of the pattern of ancestral H3 retention we observe.
Indeed, ancestral histone retention is broadly correlated with
‘‘cold’’ regions of low H3/H4 turnover (Figure 3A). Importantly,
however, for a given level of H3/H4 turnover, ancestral H3
retention varied significantly—retention at a given nucleosome
was better correlated with the average turnover rate of several
surrounding nucleosomes than with the immediate turnover rate
(see, for example, Figure 3B–C). This observation suggests that
maternal histones preferentially re-associate with daughter ge-
nomes near the location from which they originated—if old
histones scattered randomly at replication, ancestral H3 retention
patterns should more precisely anticorrelate with replication-
independent turnover patterns, as is discussed in more detail
Figure 1. Overview of system for tracking ancestral histone proteins. (A) Recombination-based swapping of epitope tags on histone H3.
Histone H3 is tagged at its endogenous locus with a C-terminal HA epitope tag surrounded by LoxP sites. Upon induction of Cre recombinase with b-
estradiol, the HA tag is recombined out and H3 is left with a C-terminal T7 tag. (B) Experimental overview. Yeast carrying HA-tagged H3 are arrested
by nutrient depletion, and the HART7 swap is induced by overnight incubation with b-estradiol. After the tag swap, yeast are released from arrest
and HA and T7 tags are mapped across the genome at varying times post-release. (C) Chromosome III overview. HA/T7 ratios are shown as a heatmap
across chromosome III at 3 generations after release. Notable in this view is a lack of accumulation of H3-HA at TEL3L or the silent mating loci. (D)
Close-up views of two genomic loci. Data are shown as a heatmap for 3 and 6 generations after the tag swap.
Transgenerational Histone Retention in Yeast
PLoS Biology | www.plosbiology.org3 June 2011 | Volume 9 | Issue 6 | e1001075
Transgenerational Histone Retention in Yeast
PLoS Biology | www.plosbiology.org4 June 2011 | Volume 9 | Issue 6 | e1001075
Accumulation of Ancestral Histones at 59 Ends of Genes
Why do old histone proteins accumulate near the 59 ends of
genes? We considered two alternative possibilities for classes of
mechanisms causing this pattern. In the first mechanism, we
reason that if histone proteins tend to maintain their locations
along the genome, the 59 enrichment of old histones implies that
old 39 histones are evicted and replaced by new histones during
some phase of the cell cycle. However, previous measures of
turnover in G1- or G2/M-arrested yeast [15,26] cannot explain
the 59/39 ratios we observe. Furthermore, we found that mutations
in candidate 59/39-marking complexes such as cohesin [27,28] or
H3K4/K36 methylases  did not affect 59-biased retention of
old histones at target loci (Figure S8A).
A second possible explanation for widespread 59 accumulation
of ancestral histone proteins is that the histone proteins move from
39 to 59 over genes over time. This could result from RNA
polymerase passage, because some RNA polymerases pass histone
octamers in a retrograde direction during transcription [30,31].
Although it is debatable whether this is true of Pol2 in vitro
[32,33], in vivo we previously observed that inactivation of Pol2
leads to a modest shift of nucleosomes from 59 to 39 ,
consistent with the idea that Pol2 movement normally shifts
nucleosomes in a 59 direction. To test whether this movement was
related to transcription, we asked whether the 59 peak of H3-HA
accumulation shifted further 59 with increasing transcription rate.
We normalized all gene lengths to one, then plotted the HA/T7
ratio for all genes sorted by transcription rate (Figure S9).
Consistent with the prediction of transcription-dependent retro-
grade movement, we did observe a subtle signal of H3-HA peaks
shifting further 59 at higher transcription rates. While this analysis
could be confounded by the higher transcription rates seen over
shorter genes, even when we focus on 1–2 kb genes, we observe
that poorly transcribed genes exhibit a much flatter profile than
genes expressed at average levels (Figure 2F), as expected if Pol2
transit were required for H3/H4 ‘‘passback.’’ Finally, we show
below that per-gene estimates of passback exhibit significant
correlation with Pol2 levels. Together, these results are most
consistent with a model in which histone proteins move from 39 to
59 over coding regions over time (further detailed in the
Quantitative Estimation of Nucleosome Dynamics During
A key question we sought to address in this study is whether
maternal histones re-associate near their original positions after
passage of the replication fork. We reasoned that changes in HA/
T7 patterns over the course of several generations might provide
insight into the effects of replication on nucleosome dynamics.
HA/T7 patterns change dramatically between arrest and 1
generation of release (with or without G2/M arrest) and then
are very similar between 1 and 3 generations, before the
background of nonswitching cells starts to dominate the profile
at 6 generations (Figure 4). As expected, HA/T7 data at
generation 0 exhibited widespread HA loss/T7 gain at promoters
and +1 nucleosomes as a result of the rapid replication-
independent turnover at these loci [13,15,17]. Importantly, to
rule out the possibility that 59 accumulation of H3-HA was an
effect of our arrest-release protocol, we also measured HA/T7
distributions 6 h after inducing recombination in actively growing
midlog cultures of yeast (Figures S3 and S10). Despite heteroge-
neity in switching times in this protocol (only 65% of yeast have
switched from HA to T7 3 h after switch induction, 85% after
6 h), we nonetheless observed that HA/T7 distributions were
remarkably similar in midlog-switching cells to HA/T7 patterns
observed in cells undergoing the arrest/release protocol, with
preferential ancestral histone accumulation at the 59 ends of long,
poorly transcribed genes.
We asked whether these dynamic observations could be used to
quantitatively rule out specific models concerning the mechanisms
for segregation of maternal histones to daughter genomes.
However, the resolution of this question is complicated by
replication-independent processes we discuss above that can
remove or shift ancestral histones, and that cannot be fully
removed experimentally (for example, yeast will not proceed
through the cell cycle in the absence of RNA polymerase). To
understand the relationship between these issues, we designed an
analytical model that accounts for three processes that affect H3
molecules in coding sequences (Figure 5A) and then examined the
effect of removing any of the three. Briefly, our model includes a
nucleosome-specific term for H3 turnover taken from prior
experimental results , with H3 turnover resulting in loss of
HA. In addition, it includes a gene-specific parameter accounting
for lateral movement of histones (‘‘passback’’). Further, the model
also includes a global parameter that describes the extent of
histone ‘‘spreading’’ via dissociation/re-association during repli-
cation. Finally, the experimentally measured background of 2%
nonswitching cells (Figure S1) was included. The free parameters
of the model (describing global histone spreading and gene-specific
lateral movement per generation) were estimated to maximize the
likelihood of experimental observations (Text S1).
To account for any first-pass effects of Pol2 behavior during
initial re-feeding of nutrient-depleted yeast (Figures S3 and S10),
we examined this model with two starting conditions—the first
started with a uniform genomic distribution of H3-HA, while the
second started with the experimental distribution of HA/T7
observed after release into G2/M arrest (Figure 4). Both model
variants predicted HA/T7 ratios with good correlations to the
experimental data (Figure 5B shows data starting from a uniform
distribution, Figure 5E and Figure S11 start from the G2/M
distribution). Examination of estimated parameters revealed
expected behaviors. For instance, the distribution of lateral histone
movement estimates (Figure 5C) was strongly biased towards
Figure 2. Ancestral H3 molecules accumulate at the 59 ends of long, poorly transcribed genes. (A–B) Heatmap of sites of ancestral H3
accumulation. Genes are aligned by TSS (indicated), and Log2HA/T7 ratios are indicated as a heatmap. Genes are ordered by the median HA/T7 ratio
over the 59-most 1 kb at 3 generations. Grey over coding regions indicates missing data; grey downstream of genes indicates sequence downstream
of the 39 end of the gene to show gene length. Accumulation of ancestral histones at the 59 ends of genes peaks around the +3 nucleosome, as
expected given that the +1 and +2 nucleosomes are generally subject to high rates of replication-independent H3/H4 replacement [15,17]. (C) An 80
gene sliding window average of Pol2 ChIP levels  for genes ordered as in (A–B), showing that genes with low levels of ancestral H3 retention are
highly transcribed. (D) 80 gene sliding window average of gene lengths, showing that genes with high levels of ancestral H3 retention tend to be
long. (E) The median HA/T7 ratio over the 59 end of genes (1 kb) was calculated for all genes, and median values of this retention metric are shown
for groups of genes ordered by transcription rate (x-axis) and gene length (y-axis). While these are not independent—highly expressed genes tend to
be short—for a given gene length genes transcribed at higher levels exhibit low HA retention levels. This is true mostly of genes shorter than 3 kb,
which encompasses the majority of yeast genes. (F) Average HA/T7 ratios (Log2) for genes between 1 and 2 kb, broken into high (red), low (green),
and intermediate (blue) transcription rates.
Transgenerational Histone Retention in Yeast
PLoS Biology | www.plosbiology.org5June 2011 | Volume 9 | Issue 6 | e1001075
Figure 3. H3 retention anticorrelates with replication-independent turnover in a gene length-dependent manner. (A) Scatterplot of
ancestral H3 retention (median Log2 HA/T7 for the 59 1 kb, y-axis) versus replication-independent turnover (Dion et al. , Z score, x-axis). (B) HA
retention is plotted against 59 H3 turnover as above but with short and long genes plotted separately. For a given level of H3 turnover, ancestral
retention is greater at longer genes. (C) Averages of the 59 HA/T7 retention parameter (median HA/T7 for the 59-most 1 kb) are shown for genes
broken into different length and 59 turnover groups. For all turnover levels, longer genes retain more H3-HA than do shorter genes.
Transgenerational Histone Retention in Yeast
PLoS Biology | www.plosbiology.org6 June 2011 | Volume 9 | Issue 6 | e1001075
retrograde 39 to 59 movement of histones, consistent with the
previously measured effects of rpb1-1 inactivation on nucleosome
(p=9.6439e-19) correlated with transcription rate (Figure S12).
Our model allows us to estimate the extent of histone movement
during replication. Figure 5D shows the likelihood of the full
model plotted for various values for replication-dependent histone
spreading. The best fit model allowed histones to spread ,400 bp
in either direction, or roughly two nucleosome widths, during
replication (more precisely, in this model two-thirds of histones
stay within 400 bp of their original locations, as this value is the
standard deviation of a Gaussian function describing spreading;
see Text S1). Results from models with 800 and 1,600 bp
spreading parameters are shown in Figure 6 for comparison. Our
estimate of 6400 bp spreading is particularly interesting given
electron microscopy results demonstrating that nucleosomes are
destabilized over 650–1,100 bp around the replication fork on
replicating SV40 minichromosomes [35,36].
Elimination of any one or two of the three components of the
model—spreading, turnover, or passback—resulted in significantly
worse fits between model predictions and experimental data
(Figure 5E). This can be intuited as follows. First, in the absence of
histone spreading, unmitigated histone movement from 39 to 59
results in a much tighter 59 ancestral histone peak and results in
much more extensive change from one generation to the next than
we observe. Second, eliminating histone turnover shifts the 59
ancestral peak closer to the +1/+2 nucleosome. Third, preventing
lateral histone movement results in a 39-shifted, flatter ancestral
While our model provided good quantitative fits of ancestral H3
patterns for many genes, we nonetheless note that many genes
were not perfectly fit by this model. Generally, we found that the
model poorly fit short genes, and overall the model almost
universally predicted lower HA/T7 at the +N nucleosome (the last
nucleosome in a gene) than was observed (Figures S11, S13). We
ascribe these failures to the fact that we considered each gene in
isolation and therefore did not model shifts of old nucleosomes
from adjacent genes, which would result in poor fits over short
genes in particular. Interestingly, the better fit at the +1
nucleosome than at the +N nucleosome is consistent with rapid
promoter turnover more effectively isolating genes from one
another at their 59 ends in vivo.
Overall, the strong correlation between our model and the
experimental data supports the hypothesis that at least three
dynamic processes affect nucleosomes and shape the landscape of
ancestral histone retention and provide the first quantitative
estimate of maternal histone dynamics during replication.
Topoisomerase I and the H4 N-Terminal Tail Play Roles in
Establishing the 59/39 Gradient of Ancestral H3 Molecules
To further investigate the mechanism of 59 accumulation, we
asked whether gene-specific passback parameters were correlated
with specific gene annotations (Table S2) [34,37]. Interestingly, we
find that the estimated passback distance was much greater at
TFIID-dominated (‘‘growth’’) genes than at SAGA-dominated
(‘‘stress’’) genes (Figure 7A) . As a result, 59 accumulation was
much more pronounced at TFIID-dominated than at SAGA-
dominated genes (Figure 7B). Almost every described aspect of
chromatin structure and gene expression, from nucleosome
positioning to evolutionary lability (reviewed in [39,40,41]), differs
between these two broad types of genes. Mechanistically, one
interesting correlate is that TFIID recruitment has been proposed
to be mediated in part by acetylation of the N-terminal tail of
histone H4 [38,42].
To investigate this link experimentally, we examined whether
mutations of the H4 tail influenced ancestral histone H3 retention.
In an H4K5,12R mutant that cannot be acetylated on these two
tail residues, the 59-biased HA/T7 was partially lost (Figure S8B),
Figure 4. Kinetic analysis of ancestral H3 retention. HA/T7 ratios were measured genome-wide after recombination but before release (Gen 0),
after release into nocodazole (G2/M), and after 1, 3, or 6 generations of growth post-release. Data for all genes were averaged and are plotted as
Transgenerational Histone Retention in Yeast
PLoS Biology | www.plosbiology.org7 June 2011 | Volume 9 | Issue 6 | e1001075
Figure 5. Quantitative modeling reveals three distinct dynamic processes. (A) Outline of quantitative model. From a given starting
distribution, histones are subject to turnover , transcription-associated lateral movement (‘‘passback’’), and replication-mediated spreading.
Model is described in detail in Text S1. (B) The model captures major features of the experimental data. HA/T7 ratios for experimental data and model
predictions are shown for all genes as a heatmap. (C) Distribution of lateral passback parameter (per generation) for all genes. Note that the vast
majority (92%) of genes were associated with retrograde 39 to 59 movement along coding regions. (D) Estimation of replication-based spreading of
maternal histones. Model likelihood (Text S1) is plotted on the y-axis for various width spreading distributions (defined as 1 standard deviation of the
Gaussian describing histone movement at replication—see Text S1 for model details). (E) Eliminating any of the three model features worsens fit to
data. Plotted are averages at 3 generations for genes over 2 kb for data versus predictions of various models (‘‘STP’’ refer to replication-mediated
Transgenerational Histone Retention in Yeast
PLoS Biology | www.plosbiology.org8 June 2011 | Volume 9 | Issue 6 | e1001075
Spreading, replication-independent Turnover, and Passback). Note that the model eliminating turnover underestimates turnover effects, as histones
that spread or are passed over the 59 end of the gene are still eliminated in this model (i.e., in this model we effectively only eliminate turnover within
CDS, not in intergenic regions), providing another basis for high loss of 59 histones.
Figure 6. Dependency of histone dynamics model on spreading parameter. (A–B) Parameters in the quantitative model described in
Figure 5 were re-optimized after fixing the spreading term to 400 bp (as in Figure 5), 800, or 1,600 bp. Data and simulations are shown averaged for
genes over 2 kb for models starting with a uniform H3-HA distribution (A) or starting with the experimentally measured G2/M HA/T7 distribution (B).
(C–D) Examples of data and three models with different spreading parameters. Genomic coordinates are chromosome 2 490–540 kb (C), and
chromosome 1 60–110 kb (D). Y-axis shows measured (Data) or predicted HA/T7 values, in Log2.
Transgenerational Histone Retention in Yeast
PLoS Biology | www.plosbiology.org9 June 2011 | Volume 9 | Issue 6 | e1001075
Figure 7. Mutants affecting ancestral histone retention. (A) Distribution of lateral nucleosome distances from model (Figure 5). Shown are the
passback parameters for SAGA-dominated and TFIID-dominated genes as defined in Huisinga et al. . (B) TFIID-dominated genes preferentially
accumulate 59 H3-HA. Averages of 3 generation experimental data are shown for the indicated gene classes. (C) H4 tail deletion dramatically reshapes
the landscape of ancestral histone retention. Yeast carrying an N-terminal H4 tail deletion were processed as in Figure 1A–B, and averages for all
genes are plotted as indicated. We note that this strain has retained a wild-type HHT2-HHF2 locus for viability, so results must be interpreted with
Transgenerational Histone Retention in Yeast
PLoS Biology | www.plosbiology.org 10June 2011 | Volume 9 | Issue 6 | e1001075
consistent with the possibility that acetylation of H4 tail lysines
may contribute to H3/H4 passback. We also deleted the H4 N-
terminal tail, although in this strain background this mutation
proved lethal and so all recovered strains retained a wild-type copy
of the H4-H3 locus (HHF2-HHT2). Thus, results with this strain
must be interpreted with extreme caution, as we do not know the
effect of wild-type, untagged nucleosomes on the behavior of the
Nonetheless we present here results of mapping of HA and T7 3
generations after release from the HA/T7 tag swap, since the H4
tail deletion has dramatic effects on global nucleosome dynamics
(Figure 7C), with low HA/T7 at 59 ends followed by a nearly flat
profile over the remainder of coding regions. This profile suggests
a requirement for the H4 tail in H3/H4 passback, and possibly on
replication-mediated spreading (see Figure 5E). Interestingly, the
effect of H4 tail deletion was much more pronounced at TFIID-
dominated genes (Figure 7D), suggesting that the exaggerated H3/
H4 passback inferred at these genes involves the H4 tail. The
effects of the H4 tail deletion were not simply due to the extensive
changes in the transcriptome , as we measured changes in
genome-wide RNA Pol2 localization in our H4 mutant strains,
finding that the relationship between Pol2 levels and HA/T7
behavior qualitatively changed in this mutant (Figure S14). While
we must be cautious interpreting results obtained with the H4 tail
deletion, the fact that H4K5,12R mutants (which were viable and
did not retain any wild-type H3/H4) also exhibit diminished 59
bias in ancestral H3 retention provides independent support for a
key role for the H4 tail in H3/H4 passback.
We also explored the role of supercoiling in the 59-biased
retention of old histones. Topoisomerases relax DNA supercoiling
and thereby help to maintain chromatin architecture. Transcrip-
tion of DNA templates by Pol2 differentially affects supercoiling in
front of and behind the passing polymerase, thereby differentially
affecting 59 and 39 nucleosomes [44,45]. To assess the role of this
activity in 59 accumulation of old histones, we examined the
consequences of inactivation of the major topoisomerase Top1,
which in vitro can resolve both negative and positive supercoils
[46,47]. Cells lacking Top1 showed reduced 59 bias in ancestral
nucleosome accumulation (Figure 7E), indicating that resolving
DNA topology problems before or after passage of the
transcription or replication machinery influences the mobility
and/or stability of nucleosomes. Consistent with expectations of a
greater buildup of supercoils over longer transcription units, we
confirmed a stronger effect of TOP1 deletion at longer genes
Replication Timing, Chaperones, and Ancestral Histone
We finally turn to the role of replication factors in ancestral
histone retention. We first asked whether replication timing
affected H3-HA retention. Nucleosomes surrounding early-firing
origins tended to lose H3-HA more rapidly than late-firing origins
(unpublished data), but this likely stems from the fact that
replication timing correlates with replication-independent turn-
over [23,26]. Focusing only on nearby coding regions (Figure S15),
we found that late-replicating genes were associated with slightly
59 shifted ancestral H3 peaks relative to genes near early origins
(consistent with decreased spreading or turnover), suggesting that
different replication forks might affect chromatin in different ways,
although the modest effect precludes a stronger interpretation.
To directly address the role of fork-associated chromatin
proteins in histone spreading at replication, we examined
mutations of PCNA and Chromatin Assembly Factor (CAF-1),
which plays a key role in replication-coupled histone deposition
[48,49]. Three different mutants of PCNA that disrupt interac-
tions with replication proteins or with replication-coupled
chromatin-assembly factors showed only minor effects on 59
retention of ancestral H3 at target genes SPA2 and BUD3 (Figure
S8C). In contrast, ancestral H3 retention at the 59 ends of these
target genes was slightly increased upon deletion of the CAF-1
subunit CAC1 (unpublished data). To further explore the role of
CAF-1 in histone retention patterns, we deep sequenced HA and
T7 tags from cac1D yeast 3 generations after release (Figure 8).
These data show a dramatic 59 shift in the peak of ancestral H3
retention in these mutants. This shift is most consistent with a
decrease in histone turnover at the 59 ends of genes in this mutant,
which we have independently confirmed using G1-arrested yeast
expressing pGAL-driven Flag-H3 . However, we cannot rule
out the possibility that the role of CAF-1 in retention of old
histones in 59 and promoter regions involves interactions with
PCNA during DNA replication. Interestingly, the 59 accumulation
observed in wild-type yeast is otherwise little changed in the CAF-
1 mutant over long genes (Figure S16), suggesting that 39 to 59
movement of histones is normal, and that preferential retention of
old histones at their maternal locations may be carried out by
alternative histone chaperones such as the Hir complex or Asf1 in
this mutant. Unfortunately, both hir and asf1 mutants are lethal in
our strain background (likely because our strain carries only one
copy of the H3/H4 gene pair ), preventing us from testing this
Consequences for Histone Modification Patterns
Our results are most consistent with histone retrogression from
39 to 59 over genes, which raises the question of whether old
histones carry modifications associated with mid- and 39 coding
regions (e.g., H3K36 and H3K79 methylation) towards the 59 end
of genes. Alternatively, there could be active erasure of these
modifications. We therefore compared genes exhibiting high levels
of ancestral H3 retention with prior genome-wide analyses of
histone modifications [52,53]. Histone modification patterns
generally conformed to the patterns expected based on transcrip-
tional behavior—genes that retain high levels of ancestral histones
are poorly transcribed (Figure 2D), and correspondingly exhibit
low levels of transcription-related marks H3K9ac, H3K14ac,
H4ac, and H3K4 methylation (Figure S17 and unpublished data).
However, these are all 59-biased marks [19,29,52,54], and based
on retrograde movement of old histones are therefore not expected
to accumulate with age.
More interestingly, we found that genes with high levels of old
regions, particularly at the 59 end (Figure 9, Figure S17). The
H3K79 methylase is nonprocessive, indicating that K79 methyla-
tion status should essentially act as a timer . Further, analyses of
genome-wide H3K79me3 patterns show anticorrelation between
caution. However, we find similar but less dramatic effects in an H4K5,12R mutant (Figure S8B), supporting the observation here that passback is
affected by the H4 N-terminal tail. (D) H4 tail deletion preferentially affects TFIID-dominated genes. Data for wild-type and H4tailD yeast are plotted
for the indicated gene classes. (E) Topoisomerase I plays a role in 59 accumulation of ancestral histones. top1D yeast were processed as in Figure 1A–
B, and averages for all genes are plotted as indicated. (F) TOP1 deletion affects 59 passback preferentially at long genes. Data for wild-type and top1D
yeast are plotted for the indicated gene classes.
Transgenerational Histone Retention in Yeast
PLoS Biology | www.plosbiology.org11June 2011 | Volume 9 | Issue 6 | e1001075
this modification and locations of high nucleosome turnover
[15,56], supporting the idea that K79me3 identifies old H3 protein.
We recently confirmed that old H3 protein is enriched for
H3K79me3 by mass spec analysis of old nucleosomes (D. DeVos,
FvL et al., submitted).
Finally, we also observed higher levels of H3K36me3 at 59 and
mid-CDS of genes exhibiting elevated ancestral histone retention
relative to genes with intermediate H3-HA retention (Figure S17).
This observation is consistent with the above hypotheses that old
histones move from 39 to 59 and thus might carry typical mid-CDS
and 39-end histone modifications to the 59 ends of genes (Figure
S18). Together, these results provide further evidence that our
system accurately captures the behavior of old histones.
The fate of stable proteins in rapidly dividing cells is of great
interest for fields from protein damage to aging to epigenetic
inheritance. In particular, models for the inheritance of chroma-
tin-based information [5–7,12,57,58] require a quantitative
understanding of the fate of specific maternal histone proteins
during the disruptive replication process. Some models for
epigenetic inheritance of chromatin states require that old
nucleosomes are retained near their original positions, whereas
other models (such as those based on replication timing; ) are
less sensitive to the fates of old histones. However, due to the lack
of methods to directly track histone dispersal during replication,
these models have not been experimentally tested in vivo. Here, by
using a novel genetic pulse-chase assay, we characterize ancestral
histone retention patterns across the yeast genome. By accounting
for known replication-independent processes, we used these data
to estimate the effects of replication on histone movement, finding
that H3/H4 are retained close to their original locations during
replication. We also identified a number of mutants that affect
various aspects of ancestral H3/H4 movement and retention.
Ancestral Histones Accumulate at the 59 Ends of Genes
Most surprising to us was the observation that ancestral H3
molecules accumulate near the 59 ends of coding regions, peaking
around the +3 nucleosome. The high HA/T7 ratio observed at the
59 ends of genes is not an artifact of the epitope tags used, as we
have observed the converse behavior (high T7/HA) when we
switch the epitope tags used (unpublished data). Furthermore, this
unusual behavior is not an artifact of the conditions used for
growth arrest and release, as we observe a similar 59/39 gradient of
H3-HA when yeast are subjected to the epitope switch during
active midlog growth (Figure S10).
What is the mechanistic basis for the 59/39 gradient of HA/T7
we observe? We consider two classes of mechanisms—in one,
histone proteins do not move laterally and the 59/39 gradient
results from preferential loss of 39 H3/H4, while in the other the
gradient results from lateral histone movement combined with loss
at the 59 end. While we cannot definitively answer which
mechanism explains our results, we strongly disfavor a model
with preferential 39 nucleosome eviction and no lateral movement
based on the following observations. First, we tested a number of
relevant mutants for changes in the 59/39 HA/T7 bias (Figure S8).
Loss of H3K4 methylation (a 59-biased histone mark) or H3K36
methylation (a mid and 39-biased histone mark) did not affect HA/
T7 patterns at selected target genes. Similarly, 59 retention of
ancestral H3 was unaffected by mutants of cohesin, whose loading
is associated with regions of high H3/H4 turnover and which
accumulates at the 39 ends of genes [27,28]. Second, direct
measurements of H3/H4 turnover using a pGAL-driven epitope
tagged H3 do not provide evidence for ubiquitous 39 histone
replacement during G1 arrest , during G2/M arrest , or in
Figure 8. Effects of Chromatin Assembly Factor-1 complex on ancestral H3 patterns. Yeast lacking CAF-1 subunit Cac1 were processed as
in Figure 1A, and HA/T7 ratio averages are shown for all genes in wild-type and cac1D mutants 3 generations after release.
Transgenerational Histone Retention in Yeast
PLoS Biology | www.plosbiology.org12 June 2011 | Volume 9 | Issue 6 | e1001075
unsynchronized yeast . Thus, while we cannot definitively rule
out some cryptic 39 replacement event in this system, all direct tests
have failed to support this hypothesis.
Conversely, multiple observations support the hypothesis that
H3/H4 proteins move from 39 to 59 over protein-coding regions
over time. First, seminal in vitro studies on transcription of
nucleosomal templates showed that several RNA polymerases
can transcribe through a nucleosome without displacing the
H3/H4 tetramer. The proposed mechanism by which histones
remain associated with the DNA is a ‘‘bubble propagation’’
mechanism—DNA partially unwraps from the histones, RNA
polymerase enters, and DNA behind the polymerase re-
associates with the histone octamer, resulting in a net retrograde
movement of histones after the polymerase has passed. This
mechanism is relatively well established for SP6 polymerase and
RNA Polymerase III [30,31], whereas there is some controversy
regarding the effect of RNA Polymerase II on nucleosome
positioning [32,33]. Of course, it is not unreasonable to expect
that nucleosome movement during transcription in vivo will also
be affected by polymerase-associated factors such as histone
chaperones and ATP-dependent remodelers that are not present
in the in vitro systems. In any case, these studies provide a
plausible mechanism by which RNA polymerase transit results
in retrograde nucleosome movement.
Second, we have previously found that inactivation of Pol2
using the temperature-sensitive rpb1-1 allele results in a net 59 to
39 shift in the majority of coding region nucleosomes ,
consistent with the hypothesis that polymerase transit normally
shuttles nucleosomes from 39 to 59. Third, highly transcribed
genes (such as those encoding ribosomal proteins) in yeast
paradoxically exhibit very tightly spaced coding region nucleo-
somes (e.g., 155–160 bp between adjacent nucleosomes rather
than ,165 bp), and this tight spacing relaxes upon Pol2
inactivation, again consistent with nucleosomes being passed
upstream during transcription . Taken together with the
absence of any evidence for 39 H3/H4 eviction, we therefore
argue that the most parsimonious explanation of the surprising 59
accumulation of ancestral histones is retrograde movement of
histones over genes against the direction of transcription. Note
that while we favor the hypothesis that the act of RNA
polymerase transit itself is the mechanism linking transcription
to H3/H4 passback, polymerase is not the only candidate factor
leading to retrograde histone movement. Notably, we found that
top1D mutants exhibit diminished signatures of H3/H4 passback
(Figure 7), and this decrease was stronger at longer genes,
suggesting the possibility that some aspect of cleavage and
rotation of twisted DNA by Top1 contributes to the passback
observed. However, it is also possible that Top1 differentially
affects histone turnover in 59 and 39 regions or affects passback by
affecting Pol2 passage .
We analytically assess several predictions of the ‘‘passback’’
model. First of all, if RNA polymerase transit were the driver of
retrograde histone movement, then one might predict that
passback should correlate with transcription rate. We find the
expected correlation to be statistically significant (p=9.6439e-19,
Figure S12) but weak nonetheless (R=0.12). Importantly, we
previously observed that 59 to 39 nucleosome movement in rpb1-1
mutants was also significantly but poorly correlated with
transcription rate . The reason for the mediocre correlation
between polymerase abundance and passback is hinted at by the
fact that TFIID-dominated genes exhibit much greater passback
values than do SAGA-dominated genes (Figure 7A). We have
previously noted that SAGA-dominated (‘‘stress’’) genes exhibit
higher levels of H3 turnover, per polymerase, than do TFIID-
dominated genes . In vitro, a single polymerase’s transit
displaces an H2A/H2B dimer from the histone octamer, but a
second polymerase encountering a histone hexamer will displace
the remaining histones [61,62]. Coupled with the observation
that SAGA-dominated genes exhibit larger ‘‘bursts’’ of polymer-
ase, this suggests that closely spaced polymerases are required for
H3/H4 eviction over coding regions, but evenly spaced
polymerases leave time for dimer replacement on damaged
nucleosomes [54,62]. We believe this model also explains some of
the behavior of ancestral histones in this study—SAGA-
dominated genes display little passback and overall diminished
levels of ancestral H3 (Figure 7A–B), an expected consequence of
the loss of old histones via turnover. Correlations between
polymerase and passback are therefore expected to be subtle—at
increasing transcription rates, we expect an increased likelihood
of a closely spaced pair of polymerases, and the resulting H3/H4
eviction would eliminate any trace of the passback that had
occurred to that point.
It is important to note that the transcription-dependent
passback postulated here cannot simply be interpreted as a
model in which every round of polymerase passage shifts the
histone octamer upstream by one position (,165 bp). In
Figure 9. Ancestral H3 retention and histone modification
patterns. (A) Scatterplot of previously measured H3K79me3  levels
averaged over the 59 CDS of genes versus the median HA/T7 for the 59
1 kb of each gene. (B) As in (A), for K79me3 averaged over mid-CDS of
Transgenerational Histone Retention in Yeast
PLoS Biology | www.plosbiology.org13June 2011 | Volume 9 | Issue 6 | e1001075
Figure 7A, our estimates of passback per cell cycle have a mean of
,90 bp at TFIID-dominated genes, less than the spacing
between adjacent nucleosomes. If taken literally, these values
would be difficult to reconcile with the observation that the
majority of yeast nucleosomes are well positioned . Instead,
we interpret the passback values in terms of probability that an
octamer will be passed back in a given cell cycle in each cell—a
passback value of ,80 bp suggests that there is a 50% chance
that histones on a given gene will be shifted back one position
towards the 59 end in a single cell cycle. Physically, we imagine
that polymerase passage results in relatively short retrograde
movement of H3/H4, which then have some probability of
returning to their original position, and some probability of
shifting to a new upstream location.
Our results show a surprising pattern of ancestral histone
retention in yeast, with old histone proteins accumulating near
the 59 ends of genes—the histone proteins located at the +3
nucleosome are the oldest histone proteins over a typical yeast
gene. These data are best explained by a model in which H3/H4
proteins shuttle from 39 to 59 over coding regions over time, with
eventual loss of old histone proteins when they are eventually
moved into the +1 and +2 nucleosome positions.
Maternal Histone Spreading During Replication
The process of genomic replication is enormously disruptive to
chromatin structure, as the melting of the DNA double helix is
accompanied by histone dissociation from the genome [4–7].
Thus, understanding where maternal histones re-associate relative
to the locus from which they were evicted is a key constraint for
understanding the potential of chromatin as an epigenetic
information carrier. The ideal experiment for measuring this
would be to epitope tag the histones at one specific locus (e.g., the
+5 nucleosome over BUD3) in a large population of yeast, allow
replication to proceed, and measure the new locations of the
tagged histones. Despite numerous attempts, this type of tagging
has proven technically intractable to date. Here we measure
instead the bulk distribution of ancestral histones. Importantly, this
still provides information on locus-specific histone behavior—as
turnover rates are not homogeneous across the genome, even
before we release yeast into the cell cycle the landscape of H3-HA
exhibits variability (Figure 4, see generation 0), and so in effect
only a subset of ancestral locations are epitope-tagged before
release. This enables us to infer the dynamic behavior of histone
proteins during replication via analysis of the evolution of the H3-
HA distributions over time.
Two observations provide an intuition regarding the effects of
replication on histone locations. First, ancestral histone retention
exhibits the expected anticorrelation with replication-independent
turnover (Figure 3A). However, old histones are more efficiently
retained at cold (low turnover) loci that occur in long cold
domains, whereas short domains of cold nucleosomes lose
ancestral histones over time. This observation is inconsistent with
two extreme models for histone behavior during replication—if
old histones were to completely dissociate from the genome during
replication and randomly re-associate with the genome, then
ancestral histone retention should precisely recapitulate turnover
measurements. Conversely, if old histones were to reassociate
precisely with their original locations, then ancestral retention
should essentially integrate turnover for multiple generations.
Thus, some process that shuffles histone proteins locally must be
invoked along with turnover to shape the ancestral retention
In principle, the preferential retention of old histones on longer
genes could simply result from passback—shorter genes will more
quickly have all of their histones passed ‘‘over a cliff’’ at the 59 end.
However, we find relatively static 59/39 gradients of old histone
retention over time (Figure 4). While it is the case that H3-HA
domains gradually shorten over time as predicted by the model
that passback results in old histones being moved to promoters
where they are replaced (unpublished data), this effect is subtle and
is quantitatively much less dramatic than predicted from passback
alone. This leads to the second intuition regarding histone
spreading during replication. Many examples exist for relatively
static gradients in biological systems being established via a
combination of directional active transport coupled with passive
diffusion. Most relevant in our opinion is the ‘‘pump leak’’ model
 for membrane ion gradients—active transport of ions across
membranes, coupled with a passive leak of ions back into the cell,
results in a static gradient. Here, we envision transcription-related
passback as the active transport mechanism, with spreading during
replication being somewhat analogous to the leak that results in a
steady gradient rather than a continuous 39 to 59 march of histone
We present a quantitative model that recapitulates our
experimental data with only three dynamic processes—turnover,
passback, and spreading. Locus-specific turnover rates were
previously measured  and are not fit by the model. Passback
is estimated for each gene separately, while spreading is a single
global parameter affecting all histones. Thus, our model has 4,811
free parameters, which are used to fit over 100,000 HA/T7 ratios.
This model does not overfit the data, and this can best be
appreciated by the fact that eliminating a single parameter
(spreading) greatly diminishes the agreement between model and
Using this model, we estimate that maternal histones spread
little (,1–2 nucleosomes) during replication. This value has not
been measured before but is consistent with several related
observations. First, electron microscopy studies on replicating
chromatin show a stretch of ,650–1,100 bp of nucleosome-free
DNA surrounding replication forks [35,36], consistent with
histone movement of 6400 bp we estimate here. Second, histone
proteins are retained in cis during in vitro replication even in the
presence of competitor DNA [64–67], indicating that histones do
not freely diffuse away from replication forks but likely are
retained locally. Finally, we previously observed that upon gene
repression, loss of the active chromatin mark H3K4me3 occurs
during S phase, but at very highly methylated nucleosomes
H3K4me3 does not return to baseline levels immediately, with
methylation levels falling little more than the 2-fold predicted by a
dilution-based mechanism (see Figure S5 in ). This final result
indicates that ‘‘overmethylated’’ old histone proteins are retained
near their original location, since extensive spreading of old
histone proteins would enable a greater than 2-fold drop in
methylation levels during S phase.
Together, these results support the prospect of chromatin as a
‘‘sloppy’’ epigenetic information carrier (‘‘sloppy’’ in the sense that
some spreading of histones will preclude mononucleosome-
resolution information passage) , even if chromatin-based
inheritance occurs infrequently . Thus, chromatin states are
unlikely to be inherited with mononucleosome precision, a view
consistent with the fact that most or all proposed epigenetic
chromatin domains are associated with long (.1 kb) blocks of
histone modifications such as H3K9me3 or deacetylated H4K16
(reviewed in ).
To further investigate the mechanisms underlying the patterns
of ancestral H3 retention, we assessed HA/T7 ratios at target
Transgenerational Histone Retention in Yeast
PLoS Biology | www.plosbiology.org14 June 2011 | Volume 9 | Issue 6 | e1001075
genes in 12 mutants and further characterized HA/T7 genome
wide for three of these mutants. Interestingly, a number of histone
modifying factors, including Swd1, Swd3, Rtt109, Nhp6, and
Set2, had either no effect or subtle effects (e.g., Rtt109) on the 59
accumulation of old histones at our target genes (Figure S8). These
results suggest either that these mutants will have subtle global
effects on ancestral H3 retention or that they have more localized
roles that do not extend to the two target genes on which we
The three mutants we characterized at full genome coverage
each had a distinct effect on ancestral H3 retention. Most
dramatically, loss of the H4 N-terminal tail abolished the 59
accumulation of ancestral histones—while the H4 tail deletion
results are complicated by the retention of wild-type H3/H4 in
this strain, the fact that similar results were obtained with clean
H4K5,12R mutants (Figure S8B) provides independent support
for observations obtained with the H4 tail deletion. The
mechanistic basis for the loss of 59 H3-HA retention is unknown
to us—a flat HA/T7 profile is of course consistent with complete
loss of passback. Alternatively, the observed profile in this mutant
could be consistent with complete shuffling of maternal histones
every generation, which as described above would be expected to
more closely recapitulate a turnover-dominated profile. Impor-
tantly, loss of the H4 tail also affects H3/H4 turnover—in
Figure 7D, the increased HA/T7 ratio at the 59 ends of SAGA-
dominated genes suggests a decrease in histone turnover in this
mutant, and we have independently confirmed a decrease in
replication-independent turnover in this mutant (F.v.L., manu-
script in preparation). Analysis of Pol2 ChIP in H4 tail deletions
shows that the effects of the H4 tail do not simply reflect altered
transcription but instead reflect a change in the relationship
between RNA Polymerase and histone dynamics over genes in this
mutant (Figure S14).
We also observe a similar, albeit muted, effect of Topoisomerase
I on the 59 accumulation of ancestral histones. Interestingly, loss of
both topoisomerase I and II affects nucleosome occupancy and
dynamics in S. pombe, indicating that topoisomerases play key roles
in histone dynamics . Here, we find that top1D mutants exhibit
diminished 59 accumulation of ancestral H3 and that this effect is
stronger at longer genes than at shorter genes. As RNA
polymerase passage will cause greater changes in supercoiling
over longer genes, the preferential effects of Top1 on longer genes
is consistent with the observation in S. pombe that topoisomerase
mutants show evidence of stalled or slowed RNA polymerase over
longer genes . In addition to its role in transcription,
topoisomerase I plays a key role in replication . We note that
the profile of top1D mutants here most closely mimics the
predictions of our analytical model with both passback and
spreading being compromised, but since neither of these is likely to
be completely eliminated in top1D mutants, more detailed kinetic
analyses will be required to make a quantitative statement about
the role of Top1 in replication-related movement of histones.
Finally, we assessed the role of the histone chaperone CAF-1 in
ancestral H3 retention. To our surprise, we found that H3-HA
exhibited even stronger 59 accumulation in this mutant, with the
59 peak of HA/T7 occurring closer to the +1 or +2 nucleosome
(compared to the +3 peak location for wild-type strains). This
result most closely matches the predictions of a model in which
H3/H4 turnover has been slowed without loss of passback or
spreading (Figure 5E). We recently tested this prediction using an
alternative system for measuring replication-independent turnover
(pGAL-driven Flag-H3) and confirmed the prediction that caf
mutants affect replication-independent histone replacement .
As CAF-1 and the Hir complex are known to complement one
another in yeast, we predict that a caf hir double mutant would be
necessary to uncover effects of replication-coupled spreading.
Unfortunately, since both hir and asf mutants are lethal in our
strain background, this prediction cannot be tested at present.
Taken together, our results provide a surprising view of histone
dynamics over multiple generations, with 59 accumulation of
ancestral histone proteins over coding regions and little evidence
for preferential histone retention at epigenetically regulated loci
such as subtelomeric genes. One unanticipated implication of this
observation is that 39 histone marks are expected to move towards
the 59 ends of genes over time, thereby shaping histone
modification profiles (as we document in Figure 9 and Figure
S17). This potentially necessitates mechanisms for erasure of these
inappropriate marks in order to maintain accurate encoding of
gene polarity. However, we note that active erasure of H3K4me3
after gene repression occurs most efficiently at 59 ends of genes,
whereas nucleosomes over coding regions mostly lose H3K4me3
by passive dilution (, see Figure S9). If other old histone marks
are not erased over coding regions, then we speculate that the
accumulation of old histone proteins at the +3 nucleosome could
potentially provide a mechanism by which a gene’s transcriptional
history could be integrated to play a role in regulation of the
transition from transcriptional initiation to elongation.
Most importantly, we find that old histones do not re-associate
with daughter genomes at precisely the locus from which they
dissociated. Thus, any inheritance of chromatin states must occur
at the scale of ,5–10 nucleosome domains rather than at single
nucleosome resolution. These results therefore constrain the
maximum amount of information theoretically carried by
chromatin between generations. It will be of great interest in
future studies to identify mutants that affect histone movement
during replication and to measure their effects on the stability of
epigenetic inheritance and to measure how maternal histone
incorporation differs between leading and lagging strand daughter
Materials and Methods
Yeast Strains and Growth Conditions
For tag switch experiments, yeast cells were grown overnight in
YPD in the presence of Hygromycin B (200 mg/mL, Invitrogen).
The cells were then diluted 1:10 into fresh YPD and incubated for
30–36 h. Recombination was induced by the addition of 1 mM b-
estradiol (E-8875, Sigma-Aldrich). Subsequently, cells were diluted
1:25 in fresh YPD media to release the cells back into the cell cycle
and kept in log phase by 1:2 dilutions into fresh media after each
population doubling. Samples were taken after 1, 2, 3, and 6 cell
divisions or after 5 h of G2/M arrest. The number of population
doublings was determined by microscopy and OD. G2/M arrest
was induced by addition of 15 mg/ml Nocodazole (Sigma-Aldrich)
and confirmed by FACS analysis. Strains are listed in Table S3.
Gene deletion mutants isogenic to strains NKI2048, NKI2148, and
NKI2048 were made by homologous recombination using KanMX
and/or NatMX selection markers. Gene deletion mutants isogenic
to NKI4128 were made by crossing NKI4114 with gene deletion
mutants from the MATa yeast knock-out collection using Synthetic
Genetic Array methods. Histone mutants were made by transfor-
mation of strain NKI2148 with a HHF2-HHT2 CEN plasmid
(pMP9), subsequent deletion of the tagged HHF2-HHT2 locus,
followed by transformation with a PCR fragment encoding wild-
typeormutated HHF2incombination with tagged HHT2. Deletion
of the wild-type locus was confirmed in H4K5,12R mutants,
Transgenerational Histone Retention in Yeast
PLoS Biology | www.plosbiology.org 15June 2011 | Volume 9 | Issue 6 | e1001075
whereas all surviving H4 taildeletion mutants retained a copy of the
wild-type HHT2-HHF2 locus.
Chromatin Immunoprecipitation (ChIP)
ChIP was performed as described previously [13,70] with the
following modifications. All steps were done at 4uC unless
otherwise indicated. Following cell lysis by bead beating the
insoluble chromatin of 16109cells was washed, resuspended in
400 ml FA lysis buffer (50 mM HEPES-KOH [pH 7.6], 150 mM
NaCl, 1 mM EDTA, 1% Triton X-100, 0.1% sodium deoxycho-
late), and sheared using a Bioruptor (Diagenode) for 6 min with
30 s intervals at high. The soluble fraction was diluted 3-fold in
buffer 15 mM Tris-HCl pH 7.4, 50 mM NaCl, 1.5 mM CaCl2,
5 mM b-meracptoethanol, 5 mM MgCl2, after which 25 units of
micrococcal nuclease (Worthington) were added. The digestion
reaction was incubated 20 min at 37uC and stopped by the
addition of 10 mM EDTA and 10 mM EGTA; tubes were placed
on ice. The majority of obtained fragments was around 150 bp, as
determined on a 2% TAE agarose gel stained with ethidium
bromide. The isolated chromatin of the equivalent of 36108cells
was immunoprecipitated overnight at 4uC using magnetic
Dynabeads (Invitrogen), which were previously incubated with
antibody O/N at 4uC.
ChIP DNA was quantified by real-time quantitative PCR using
the SYBR Green PCR Master Mix (Applied Biosystems) and the
ABI PRISM 7500. An input sample was used to make a standard
curve, which was then used to calculate the IP samples, all
performed in the 7500 fast system software. Primers used for
qPCR are listed in Table S4.
Linear amplification of DNA.
with a starting amount of up to 75 ng for ChIP samples, using the
DNA linear amplification method described previously .
the linear amplification were used to label probe via the amino-
allyl method as described on www.microarrays.org. Labeled
probes were hybridized onto a yeast tiled oligonucleotide
microarray  at 65uC for 16 h and washed as described on
www.microarrays.org. The arrays were scanned at 5 micron
resolution with an Axon Laboratories GenePix 4000B scanner
running GenePix 5.1. Image analysis and data normalization were
performed as previously described .
The samples were amplified,
3 mg of aRNA produced from
Deep Sequencing Library Construction
ChIP DNA was treated with CIP (calf alkaline phosphatase
NEB; in 16 NEB buffer 3, 0.25 U/ml CIP; 45 min at 37uC,
reaction clean up with Qiagen MinElute spin columns). 20–150 ng
of CIP treated ChIP DNA fragments were blunt ended and
phosphorylated with the EPICENTRE End-it-Repair kit (16
buffer, 0.25 mM dNTPs,1 mM ATP, 1 ml/50 ml reaction of
Enzyme mix) for 1 h at RT and cleaned up with Qiagen
MinElute spin columns. Adenosine nucleotide overhangs were
added using EPICENTRE exo-Klenow for 45 min at RT (with
0.2 mM dATP). Illumina genome sequencing adaptors were then
ligated using the EPICENTRE Fast-Link ligation kit: 11.5 ml A
tailed DNA eluted from a MinElute column was mixed with 1.5 ml
106 ligation buffer, 0.75 ml 10 mM ATP, 0.5 ml Illumina DNA
adaptors, and 1 ml Ligase. The reaction was incubated for 1 h at
RT and subsequently supplemented with 7.5 ml water, 1 ml 106
buffer, 0.5 ml 10 mM ATP, and 1 ml ligase, and incubated
overnight at 16uC.
The ligation reaction was cleaned up with MinElute columns
(with an additional wash step to eliminate all the excess adaptors)
and the adaptor ligated fragments were amplified by PCR as
follows: 0.5 ml of each Illumina genomic DNA sequencing primers,
10 ml 106Pfx buffer 3 ml 10 mM dNTPs, 2 ml 50 mM MgSO4,
and 1 ml Pfx DNA polymerase (Invitrogen) were added to 30 ml
DNA template in a 100 ml reaction. The cycling parameters were:
(1) 94uC, 29; (2) 94uC, 150; (3) 65uC, 19; (4) 68uC, 300; (5) repeat
from (2) 17 times; (6) 68uC, 59. The PCR product (200 to 300 bp
in size) was gel purified from a 2% TAE agarose gel using the
Freeze’N Squeeze columns (BioRad). Gel purified fragments were
finally precipitated with Sodium acetate and Ethanol and pellets
were resuspended (25 nM final concentration) in TE buffer and
sent for SOLEXA sequencing at the UMass Worcester core deep
RNA Pol II ChIP and Microarray Hybridization
Cells were grown as described above. Cell pellets (,109cells)
were flash frozen after formaldehyde crosslinking (1%) and kept at
280uC overnight. Frozen cell pellets were resuspended in 300 ml
cell braking buffer (100 mM Tris pH 7.9, 20% glycerol, 16Sigma
Protease inhibitors cocktail) and cell walls were broken down by
bead beating using 400 ml of 0.5 mm zirconia/silica beads
(BioSpec Products) in the BioSpec Mini-BeadBeater Model 8
three times for 1 min with 1 min pauses in between. Cell pellets
(5 min max speed spin in refrigerated microcentrifuge) were then
washed once and resuspended in 800 ml FA lysis buffer (with 16
Sigma Protease inhibitors cocktail). Chromatin was sheared by
sonication in a cup sonicator (Branson, 50% pulse at strength 7 for
3.5 min) to 250–400 bp fragments.
The sheared chromatin suspension was pre-cleared with 100 ml
Protein A-agarose slurry (IPA 400 HC RepliGen) at 4uC for 1 h.
100 ml of the pre-cleared solution was saved for the ChIP input
sample and 7 ml of RNA Pol II antibody (abcam ab81859, lot #:
933570 and GR6094-1) was added to the rest and incubated
overnight at 4uC with rotation. ChIP DNA isolation and
amplification by TLAD was done as described previously .
2.5 mg of aRNA produced from the linear amplification were
used to label probes via the amino-allyl method as described on
www.microarrays.org. Labeled probes were hybridized onto a
4X44K yeast whole genome array (Agilent) at 65uC for 16 h. The
arrays were scanned with the Agilent microarray scanner.
Data are downloadable at http://www.
umassmed.edu/bmp/faculty/rando.cfm and have been deposited
in GEO (Accession # GSE28269).
Raw sequencing data of HA and T7 libraries after
the tag swap before release from arrest (0 generations), and at 1, 3,
and 6 generations after release were uniquely mapped to the S.
aggregated HA and T7 sequencing of the 3 generation sample
using Template-Filtering . For each nucleosome, we counted
the number of supporting reads for each sample separately and
calculated the ratio of HA reads to T7 reads at each nucleosome
for each time point. Note, for the H4tailD data the T7 data quality
was poor, so we used wild-type T7 sequences for this comparison.
For aggregated analyses such as those shown in Figure 2E or
Figure 3, we calculated the median of the Log2 HA/T7 over the
1 kb starting at a given gene’s transcription start site to provide a
summary retention score per gene.
Described in Text S1.
nonselective media and onto media selecting for the HA tag
(linked to Hygro), before (t=0) and after (t=o/n) inducing
Recombination efficiency. Yeast were plated onto
Transgenerational Histone Retention in Yeast
PLoS Biology | www.plosbiology.org 16 June 2011 | Volume 9 | Issue 6 | e1001075
recombination. Roughly 2% of yeast fail to swap out the HA-
ratios (Log2) for individual nucleosomes are scatterplotted as
indicated, showing good correlation but a slope ,1 (red line),
consistent with the background of nonswitching cells observed in
HA/T7 at 3 and 6 generations after release. HA/T7
data (3 generations) for SPA2 (A) and BUD3 (C). (B, D) qPCR
shown for the 59 and 39 ends of SPA2 (B) and BUD3 (D) at the
indicated number of generations after tag-swap and release.
Midlog refers to samples taken 3 h after the tag-swap was induced
in exponentially growing cells that had not undergone a recent
arrest. Note that only a fraction of all the cells had recombined out
the old tag during the 3 h. qPCR amplicon locations are indicated
under the gene annotation.
Validation of target genes. (A, C) Deep sequencing
ratios at 3 generations after release are shown as a heatmap for all
genes, aligned by transcription start site (TSS) and clustered (K
means, K=5). Selected Gene Ontology (GO) enrichments for the
various clusters are indicated to the right of the clusters.
K means clustering of HA/T7 ratios. Log(2) HA/T7
Average profiles for the 5 clusters from Figure S4 are plotted
relative to TSS-aligned coding regions. (B) Replication-indepen-
dent turnover (Z score, ) was averaged for 59 CDS, mid-CDS,
and 39 CDS for all genes in each cluster. Note that Cluster 2,
which exhibits a somewhat 39-shifted peak of HA/T7 relative to
Clusters 3–5 (see A), consists of genes with relatively high 59
turnover, which presumably explains the downstream location of
the HA/T7 peak in this cluster.
Histone retention anticorrelates with turnover. (A)
frequency. (A–B) As in Figure 2B–C. (C) As in (B), but using
‘‘transcription frequency’’ defined in Holstege et al.  rather
than Pol2 ChIP. (D) As in Figure 2F, but using Holstege et al. data
rather than Pol2 ChIP data.
Histone retention anticorrelates with transcription
over the 59 1 kb of all genes is plotted versus distance from the
closest telomere, with an 80 gene running window average shown
in red. No specific enrichment of H3-HA is observed near
telomeres. Similar results are found for repetitive subtelomeric
genes (unpublished data).
H3 retention at subtelomeric genes. Median HA/T7
ratio at SPA2 or BUD3 were measured by q-PCR for the various
mutants 3 generations after release or after one round of replication
arrestedinG2/M.59/39ratio relative towild-typelevelis plotted on
the y-axis. Mutants are as indicated, with swd1, swd3 referring to an
average of single replicates with each individual mutant and scc1
referring to one experiment using a pGAL1-SCC1allele that wasshut
off by release into glucose media after the tag switch (leading to a
G2/M arrest). Average of mutant/wt, 6 S.E.M. (n=2). Swd1 and
Swd3 are components of the Set1 complex, which methylates
H3K4. Set2 is the H3K36 methylase and Scc1 is part of the cohesin
complex. (B)59/39 ratio atSPA2 orBUD3 were measuredbyq-PCR
for the various mutants after one round of replication arrested in
Mutant analysis of ancestral H3 retention. (A) 59/39
G2/M. 59/39 ratio is plotted on the y-axis with wild-type set to 1.
Average 6 S.E.M. (n=2). H4K5,12R is a mutant in which two of
the acetylatable lysines of the H4 tail have been replaced by
arginine, mimicking the unacetylated state. Rtt109 is a histone
acetyltransferase that binds to Asf1 and acetylates new histone H3
on K56 . Nhp6a/b are non-essential HMGB proteins  that
arerequired for FACT activity. The FACT core subunits Spt16 and
Pob3 are essential, precluding us from testing their functions in
ancestral histone inheritance. (C) As in (A–B), for the indicated
PCNA (POL30) point mutants. Average of mutant/wt, 6 S.E.M.
genes were normalized to a length of 1, and genes are ordered by
Pol2 ChIP. Log2 HA/T7 ratios are shown as a heatmap. (B)
Running window average of data from (A). Note the 59 shift of the
downstream edge of the HA/T7 peak with increasing transcrip-
tion rates. (C) Pol2 ChIP for genes as ordered in (A–B). (D)
Averages for all length-normalized genes grouped into 6 bins of
HA retention on length-normalized genes. (A) All
absence of nutrient stress. (A) Yeast carrying the HA/T7
recombination cassette were grown continuously in YPD, then
were treated with b-estradiol for 6 h to induce recombination.
HA and T7 ChIPs were carried out after 6 h and deep
sequenced, and normalized HA/T7 ratios were calculated. Here,
genes are ordered in 5 clusters as in Figure S4. (B) Averaged data
for cells arrested, switched, and released for 3 generations (‘‘3
gen’’) are shown alongside data from the midlog switch. Note that
59 accumulation occurs in both conditions but to a lesser extent in
the midlog swap. This is an expected result of the heterogeneity
of switch timing in midlog cells—only 65% of yeast have
completed recombination after 3 h, with 85% complete by 6 h
(unpublished data), meaning that the midlog switch represents a
mixture of cells that have recently swapped tags with those that
swapped tags ,1–3 generations prior. (C) As in (B), for
intermediate and long genes.
59 accumulation of ancestral histones occurs in the
H3 retention patterns. Model predictions (red lines) and data (blue
lines) for HA/T7 ratios at 1, 3, and 6 generations after tag swap
are shown for 1–2 kb genes (A) and .2 kb genes (B). The model
performs better on longer genes than on short genes.
A quantitative model accurately captures ancestral
ed passback parameters for each gene were compared to Pol2
ChIP values for each gene. Scatterplot is colored by density of
points—red indicates greater density of points. White line indicates
linear fit to dataset, R=0.12.
Passback correlates with transcription rate. Estimat-
model. Genes are ordered by length, and the difference between
model predictions for 3 generations and actual data are shown as a
heatmap—yellow indicates the model predicts excessive old
nucleosome loss, or lower HA/T7 ratios than measured. Notably,
the +N nucleosome is universally predicted to lose more H3-HA
than is measured. This is almost certainly a consequence of the fact
that our model considers all genes in isolation—there is no way for
histones to spread onto the 39 end of a gene from adjacent
genomic loci in this model, although this likely occurs in vivo.
Short genes and 39 ends are poorly predicted by the
Transgenerational Histone Retention in Yeast
PLoS Biology | www.plosbiology.org17June 2011 | Volume 9 | Issue 6 | e1001075
relationship between transcription and ancestral H3 retention. (A)
Tag swap strains carrying an H4 N-terminal tail deletion were
processed as in Figure 2B. Genes are ordered by the median HA/
T7 over the 59 1 kb. (B) Pol2 ChIP was carried out in the H4 tail
deletion strain 2 generations after release from arrest. Data here
show an 80 gene running window average of Pol2 ChIP level per
gene. (C) As in Figure 2D. Genes with high HA/T7 ratios in the
H4 tail deletion mutant actually tend to be slightly more enriched
for Pol2 than those with low HA/T7 levels, the opposite of what is
seen in wild-type (although it is important to note that the
correlation with Pol2 levels in this mutant is very weak—note that
the scale bar for Pol2 ChIP here ranges from 20.1 to 0.1, whereas
the scale bar in Figure 2D ranges from 20.2 to 1). Thus, we can
conclude with confidence that the effects of the H4 tail deletion do
not simply result from extensive transcription reprogramming in
these mutants, since the relationship between Pol2 and H3-HA
retention qualitatively changes in this mutant.
The H4 N-terminal tail qualitatively changes the
patterns. (A) Data for the 20% earliest and 20% latest-replicating
 genes is averaged as indicated. (B) As in (A), with gene lengths
normalized to one.
Replication time has subtle effects of ancestral H3
H3. Averaged data for clusters 4 and 5 (Figure S4) are shown for
wild-type and cac1D.
CAF-1 mutation affects far-59 end levels of ancestral
patterns. Modification levels  compared to ancestral H3
retention. For each modification, genes were grouped into high,
middle, and low HA/T7 (based on the 59-most 1 kb median
HA/T7), and for each group of genes modifications were
averaged for 5-CDS (‘‘5’’), mid-CDS (‘‘m’’), and 39-CDS (‘‘3’’)
as previously described [15,19]. Groupings are indicated for
H3K9ac and for H3K79me3 and are the same for the other
Ancestral H3 retention and histone modification
landscapes. Schematic for retrograde histone movement in
shaping histone modification landscapes. Initial 59 (green) and 39
Role for histone movement in shaping modification
(purple) histone modification states could, in the absence of erasing
enzymes, eventually give rise to skewed distributions via retrograde
motion of old histones bearing 39 modifications such as
H3K36me3 (purple). Importantly, after a few cell divisions old
histones on average constitute only a minor fraction of all histones
at any given locus (e.g., see Figure S3B, D). Modifications on
ancestral histones will therefore make subtle contributions to
overall average modification patterns.
genes is shown at varying distances with respect to the TSS. Genes
are sorted according to the median HA/T7 over the 59 1 kb.
HA/T7 ratios for 3 generations. Log2(HA/T7) for all
passback values. A variety of gene sets were searched for
enrichment of relatively high or low values for model estimates
of lateral nucleosome movement. Negative KS values indicate
large 39 to 59 lateral movements; positive values indicate the
Enrichments of genesets for high or low retrograde
Strain list. Genotypes of strains used in this study.
Primers used. Primer sequences.
dynamics. Detailed description of model shown in Figures 5 and 6.
Quantitative model for multigenerational histone
We would like to thank P. Kaufman and K. Ahmad for critical reading of
the manuscript and helpful discussions, P. Kaufman and P. Burgers for
PCNA vectors, and E. Reinen for help with strain constructions.
The author(s) have made the following declarations about their
contributions: Conceived and designed the experiments: FvL OJR KFV
MRL. Performed the experiments: KFV MRL. Analyzed the data: MRL
AW NF OJR. Contributed reagents/materials/analysis tools: TvW. Wrote
the paper: MRL KFV AW NF OJR FvL.
1. Ptashne M (2007) On the use of the word ‘epigenetic’. Curr Biol 17:
2. Ringrose L, Paro R (2004) Epigenetic regulation of cellular memory by the
Polycomb and Trithorax group proteins. Annu Rev Genet 38: 413–443.
3. Rusche LN, Kirchmaier AL, Rine J (2003) The establishment, inheritance, and
function of silenced chromatin in Saccharomyces cerevisiae. Annu Rev Biochem
4. Annunziato AT (2005) Split decision: what happens to nucleosomes during DNA
replication? J Biol Chem 280: 12065–12068.
5. Groth A (2009) Replicating chromatin: a tale of histones. Biochem Cell Biol 87:
6. Groth A, Rocha W, Verreault A, Almouzni G (2007) Chromatin challenges
during DNA replication and repair. Cell 128: 721–733.
7. Kaufman PD, Rando OJ (2010) Chromatin as a potential carrier of heritable
information. Curr Opin Cell Biol.
8. Jackson V, Chalkley R (1985) Histone segregation on replicating chromatin.
Biochemistry 24: 6930–6938.
9. Jackson V (1987) Deposition of newly synthesized histones: new histones H2A
and H2B do not deposit in the same nucleosome with new histones H3 and H4.
Biochemistry 26: 2315–2325.
10. Jackson V (1988) Deposition of newly synthesized histones: hybrid nucleosomes
are not tandemly arranged on daughter DNA strands. Biochemistry 27:
11. Corpet A, Almouzni G (2009) Making copies of chromatin: the challenge of
nucleosomal organization and epigenetic information. Trends Cell Biol 19:
12. Dodd IB, Micheelsen MA, Sneppen K, Thon G (2007) Theoretical analysis of
epigenetic cell memory by nucleosome modification. Cell 129: 813–822.
13. Verzijlbergen KF, Menendez-Benito V, van Welsem T, van Deventer SJ,
Lindstrom DL, et al. (2010) Recombination-induced tag exchange to track old
and new proteins. Proc Natl Acad Sci U S A 107: 64–68.
14. Schermer UJ, Korber P, Horz W (2005) Histones are incorporated in trans
during reassembly of the yeast PHO5 promoter. Mol Cell 19: 279–285.
15. Dion MF, Kaplan T, Kim M, Buratowski S, Friedman N, et al. (2007) Dynamics
of replication-independent histone turnover in budding yeast. Science 315:
16. Jamai A, Imoberdorf RM, Strubin M (2007) Continuous histone H2B and
transcription-dependent histone H3 exchange in yeast cells outside of
replication. Mol Cell 25: 345–355.
17. Rufiange A, Jacques PE, Bhat W, Robert F, Nourani A (2007) Genome-wide
replication-independent histone H3 exchange occurs predominantly at promot-
ers and implicates H3 K56 acetylation and Asf1. Mol Cell 27: 393–405.
18. Au WC, Crisp MJ, DeLuca SZ, Rando OJ, Basrai MA (2008) Altered dosage
and mislocalization of histone H3 and Cse4p lead to chromosome loss in
Saccharomyces cerevisiae. Genetics 179: 263–275.
19. Liu CL, Kaplan T, Kim M, Buratowski S, Schreiber SL, et al. (2005) Single-
nucleosome mapping of histone modifications in S. cerevisiae. PLoS Biol 3:
20. Yuan GC, Liu YJ, Dion MF, Slack MD, Wu LF, et al. (2005) Genome-scale
identification of nucleosome positions in S. cerevisiae. Science 309: 626–630.
21. Mito Y, Henikoff JG, Henikoff S (2005) Genome-scale profiling of histone H3.3
replacement patterns. Nat Genet 37: 1090–1097.
Transgenerational Histone Retention in Yeast
PLoS Biology | www.plosbiology.org18 June 2011 | Volume 9 | Issue 6 | e1001075
22. Thiriet C, Hayes JJ (2005) Replication-independent core histone dynamics at Download full-text
transcriptionally active loci in vivo. Genes Dev 19: 677–682.
23. Deal RB, Henikoff JG, Henikoff S (2010) Genome-wide kinetics of nucleosome
turnover determined by metabolic labeling of histones. Science 328: 1161–1164.
24. Gottschling DE, Aparicio OM, Billington BL, Zakian VA (1990) Position effect
at S. cerevisiae telomeres: reversible repression of Pol II transcription. Cell 63:
25. Pillus L, Rine J (1989) Epigenetic inheritance of transcriptional states in S.
cerevisiae. Cell 59: 637–647.
26. Kaplan T, Liu CL, Erkmann JA, Holik J, Grunstein M, et al. (2008) Cell cycle-
and chaperone-mediated regulation of H3K56ac incorporation in yeast. PLoS
Genet 4: e1000270. doi:10.1371/journal.pgen.1000270.
27. Glynn EF, Megee PC, Yu HG, Mistrot C, Unal E, et al. (2004) Genome-wide
mapping of the cohesin complex in the yeast Saccharomyces cerevisiae. PLoS
Biol 2: E259. doi:10.1371/journal.pbio.0020259.
28. Lengronne A, Katou Y, Mori S, Yokobayashi S, Kelly GP, et al. (2004) Cohesin
relocation from sites of chromosomal loading to places of convergent
transcription. Nature 430: 573–578.
29. Kouzarides T (2007) Chromatin modifications and their function. Cell 128:
30. Studitsky VM, Kassavetis GA, Geiduschek EP, Felsenfeld G (1997) Mechanism
of transcription through the nucleosome by eukaryotic RNA polymerase.
Science 278: 1960–1963.
31. Studitsky VM, Clark DJ, Felsenfeld G (1994) A histone octamer can step around
a transcribing polymerase without leaving the template. Cell 76: 371–382.
32. Kulaeva OI, Gaykalova DA, Pestov NA, Golovastov VV, Vassylyev DG, et al.
(2009) Mechanism of chromatin remodeling and recovery during passage of
RNA polymerase II. Nat Struct Mol Biol 16: 1272–1278.
33. Hodges C, Bintu L, Lubkowska L, Kashlev M, Bustamante C (2009)
Nucleosomal fluctuations govern the transcription dynamics of RNA polymerase
II. Science 325: 626–628.
34. Weiner A, Hughes A, Yassour M, Rando OJ, Friedman N (2010) High-
resolution nucleosome mapping reveals transcription-dependent promoter
packaging. Genome Res 20: 90–100.
35. Gasser R, Koller T, Sogo JM (1996) The stability of nucleosomes at the
replication fork. J Mol Biol 258: 224–239.
36. Sogo JM, Stahl H, Koller T, Knippers R (1986) Structure of replicating simian
virus 40 minichromosomes. The replication fork, core histone segregation and
terminal structures. J Mol Biol 189: 189–204.
37. Wapinski I, Pfeffer A, Friedman N, Regev A (2007) Natural history and
evolutionary principles of gene duplication in fungi. Nature 449: 54–61.
38. Huisinga KL, Pugh BF (2004) A genome-wide housekeeping role for TFIID and
a highly regulated stress-related role for SAGA in Saccharomyces cerevisiae.
Mol Cell 13: 573–585.
39. Radman-Livaja M, Rando OJ (2010) Nucleosome positioning: how is it
established, and why does it matter? Dev Biol 339: 258–266.
40. Tirosh I, Barkai N, Verstrepen KJ (2009) Promoter architecture and the
evolvability of gene expression. J Biol 8: 95.
41. Jiang C, Pugh BF (2009) Nucleosome positioning and gene regulation: advances
through genomics. Nat Rev Genet 10: 161–172.
42. Matangkasombut O, Buratowski S (2003) Different sensitivities of bromodomain
factors 1 and 2 to histone H4 acetylation. Mol Cell 11: 353–363.
43. Yu C, Palumbo MJ, Lawrence CE, Morse RH (2006) Contribution of the
histone H3 and H4 amino termini to Gcn4p- and Gcn5p-mediated transcription
in yeast. J Biol Chem 281: 9755–9764.
44. Pommier Y (2006) Topoisomerase I inhibitors: camptothecins and beyond. Nat
Rev Cancer 6: 789–802.
45. Koster DA, Crut A, Shuman S, Bjornsti MA, Dekker NH (2010) Cellular
strategies for regulating DNA supercoiling: a single-molecule perspective. Cell
46. Fernandez-Beros ME, Tse-Dinh YC (1996) Vaccinia virus DNA topoisomerase I
preferentially removes positive supercoils from DNA. FEBS Lett 384: 265–268.
47. Goto T, Laipis P, Wang JC (1984) The purification and characterization of
DNA topoisomerases I and II of the yeast Saccharomyces cerevisiae. J Biol
Chem 259: 10422–10429.
48. Verreault A, Kaufman PD, Kobayashi R, Stillman B (1996) Nucleosome
assembly by a complex of CAF-1 and acetylated histones H3/H4. Cell 87:
49. Smith S, Stillman B (1989) Purification and characterization of CAF-I, a human
cell factor required for chromatin assembly during DNA replication in vitro. Cell
50. Lopes da Rosa J, Holik J, Green EM, Rando OJ, Kaufman PD (2010)
Overlapping regulation of CenH3 localization and Histone H3 turnover by
CAF-1 and HIR proteins in Saccharomyces cerevisiae. Genetics.
51. Sharp JA, Rizki G, Kaufman PD (2005) Regulation of histone deposition
proteins Asf1/Hir1 by multiple DNA damage checkpoint kinases in Saccharo-
myces cerevisiae. Genetics 171: 885–899.
52. Pokholok DK, Harbison CT, Levine S, Cole M, Hannett NM, et al. (2005)
Genome-wide map of nucleosome acetylation and methylation in yeast. Cell
53. Schulze JM, Jackson J, Nakanishi S, Gardner JM, Hentrich T, et al. (2009)
Linking cell cycle to histone modifications: SBF and H2B monoubiquitination
machinery and cell-cycle regulation of H3K79 dimethylation. Mol Cell 35:
54. Rando OJ, Chang HY (2009) Genome-wide views of chromatin structure. Annu
Rev Biochem 78: 245–271.
55. Frederiks F, Tzouros M, Oudgenoeg G, van Welsem T, Fornerod M, et al.
(2008) Nonprocessive methylation by Dot1 leads to functional redundancy of
histone H3K79 methylation states. Nat Struct Mol Biol 15: 550–557.
56. Gat-Viks I, Vingron M (2009) Evidence for gene-specific rather than
transcription rate-dependent histone H3 exchange in yeast coding regions.
PLoS Comput Biol 5: e1000282. doi:10.1371/journal.pcbi.1000282.
57. Probst AV, Dunleavy E, Almouzni G (2009) Epigenetic inheritance during the
cell cycle. Nat Rev Mol Cell Biol 10: 192–206.
58. Sedighi M, Sengupta AM (2007) Epigenetic chromatin silencing: bistability and
front propagation. Phys Biol 4: 246–255.
59. Goren A, Cedar H (2003) Replicating by the clock. Nat Rev Mol Cell Biol 4:
60. Durand-Dubief M, Persson J, Norman U, Hartsuiker E, Ekwall K (2010)
Topoisomerase I regulates open chromatin and controls gene expression in vivo.
Embo J 29: 2126–2134.
61. Kulaeva OI, Gaykalova DA, Studitsky VM (2007) Transcription through
chromatin by RNA polymerase II: histone displacement and exchange. Mutat
Res 618: 116–129.
62. Kulaeva OI, Hsieh FK, Studitsky VM (2010) RNA polymerase complexes
cooperate to relieve the nucleosomal barrier and evict histones. Proc Natl Acad
Sci U S A 107: 11325–11330.
63. Hoffmann EK (2001) The pump and leak steady-state concept with a variety of
regulated leak pathways. J Membr Biol 184: 321–330.
64. Bonne-Andrea C, Wong ML, Alberts BM (1990) In vitro replication through
nucleosomes without histone displacement. Nature 343: 719–726.
65. Randall SK, Kelly TJ (1992) The fate of parental nucleosomes during SV40
DNA replication. J Biol Chem 267: 14259–14265.
66. Krude T, Knippers R (1991) Transfer of nucleosomes from parental to
replicated chromatin. Mol Cell Biol 11: 6257–6267.
67. Gruss C, Wu J, Koller T, Sogo JM (1993) Disruption of the nucleosomes at the
replication fork. Embo J 12: 4533–4545.
68. Radman-Livaja M, Liu CL, Friedman N, Schreiber SL, Rando OJ (2010)
Replication and active demethylation represent partially overlapping mecha-
nisms for erasure of H3K4me3 in budding yeast. PLoS Genet 6: e1000837.
69. Talbert PB, Henikoff S (2006) Spreading of silent chromatin: inaction at a
distance. Nat Rev Genet 7: 793–803.
70. McConnell AD, Gelbart ME, Tsukiyama T (2004) Histone fold protein Dls1p is
required for Isw2-dependent chromatin remodeling in vivo. Mol Cell Biol 24:
71. Holstege FC, Jennings EG, Wyrick JJ, Lee TI, Hengartner CJ, et al. (1998)
Dissecting the regulatory circuitry of a eukaryotic genome. Cell 95: 717–728.
72. Tsubota T, Berndsen CE, Erkmann JA, Smith CL, Yang L, et al. (2007) Histone
H3-K56 acetylation is catalyzed by histone chaperone-dependent complexes.
73. Stillman DJ (2010) Nhp6: a small but powerful effector of chromatin structure in
Saccharomyces cerevisiae. Biochim Biophys Acta 1799: 175–180.
74. Raghuraman MK, Winzeler EA, Collingwood D, Hunt S, Wodicka L, et al.
(2001) Replication dynamics of the yeast genome. Science 294: 115–121.
75. Kim TS, Liu CL, Yassour M, Holik J, Friedman N, et al. (2010) RNA
Polymerase mapping during stress responses reveals widespread nonproductive
transcription in yeast. Genome Biol 11: R75.
Transgenerational Histone Retention in Yeast
PLoS Biology | www.plosbiology.org19 June 2011 | Volume 9 | Issue 6 | e1001075