Chromosomal domains of gene expression dysregulation in Down syndrome
Audrey Letourneau1, Federico Santoni1, Ximena Bonilla1, M.Reza Sailani1, David
Gonzalez2, Jop Kind3, Youssef Hibaoui4, Emilie Falconnet1, Maryline Gagnebin1,
Corinne Gehrig1, Anne Vannier1, Michel L. Guipponi1, Laurent Farinelli5, Daniel
Robyr1, Marco Garieri1, Eugenia Migliavacca1,6, Christelle Borel1, Anis Feki4, Yann
Herault7, Bas van Steensel3, John A. Stamatoyannopoulos8, Roderic Guigo2, Stylianos
E. Antonarakis1,9
1. Department of Genetic Medicine and Development, University of Geneva Medical
School, 1211 Geneva, Switzerland.
2. Center for Genomic Regulation, University Pompeu Fabra, Barcelona, Spain
3. Division of Molecular Biology, Netherlands Cancer Institute, 1066 CX
Amsterdam, The Netherlands.
4. Stem Cell Research Laboratory, Department of Obstetrics and Gynecology, Geneva
University Hospitals, Switzerland
5. FASTERIS SA, Plan-les-Ouates, Switzerland
6. Swiss Institute of Bioinfomatics, Switzerland.
7. AneuPath 21, IGBMC, Illkirch, France
8. Department of Genome Sciences, University of Washington, Seattle, USA
9. iGE3 Institute of Genetics and Genomics of Geneva, Switzerland.
1
ABSTRACT
Down syndrome (DS) results from trisomy 21 (T21) and is the most frequent genetic
cause of mental retardation. In order to assess the perturbations of gene expression in
trisomic cells, we studied the transcriptome of fibroblasts derived from a pair of
monozygotic twins discordant for T21. The use of these samples eliminates the
genomic variability and identifies changes related to the T21 per se. Surprisingly, we
found that the differential expression between the twins is organized in domains along
the chromosomes. We identified 237 of those large regions that are either up- or
downregulated in the trisomic twin. We found that those gene expression
dysregulation domains (GEDDs) can be partially defined by the expression level of
their gene content. Interestingly, the domains are well conserved in induced
pluripotent stem (iPS) cells derived from the twins’ fibroblasts. We also compared the
transcriptome of the Ts65Dn mouse model of DS and wild-type littermates.
Remarkably, we found that the expression changes are also organized in domains
along the mouse chromosomes and that those domains correlate with the syntenic
regions in human. Interestingly, the GEDDs correlate with the lamina-associated
domains (LADs) and replication domains already described in mammalian cells. The
investigation of the LADs position in the twins’ fibroblasts revealed that the overall
genome topology is not altered in trisomic cells. We thus examined the epigenetic
profile within the GEDDs. Interestingly, while the cytosine DNA methylation is not
dramatically changed between the twins, we found that the H3K4me3 profile of the
trisomic fibroblasts is modified and follows the GEDD pattern. All together, our
results suggest that the nuclear compartments of trisomic cells undergo modifications
of the chromatin environment influencing the overall transcriptome and that GEDDs
may therefore contribute to some T21 phenotypes.
2
INTRODUCTION
Down syndrome (DS) results from total or partial trisomy of human chromosome 21
(T21). It is the most frequent live-born aneuploidy affecting 1/750 newborns and a
leading cause of mental retardation. DS patients are characterized by a cognitive
impairment together with other manifestations such as muscle hypotonia, dysmorphic
features, Alzheimer disease neuropathology or congenital heart defects [1-3].
Importantly, the severity and the incidence of those phenotypes are highly variable
within the DS population [3]. Among the possible causes, the genetic (or epigenetic)
background of each individual may largely contribute to this phenotypic variability.
It is likely that most of the DS phenotypes are related to alteration of gene expression
due to the extra-copy of chromosome 21 (HSA21). According to the “gene dosage
effect” hypothesis, some DS features could be directly explained by the dosage
imbalance of individual genes on HSA21 [4-7]. Additionally, the phenotypes may
also be linked to a general disturbance of the cellular homeostasis due to the presence
of extra DNA material in the nucleus [5, 6, 8]. The understanding of these genomic
determinants constitutes one of the main goals in DS research.
Several DS mouse models have been created in order to mimic the gene expression
changes observed in humans. Those models are based on the duplication of syntenic
regions between HSA21 and mouse chromosomes (MMUs) 10, 16 and 17. Among
them, the Ts65Dn mouse model has been commonly used to study the molecular
mechanisms responsible for the DS features [9]. Those mice harbor a segmental
triplication of MMU16 and exhibit some of the DS phenotypes [10-13]. Other mouse
models are based on the triplication of single candidate genes such as SIM2 or
DYRK1A. In accordance with the gene dosage hypothesis, these models have
3
emphasized the role of individual HSA21 genes in specific phenotypes, in particular
the cognitive impairment [14-17].
Since DS is likely to be due to gene expression disturbances, the investigation of the
molecular mechanisms that underlie the phenotypic consequences requires the
understanding of the transcriptome differences in trisomic cells and tissues. Several
studies have explored the changes of gene expression between trisomic individuals
and control populations [18-24]. However, the natural gene expression variation
occurring in both normal and DS individuals [25-28] complicates the identification of
changes related to the T21 per se. In order to assess the perturbations of gene
expression in DS without the genetic variation among the samples, we studied a pair
of monozygotic (MZ) twins discordant for T21 [29]. For the first time, the use of
these samples eliminates the bias of genome variability and thus most of the
transcriptome differences observed are likely related to the supernumerary HSA21.
Interestingly, we have found that the differential gene expression between the
discordant twins is organized in domains along the chromosomes. We show that those
domains are conserved in the Ts65Dn mouse model and correlate with the previously
described lamina-associated domains and replication time domains. The study of
histone mark profiles confirmed the role of chromatin modifications in the gene
expression changes observed between the twins. All together, our results suggest a
new molecular mechanism in the regulation of gene expression in trisomic cells.
4
RESULTS
Differential expression is organized in chromosomal domains
We sought to investigate the gene expression changes in DS by comparing a pair of
MZ twins discordant for T21. We used mRNA-Sequencing to study the transcriptome
of fetal skin primary fibroblasts derived from both the trisomic (T1DS) and the
normal (T2N) twin. In total, 107-315 million 100bp paired-end reads were generated
from each sample and mapped against the reference genome using GEM [30]. All the
data were further confirmed using TopHat [31] (data not shown). The normalized
gene expression (RPKM, Reads Per Kilobase per Million) was then compared
between the twins. We first used EdgeR [32] to evaluate the differences of gene
expression between the samples and found nine genes significantly differentially
expressed between the twins (False discovery rate FDR<0.05). In total, 1816 genes
showed a modified expression of at least 4 fold in T1DS compared to T2N (737 with
a decreased expression and 1079 with an increased expression) (data available to the
reviewers, see Appendix 1). Next, we assessed the genome-wide differential
expression between the discordant twins by looking at the distribution of gene
expression fold changes along the chromosomes. Surprisingly, this comparison
revealed well-defined chromosomal domains composed of neighboring genes sharing
the same differential expression profiles. Indeed, in most of the chromosomes, large
regions of upregulated genes alternate with large downregulated domains (Figures 1a,
1b and S1). This observation suggests that the differential expression between T1DS
and T2N is not randomly organized but follows a specific pattern along the
chromosomes. We used a smoothing function to define more precisely the domain
borders and identified a total of 237 gene expression dysregulation domains (GEDDs)
in the discordant twins (Figure S2a and Table S1). Those domains vary greatly in size
5
(from 70kb to 114Mb, median size of 5.5Mb) and can contain up to 813 genes with a
median number of 46. An independent replicate experiment confirmed the existence
of the GEDDs in the discordant twins (Figure S2b). We also compared the gene
expression profiles of fibroblasts derived from a healthy pair of monozygotic twins
(MZ1/MZ2) as a control (Figure S2c). The absence of domains in this comparison
suggests that the domain organization observed in the discordant twins can be mainly
attributed to the supernumerary chromosome 21.
We then wanted to verify if this domain organization could be identified when
comparing T21 to normal samples from unrelated individuals. We hypothesized that
the natural genetic variability between unrelated individuals would mask the domains.
Therefore, we sequenced the mRNA from fibroblasts of 8 trisomic individuals and 8
sex-matched controls and averaged the expression values within each group in order
to reduce the influence of individual genetic backgrounds. We calculated the
expression fold change between the two groups but this comparison did not reveal
chromosomal domains (Figure S2d). We concluded that the natural gene expression
variation occurring between unrelated individuals is too strong and cancels the effects
observed with the discordant twins.
Weakly and highly expressed genes contribute differently to the domains
Next, we investigated the expression level of the genes within the domains. We
classified the genes according to low (0.1<rpkm<10), medium (10<rpkm<100) or
high (rpkm>100) expression level and looked at their position along the chromosomes
(Figures 2a and S3). For each category, we compared the expression fold change
distribution with the expected distribution given by the healthy MZ twins (Figure 2b).
Interestingly, we found that the highly expressed genes contribute substantially to the
6
downregulated domains in the discordant twins (p<2.2e-16 and p=1.4e-15 for medium
and high expression, respectively). On the contrary, the genes expressed at lower
levels are mainly associated to the upregulated regions (p=4.4e-04). All together,
these data show that the expression fold change between trisomic and normal cells is
organized in chromosomal domains and that those domains can be partially defined
by the expression level of their gene content.
GEDDs are conserved in induced pluripotent stem cells
We then wanted to verify in the domain organization observed in the twins’
fibroblasts could be maintained when we de-differentiate the cells into induced
pluripotent stem (iPS) cells. We derived iPS cells from each of the fibroblast lines by
using a lentiviral vector expressing the OCT4, SOX2, KLF4 and C-MYC genes
(Hibaoui et al, submitted). We performed mRNA-sequencing on both iPS lines and
compared the normalized gene expression values along the chromosomes, similarly to
the fibroblasts analysis. Interestingly, for most of the chromosomes, we found that the
differential expression patterns in iPS cells are highly similar to the ones observed in
the twins’ fibroblasts (Figures 3 and S4). These results suggest that the GEDDs are
conserved after dedifferentiation and that the iPS cells retain a specific memory of
their origin.
GEDDs are conserved in mouse Ts65Dn fibroblasts
In order to verify if the domain structure of gene expression dysregulation described
in the twins’ fibroblasts is conserved in other organisms, we performed a similar
transcriptome analysis in the partial trisomy 16 Ts65Dn mouse model of DS. We
analyzed the differential expression pattern between Ts65Dn mouse fibroblasts and
7
wild-type littermates. Importantly, we found that the expression fold change is also
organized in domains along the Ts65Dn mouse chromosomes (Figures 1c and S5).
The comparison of the largest syntenic blocks (>1Mb) between mouse and human
revealed that the domains and the associated gene expression fold changes are well
conserved between the discordant twins’ fibroblasts and the mouse fibroblasts and
that the direction of dysregulation is maintained between the two species (Figure 4a).
Importantly, this conservation seems to be independent of the genomic context since
the fragments can be located in different chromosomes in the mouse. Human and
mouse orthologous genes show rather similar expression fold changes when
comparing both experiments (Whole genome Spearman correlation 0.44) (Figure 4b,
upper panel). The same comparison using the healthy MZ twins did not show a
correlation between human and mouse fold changes (Whole genome Spearman
correlation -0.13) (Figure 4b, lower panel). Those results demonstrate that the GEDDs
observed between T1DS and T2N are conserved in the Ts65Dn mouse model for DS.
This remarkable conservation between human and mouse syntenic regions suggests a
functional role for those domains of co-regulated genes.
Correlation with previously described chromosomal domains
The domain organization of the mammalian chromosomes has been previously
reported with the identification of the lamina-associated domains (LADs) [33-35].
Those genomic regions are defined by their interaction with the nuclear lamina and
are characterized by repressive chromatin marks and low gene expression levels. We
looked at the correlation between those LADs and the GEDDs identified in the
discordant twins’ fibroblasts. We first compared the distribution of gene expression
fold changes within and outside the LADs. We observed that the genes located within
8
the LADs were in average overexpressed in the T1DS as opposed to the genes located
outside of those regions (in inter-LADs = iLADs) (p = 8.9e-16) (Figure 5a). A similar
significant shift was observed in the Ts65Dn fibroblasts when we compared the
mouse gene expression fold changes inside and outside the LADs described in mouse
[34] (p = 4.61e-07) (Figure 5b). The healthy MZ twins did not show any expression
differences between LADs and iLADs (p = 0.54) (Figure 5c). The LADs are known to
constitute a strong repressive environment and are mostly associated with a repression
of gene expression [33, 36-38]. In our experiment, the increased fold change in LADs
suggests a derepression of the genes in the LADs of trisomic cells. Therefore, we
hypothesized that the interaction of the genome with the nuclear lamina might be
disturbed in trisomic nuclei. We used the DamID method [39] to define the position
of the LADs in the twins’ fibroblasts. The lamin B1 interaction scores along the
chromosomes revealed identical profiles between T1DS and T2N (Pearson correlation
0.75, Hidden Markov modeling concordance of 94% => van Steensel’s lab: is there a
better way of showing this?) (Figures 5d, 5e and S6). We concluded that the overall
topology of LADs is not perturbed by the presence of an extra chromosome 21 in the
T1DS cells.
Interestingly, the LADs have been shown to correlate with another type of
chromosomal regions known as replication domains [40]. During the S-phase, the
DNA replication occurs in a specific temporal order. Some regions of the
chromosomes replicate first (early replication) whereas other regions always replicate
later (late replication). Thus, the chromosomes can be physically divided in time
zones defined as replication time domains [41-43]. Early replicating domains contain
mostly active genes and tend to localize centrally in the nucleus as opposed to late
replicating domains that mainly contain less active genes at the nuclear periphery [44-
9
46]. We assessed the correlation between the replication time domains described in
human fetal fibroblasts [47, 48] and our GEDDs. Most chromosomes show a striking
correlation between both profiles, suggesting that the gene expression is increased in
late replication domains and decreased in early replication domains of the T1DS cells
(Figures 6a and S7). These data suggest that the replication time domains of trisomic
fibroblasts undergo specific modifications leading to the alteration of gene expression.
Alternatively, the GEDDs may also be linked to a shift in the replication time itself in
the T1DS cells.
Modification of the chromatin environment in the GEDDs
Since we showed that the genome topology is not changed in T1DS fibroblasts, we
hypothesized that the alterations of gene expression might be related to epigenetic
modifications within the chromosomal domains of trisomic cells.
Thus, we first decided to explore the DNA methylation changes in T1DS compared to
T2N and to investigate a possible link with the GEDDs. We performed Reduced
Representation Bisulfite-Sequencing (RRBS) on DNA extracted from the twins’
fibroblasts and evaluated the methylation percentage for each CpG dinucleotide in the
genome. We then compared the methylation level of each gene between T1DS and
T2N by considering the CpGs included either in the gene body or in the promoter
region. The profiles of gene methylation differences along the chromosomes were
compared to the GEDDs. We observed a weak correlation between both patterns for
some chromosomes when using CpGs in the gene body (Figure S8a) or in the
promoter region (Figure S8b). However, this degree of concordance between
methylation and gene expression was not sufficient to elucidate the GEDDs. We
10
concluded that the alteration of gene expression cannot be fully explained by domainwide changes of cytosine methylation in the trisomic fibroblasts.
The cellular transcription activity can also be influenced by the modification of the
chromatin environment through changes of histone marks. Among them, H3K4me3
(histone 3 lysine 4 trimethylation) has been associated to active promoters and is
known to correlate positively with the gene expression [49]. Thus, we used chromatin
immunoprecipitation coupled to high throughput sequencing (ChIP-Seq) to compare
the level of H3K4me3 in the discordant twins. We first used HOMER to identify the
genomic regions enriched for H3K4me3 in T1DS and T2N. We then investigated the
differences of signal intensity between all the peaks common to both twins. Each of
those peaks was assigned to the closest gene in order to obtain one H3K4me3 fold
change value per gene and to establish chromosomal profiles comparable to the gene
expression dysregulation domains. Remarkably, this comparison revealed a striking
correlation between both patterns for all chromosomes except HSA13 and HSA18
(Figures 6b and S9). We explain the lack of correlation in those chromosomes by the
absence of a clear domain pattern due to the overall overexpression of the genes. We
concluded that the changes of gene expression between the trisomic and the normal
twins can be attributed to the modification of the chromatin environment and more
specifically to the modification of H3K4me3 in trisomic cells.
Interestingly, the H3K4me3 mark at active promoters has been associated to other
genomic features representing open chromatin. Among them, the DNase I
hypersensitivity has been widely studied to identify the sites of open chromatin in the
genome. We decided to explore the DNase I hypersensitivity profiles of both twins to
11
verify if the changes of chromatin structure in trisomic fibroblasts could be seen at the
DNase hypersensitivity level. DNase I HS data from Stamatoyannopoulos’ lab
12
DISCUSSION
The transcriptome analysis of MZ twins discordant for T21 highlighted the existence
of chromosomal domains of gene expression dysregulation between trisomic and
normal fibroblasts. We attributed this discovery to the use of samples from MZ twins
that reduced the influence of individual genetic backgrounds. Indeed, we hypothesize
that the natural gene expression variation occurring between unrelated individuals
masks this domain organization when we compare the transcriptome of individuals
from totally different genetic contexts. Moreover, the effect could be mainly
attributed to the additional HSA21, as confirmed by the absence of domains in the
healthy twins comparison. The conservation of this transcriptome pattern in iPS cells
raises the idea that the mechanisms responsible for the dysregulation domains are not
specific to fibroblasts but can be maintained in other cell types or tissues. It also
suggests that iPS cells can keep memory of the transcriptomic state of the parental
cells.
We have demonstrated that the GEDDs are conserved in the Ts65Dn mouse model for
DS. This conservation is remarkable because the domains can occur in different
chromosomes and thus in a genomic context that is different in human and in mouse.
This observation emphasizes the functional importance of the clustering of coregulated genes. The conservation in mouse also suggests that the domain
organization may be the consequence of the overexpression of one or more HSA21
gene(s) whose mouse orthologs are found on MMU16 syntenic region. Alternatively,
and in agreement with the hypothesis of a general perturbation of the cellular
homeostasis, the domains could be related to the physical presence of extra DNA
material in the nucleus.
13
We have shown that the GEDDs correlate with the previously described laminaassociated domains and replication domains. The sites of nuclear lamina interaction
and the late replicating regions show a remarkable correlation [40] and are known to
contain mostly constitutive heterochromatin associated to a strong gene repression.
On the contrary, early replicating domains tend to localize inside the nucleus (iLADs)
and contain more actively transcribed genes. Our results suggest that trisomic cells
undergo a derepression of the genes located at the nuclear periphery together with the
repression of the early replicating active genes. Our data reveal that the LADs
topology is not disturbed in trisomic cells, indicating that the overall genome
architecture is not modified by the presence of the additional HSA21. We therefore
hypothesized that the change of gene expression is related to epigenetic modifications
of the chromatin environment. This hypothesis was validated by the changes of
H3K4me3 marking observed between T1DS and T2N. We have demonstrated that
these changes follow a pattern highly similar to the one observed for the gene
expression dysregulation, confirming the idea of a general modification of the
chromatin within the domains. The alteration of H3K4me3 suggests that other histone
marks may also be disturbed in the trisomic cells. Interestingly, it was shown that the
histone hypoacetylation plays an important role in the maintenance of the late
replicating domains. More importantly, the artificial acetylation of those regions by
inhibition of histone deacetylases was associated to advanced replication time in the
early S phase [50].
In this context, several HSA21 genes emerge as potential candidates for the epigenetic
modification of chromosomal domains. The HLCS (holocarboxylase synthetase)
14
protein, encoded on HSA21 and triplicated in the Ts65Dn mouse model, is known to
catalyze the binding of biotin to histones, participating to chromatin condensation and
gene repression [51-54]. Interestingly, HLCS is associated to the nuclear lamina [55]
and can physically interact with chromatin modifying proteins such as DNMT1 (DNA
methyltransferase 1) or MeCP2 (methyl CpG binding protein 2) [56].
The HSA21 protein HMGN1 (high mobility group N1) is also a potential candidate.
This nucleosome-binding protein, also triplicated in the Ts65Dn mice, influences the
structure and the function of the chromatin through histone modifications [57]. The
loss of HMGN1 alters for instance the level of H3K14ac (acetylation of lysine 14 in
Histone H3) [58] or histone H3 phosphorylation [59]. Consequently, HMGN proteins
are also known as important modulators of gene expression [60]. Alteration of
HMGN1 level was shown to affect the behavior of mice and emphasized the possible
role of this protein in neurodevelopmental diseases [61].
The function of the DNMT3L (DNA (Cytosine-5-)-Methyltransferase 3-Like) protein
suggests that it may also be a candidate. First, it participates to de novo DNA
methylation by interacting with the DNA methyltransferases DNMT3A and
DNMT3B [62]. Second, it was shown to contribute to histone modification and
transcriptional repression through the histone deacetylase HDAC1 [63, 64]. However,
we found that DNMT3L is not expressed in the MZ twins’ fibroblasts and is not
located in the triplicated region of Ts65Dn mice, excluding it from the list of potential
candidates for any chromatin modification in our study.
Other HSA21 proteins are directly or indirectly involved in epigenetic mechanisms.
Among them, DYRK1A (Dual-specificity tyrosine-(Y)-phosphorylation regulated
kinase 1A), BRWD1 (Bromodomain and WD repeat-containing protein 1) and
RUNX1 (Runt-related transcription factor 1) may influence the epigenetic
15
architecture of the nucleus, in particular through their interaction with the SWI/SNF
chromatin remodeling complex [65-67]. All together, those data suggest that some
candidate genes on HSA21 may explain the gene expression changes observed in the
discordant twins by the modification of the chromatin environment.
Additionally, the correlation with the replication domains raises the hypothesis of a
modified replication time in trisomic cells. Several studies have shown that the cell
cycle is extended in trisomic nuclei, mainly through the elongation of the G2 phase
[68, 69]. This perturbation of the cell cycle may influence the DNA replication time
and the associated gene expression in trisomic cells. This hypothesis is not
incompatible with a modification of the chromatin landscape. It was shown for
instance that the binding of HMGN proteins to the chromatin is cell cycle dependant
[70].
Here, we propose that the overexpression of one or more HSA21 candidate gene(s)
(or the additional DNA material in the nucleus) modifies the chromatin environment
of the nuclear compartments in trisomic cells. These modifications would lead to a
general perturbation of the transcriptome that could explain some of the DS
phenotypes.
16
MATERIAL & METHODS
Cell culture and RNA preparation
Primary fetal skin fibroblasts were collected post mortem from the T1DS and T2N
discordant twins. The replicate experiment was performed from an independant
culture of the same fibroblasts. Primary fibroblasts from MZ1 and MZ2 healthy
monozygotic twins were derived from umbilical cord tissue collected as part of the
GenCord project in the University Hospitals of Geneva. Primary skin fibroblasts from
T21 and normal unrelated individuals were taken from [25] (GM08447, GM05756,
GM00969, GM00408, GM02036, GM03377, GM03440, PM9726F, GM04616,
AG07409, AG06872, AG06922, GM02767, AG08941, AG08942, D-S99124M). All
primary fibroblasts were grown in DMEM GlutaMAX™ (Life Technologies #31966)
supplemented with 10% fetal bovine serum (Life Technologies #10270) and 1%
penicillin/streptomycin/fungizone mix (Amimed, BioConcept #4-02F00-H) at 37°C in
a 5% CO2 atmosphere.
Mouse fibroblasts => Herault’s lab: could you please add details on the way mouse
fibroblasts were collected and cultured ?
Induced pluripotent stem cells (iPS) => Youssef: could you please add some details
on the iPS culture conditions ?
Total RNA was collected using the TRIzol® reagent (Life Technologies #15596)
following the manufacturer’s instructions. RNA quality was verified on the Agilent
2100 Bioanalyzer (RNA 6000 Nano Kit, #5067) and quantity was measured on a
Qubit® instrument (Life Technologies).
mRNA sequencing, mapping and normalization
mRNA-Seq libraries were prepared from 5 to 10μg of total RNA using the Illumina
mRNA-Seq Sample Preparation kit (#RS-100-0801), following the manufacturer’s
instructions. Libraries were sequenced on the Illumina HiSeq 2000 instrument using
paired-end sequencing 2x100bp. Reads were mapped against the human (hg19) or
mouse (mm9) genomes using the default parameters of the GEM mapper (Guigo’s
17
lab: is it necessary to add more details on the mapping method?) [30]. An independent
mapping was performed using the default parameters of TopHat. Quantile
normalization was performed on the RPKM data (Reads Per Kilobase per Million) for
all the human samples on one side and for the mouse samples on the other side.
Differential expression analysis was performed on the raw read counts using the
default parameters of EdgeR on the genes expressed in at least two samples [32].
Definition of chromosomal domains
For each gene, we determined the log2 expression fold change (log2FC) by
calculating the ratio of normalized expression values (after quantile normalization)
between the trisomic and the euploid samples (i.e. a positive log2FC reflects the
overexpression of the gene in the trisomic sample). For the healthy MZ twins the
log2FC was based on the ratio between MZ1 and MZ2. For the unrelated individuals,
the log2FC was based on the ratio between the mean value of all T21 individuals and
the mean value of all normal individuals. For all the comparisons, only the genes with
at least 0.1 rpkm in the normal individual(s) (T2N, average value of MZ1 and MZ2 or
average value of the normal unrelated individuals) were conserved. The log2FC
values were plotted equidistantly based on the gene positions along the chromosomes.
We used the lowess function in R (http://www.r-project.org) to smooth the log2FC
data (smoother span 1/20) and identify upregulated and downregulated domains. An
upregulated domain was defined as a set of at least 2 consecutive genes with positive
smoothed log2FC values. On the contrary, a downregulated domain was defined as a
set of at least 2 consecutive genes with negative smoothed log2FC values.
Expression level analysis
The genes were classified according to their expression level (rpkm value after
quantile normalization) in 3 categories: low (0.1<rpkm<10), medium (10<rpkm<100)
or high (rpkm>100) expression. In the discordant twins comparison, we used the
expression level of the normal twin (T2N) as a reference. In the healthy MZ twins
comparison, we used the average value between MZ1 and MZ2 to assess the gene
expression level. For each category, the distribution of log2FC was plotted (density
18
function in R). The difference of distribution between the discordant twins and the
healthy MZ twins was assessed by the Wilcoxon test.
Human/Mouse comparison
The mouse/human syntenic blocks coordinates were downloaded from the
“Comparative
Genomics”
track
of
the
UCSC
genome
browser
(http://genome.ucsc.edu). Only the largest mouse syntenic blocks (>1Mb) were
conserved for the comparative analysis. Log2FC of gene expression between Ts65Dn
and WT mice were plotted along the mouse chromosomes. Each gene was assigned to
a syntenic block (Total overlap between the gene coordinates and the block
coordinates) and colored accordingly. The corresponding blocks in human were
shown with the same colors in the GEDD plots.
The list of mouse and human orthologous genes was obtained from the Mouse
Genome Informatics Database project, The Jackson Laboratory, Bar Harbor, Maine
(ftp://ftp.informatics.jax.org/pub/reports/HMD_Human2.rpt).
We
compared
the
log2FC of expression obtained in the mouse with the values obtained for their human
orthologs. The comparison was done first with the discordant twins and then with the
healthy MZ twins. The degree of similitude between the 2 sets was given by the
Spearman correlation.
Lamina-associated domains (LADs) analysis
The human LADs coordinates were taken from [33] and the mouse LADs coordinates
were taken from [34] (mouse embryonic fibroblasts). A gene was assigned to a LAD
if its coordinates overlap from at least 1bp with the LADs coordinates. The
distribution of log2FC for the overlapping genes was compared to the distribution of
log2FC for the non-overlapping genes using a Wilcoxon test.
DamID
19
Method + analysis => van Steensel’s lab: could you please add details on the DamID
method?
Reduced Representation Bisulfite Sequencing (RRBS) Library Preparation
Reduced representation bisulfite sequencing (RRBS) libraries were made according to
[71] with some modifications. In short, the Qiagen DNeasy Blood and Tissue Kit was
used to extract DNA from the twin fibroblast cells. We then digested 2μg genomic
DNA with 2μl 20U/μl MspI restriction enzyme which cuts DNA regardless of
cytosine methylation status at CCGG sequence in 5μl NEB buffer 4 in a total reaction
volume of 50μl. This reaction was incubated for 18 hours at 37°C followed by heat
inactivation at 80°C for 20 min to generate DNA fragments containing CpG
dinucleotides at the end. Phenol extraction and ethanol precipitation steps were used
to clean up DNA from enzymatic reaction. DNA fragments were subsequently endrepaired and A-tailed by adding 2μl dNTP mix (10mM), 2μl 5U/μl Klenow Fragment,
and 4μl of NEB2 in a total reaction volume of 40μl. This reaction was incubated at
30°C for 20 min, followed by 20°C for 37 min. Heat inactivation was done at 75°C
for 20 min. The end-repaired and A-tailed DNA was purified by phenol extraction
and ethanol precipitation steps and ligated to the methylated version of Illumina
adapters:
ilAdap
Methyl
PE1
(
ACACTCTTTCCCTACACGACGCTCTTCCGATC*T) and ilAdap Methyl PE2
(GATCGGAAGAGCGGTTCAGCAGGAATGCCGA*G); all C’s are methylated, 5’
phosphate, *=phosphorothioate bond). The ligation mediated by adding 1μl T4 DNA
ligase (2,000 U/μl ), 2μl T4 ligase buffer (10×), and 1μl methylated adapters (pairedend adapters) (15μM) in a total reaction volume of 20μl. The reaction was incubated
at 16°C overnight. The adapter-ligated DNA was purified by phenol extraction and
ethanol precipitation steps and dissolved in 15μl EB buffer. Size selection of the
adapter-ligated DNA fragments (170-bp to 350-bp) was done by electrophoresing the
15μl ligation reaction in a 2.5% NuSieve GTG agarose gel (Lonza). We subsequently
purified the DNA using a Qiagen Qiaquick Gel Extraction kit as described in the
manufacturer’s instructions, except that we eluted the purified DNA from the column
using 25μl buffer EB. We then used 20mL of this purified DNA in the sodium
bisulfite conversion, which was performed using the QIAGEN Epitect Bisulfite Kit
20
(Qiagen, Valencia CA, USA) per manufacturer’s recommended protocol with the
following modifications: incubation after the addition of bisulphate conversion
reagents was conducted in a thermocycler with the following conditions: 95°C for 5
min, 60°C for 25 min, 95°C for 5 min, 60°C for 85 min, 95°C for 5 min, 60°C for 175
min, (95°C for 5 min and 60°C for 90 min) × 6. The bisulphate converted DNA was
eluted two times in 20μl of EB buffer yielding 40μl at the end. Illumina PCR primers
PE1 and PE2 were used for the final library amplification. The sequences of PCR
primers
are
as
follow:
ilPCR
PE1
(AATGATACGGCGACCACCGAGATCTACACTCTTTCCCTACACGACGCTCT
TCCGATC*T)
and
ilPCR
PE2
(CAAGCAGAAGACGGCATACGAGATCGGTCTCGGCATTCC
TGCTGAACCGCTCTTCCGATC*T),
where
an
asterisk
indicates
a
phosphorothioate bond. The purified bisulphate treated DNA was then PCR-amplified
in a reaction containing 100μl KAPA2G Robust DNA Polymerase mixture
(Kapabiosystems), 40μl bisulphite treated DNA, 0.5mM ilPCR PE1, 0.5mM ilPCR
PE2 DNA oligonucleotide primers in a total reaction volume of 200μl. The PCR
mixtures were divided amongst eight PCR tubes and were incubated at 95°C for 2
min, followed by 20 cycles of (95°C for 20 sec and 65°C for 30 sec and 72°C for 30
sec) and 72°C for 7 min. The PCR products were pooled and purified by adding 360μl
AMPure magnetic beads (Agencourt Bioscience, Beverly, MA). We confirmed the
amplification and correct product size range by running one-fifth of the reaction on a
2% agarose gel. We quantified the purified product using the Qubit fluorometer
(Invitrogen). We also checked the template size distribution by running an aliquot of
the library on an Agilent Bioanalyzer (Agilent 2100 Bioanalyzer). We then diluted
each library to 10nM and proceeded to sequence each library (pair-ended 100bp) in a
single lane on the Illumina HiSeq 2000 according the manufacturers instructions. We
also ran one lane of a standard (non-RRBS) library (Exome) in the same flow cell as
the control lane and used this lane (rather than the RRBS lane) to calculate the dye
matrix for base calling. Also, 5% phiX genomic DNA (control) was spiked into each
lane. We also limited the cluster number for RRBS libraries to a 80% optimized
cluster number per tile on Illumina HiSeq 2000. Image capture, analysis and base
calling were performed using Illumina’s CASAVA 1.7.
21
Methylation Computational analyses
We obtained 121.7 million reads for T1DS and 104.3 million reads for T2N of which
after trimming of reads for base quality and adapter contamination (trim_galore
wrapper http://www.bioinformatics.babraham.ac.uk/projects/trim_galore/) 59% and
59.6% of reads were uniquely mapped against human genome hg19 using Bismark
aligner (-n 3 -l 20 -I 20) [72]. We used MethylKit R package [73] to calculate
methylation percentages per each CpG. Percent methylation values for CpG
dinucleotides are calculated by dividing number of methylated Cs by total coverage
on that base. CpGs with at least 20X read coverage and at least a Phred score
threshold of 20 were retained for calling CpG methylation. 3,984,938 CpGs in T1DS
and 3,690,212 CpGs in T2N met these criteria of which 2,598,760 CpGs are covered
in both samples.
H3K4me3 ChIP-Sequencing
Primary human skin fibroblasts from both twins were grown in 10 cm dishes without
reaching confluence. 21 million cells were crosslinked by adding enough
formaldehyde for final concentration 0.5% to the growing media at RT. Dishes were
incubated for 10 minutes and the reaction was then quenched by adding glycine to a
final concentration of 125mM. The crosslinked cells were collected and stored at 80°C. For cell lysis and sonication, the pellets were resuspended in lysis buffer (1%
SDS, 10mM EDTA, 50mM Tris pH8.0) and sonicated in a Covaris S2 (duty cycle
5%, intensity 5, 200 cycles per burst, 6 minutes).
Chromatin from 7.5 million cells was incubated with 10μl of H3K4me3 antibody
(Abcam ab8580 0.5mg/ml) coupled to IgG magnetic beads (Cell Signaling) at RT for
1 hour and subsequently at 4°C for another hour. The magnetic beads were washed 5
times with LiCl wash buffer (100mM Tris pH 7.5, 500mM LiCl, 1% NP-40, 1%
sodium deoxycholate) and one time with TE buffer (10mM Tris-HCl pH 7.5, 0.1mM
EDTA). Subsequently, the bound DNA was eluted at 65°C in elution buffer (1% SDS,
0.1M NaHCO3) for 1 hour and then separated from the beads with a magnetic stand.
The immunoprecipitated DNA was incubated overnight at 65°C with 2μl Proteinase K
22
to reverse cross-link. The samples were then purified with the QIAquick purification
kit (Qiagen).
ChIP-Seq libraries were prepared using a ChIP-Seq DNA Sample Prep Kit from
Illumina with the following modifications to the protocol: mRNA adapter indexes
from the TruSeq RNA kit were used instead of the Adapter oligo mix. The enrichment
PCR was carried out with PCR reagents from the TruSeq mRNA kit from Illumina.
The enriched library was purified with AMPure magnetic beads (Agencourt), its
concentration verified with Qubit (Invitrogen) and the distribution and size of
fragments with a Bioanalyzer (Agilent). Four samples were pooled in equimolar
proportion and sequenced in a HiSeq2500 (single read, 36bp). The obtained reads
were demultiplexed with Illumina CASAVA 1.8 software and mapped to hg19 with
BWA.
Peaks
were
called
using
unique
tags
with
HOMER
(http://biowhat.ucsd.edu/homer/ngs/index.html). Peak comparison between twins was
performed with HOMER and in house scripts.
Correlation analysis between log2FC fibroblast vs. iPS log2FC, IMR90 Replication
Time, DNA methylation and H3K4me3 profiles
iPS log2FC chromosomal profiles have been obtained from RNA-Seq data as
previously described.
Replication time domains of IMR90 fetal fibroblasts were downloaded from the
ReplicationDomain database (www.replicationdomain.org). Gene Replication times
have been estimated for each gene considering the median of each overlapping
replication time domain and ordered according to chromosomal coordinates.
H3K4me3 profiles have been obtained as follows. For each gene body +/-1000bp, the
overlapping H3K4me3 ChIPSeq peaks were isolated and the logFC of the size of the
matching peaks (as provided by HOMER) of the trisomic and the healthy MZ twin
were calculated. The median of the resulting H3K4me3 logFC for each gene was
eventually retained to compose the chromosomal profiles according to gene
coordinates.
23
Similarly, DNA methylation profiles in gene bodies and promoters have been
obtained from averaging percent methylation values for CpG dinucleotides in gene
bodies and in promoter regions encompassing Transcription Starting Sites [TSS2000bp, TSS+100bp], respectively.
Locally Weighted Scatterplot Smoothing (Lowess) with 30% bandwidth has been
applied to calculate the Spearman correlation between fibroblast log2FC gene
expressions and iPS log2FC, Gene Replication times, DNA methylation and
H3K4me3 profiles for each chromosome. Significance of each chromosomal
correlation has been estimated by 106 random permutations. When the empirical pvalue is zero the theoretical one is reported.
DNase I hypersensivity
Method + analysis => Stamatoyannopoulos’ lab: could you please add details on the
DNase HS method?
24
LEGENDS
Figure 1: Gene expression fold change is organized in chromosomal domains. a,
Smoothed log2 fold change (log2FC T1DS/T2N) of gene expression between T1DS
and T2N fibroblasts along the human genome. Numbers indicate the chromosomes
delimited with vertical lines. Upregulated domains are shown in dark blue and
downregulated domains in light blue. b, Log2 fold change (log2FC T1DS/T2N) of
gene expression between T1DS and T2N fibroblasts along the human chromosomes
3, 11 and 19. c, Log2 fold change of gene expression between Ts65Dn and wild-type
littermates along mouse chromosomes 10, 15 and 19. Genes (in blue) are sorted
according to their position on the chromosome, at equidistance. The human and
mouse lamina-associated domains (LADs) [33, 34] are shown in red.
Figure 2: Weakly and highly expressed genes contribute differently to the
domains. a, Log2 fold change (log2FC T1DS/T2N) of gene expression between
T1DS and T2N fibroblasts along the human chromosomes 3, 11 and 19. Genes are
sorted according to their position on the chromosome (at equidistance) and colored
according to their expression level. b, Distribution of expression fold changes (log2)
for low (0.1<rpkm<10, upper panel), medium (10<rpkm<100, middle panel) and high
(rpkm>100, lower panel) expressed genes. Solid lines represent the log2FC between
the discordant twins (T1DS/T2N) and dashed lines represent the log2FC between the
healthy twins (MZ1/MZ2). p = Wilcoxon test p-value.
Figure 3: GEDDs are conserved in iPS cells. Comparison of the log2 expression
fold change profiles between T1DS and T2N in fibroblasts (in black) and iPS cells (in
red) along human chromosomes 3, 6, 11 and 16. ρ = Spearman correlation; p =
associated p-value.
25
Figure 4: GEDDs are conserved in Ts65Dn mouse model for DS. a, Log2 fold
change of gene expression in the Ts65Dn fibroblasts (Ts65Dn/WT) along the mouse
chromosomes 15 (MMU15, upper panel) and 19 (MMU19, lower panel). Genes are
sorted according to their position on the chromosome (at equidistance). Colors
represent the mouse/human syntenic blocks. The corresponding blocks on human
chromosomes are shown with the same colors. b, Spearman correlation plot of log2
expression fold changes between human (T1DS/T2N in upper panel and MZ1/MZ2 in
lower panel) and mouse (Ts65Dn/WT) orthologous genes. Mouse genes (in red) are
ranked in the ascending order of log2FC. Human genes (in blue) are plotted in the
same order. ρ = Spearman correlation.
Figure 5: Correlation with lamina-associated domains (LADs). Distribution of
log2 expression fold changes for genes overlapping a LAD (in red) and genes located
outside of a LAD (in yellow, inter LADs = iLADs). Profiles are shown for the
discordant twins (a), the Ts65Dn mice (b) and the healthy MZ twins (c). p =
Wilcoxon test p-value. d, DamID map of lamin B1 (LMNB1) interaction scores for
human chromosome 11 in T1DS (upper panel) and T2N (lower panel). e, Correlation
plot of lamin B1 interaction scores for the entire genome between T1DS and T2N
(r=0.75).
Figure 6: Correlation with replication time domains and H3K4me3 mark.
Comparison of the log2 expression fold change between T1DS and T2N (in black)
and the replication time (a) or the H3K4me3 differences (b) (in red) profiles along
human chromosomes 3, 6, 11 and 18. ρ = Spearman correlation; p = associated pvalue.
26
Table S1: List of GEDDs identified in the twins’ fibroblasts. Table gives the domain
ID, the names and coordinates of the first genes at the left and right borders of each
GEDD, the number of genes included in the domain and the median log2 fold change
within the domain.
Figure S1: Log2 fold change (log2FC T1DS/T2N) of gene expression between T1DS
and T2N fibroblasts along all human chromosomes. Genes (in blue) are sorted
according to their position on the chromosome, at equidistance. The laminaassociated domains (LADs) [33] are shown in red.
Figure S2: Log2 fold change of gene expression between the discordant twins (First
and replicate experiments on panels a and b, respectively), the healthy MZ twins
(panel c) and the T21 and normal unrelated individuals (panel d). Genes are shown in
grey and are sorted according to their position on the chromosomes. Blue lines
represent the lowess smoothing function and define upregulated (dark blue) and
downregulated (light blue) domains. LADs are shown in red. ρ = Spearman
correlation relative to the first experiment (T1DS/T2N).
Figure S3: Log2 fold change (log2FC T1DS/T2N) of gene expression between T1DS
and T2N fibroblasts along all human chromosomes. Genes are sorted according to
their position on the chromosome (at equidistance) and colored according to their
expression level (rpkm value from the normal twin T2N): 0.1<rpkm<10 in green,
10<rpkm<100 in blue and rpkm>100 in red.
Figure S4: Comparison of the log2 expression fold change profiles between T1DS
and T2N in fibroblasts (in black) and iPS cells (in red) along all human chromosomes.
ρ = Spearman correlation; p = associated p-value.
Figure S5: Log2 fold change (log2FC Ts65Dn/WT) of gene expression in Ts65Dn
fibroblasts along all mouse chromosomes. Genes (in blue) are sorted according to
their position on the chromosome, at equidistance. The mouse lamina-associated
domains (LADs) [35] are shown in red.
Figure S6: Difference of lamin B1 interaction scores between T1DS and T2N for all
human chromosomes. Difference signal is plotted according to the genomic location.
27
Figure S7: Correlation with replication time domains. Comparison of the log2
expression fold change (in black) and the replication time (in red) profiles along all
human chromosomes. ρ = Spearman correlation; p= associated p-value.
Figure S8: Correlation with DNA methylation profile. Comparison of the log2
expression fold change (in black) and the DNA methylation differences profile (in
red) along all human chromosomes. Gene methylation levels are based on the CpGs
located in the gene body (a) or the promoter (b). ρ = Spearman correlation; p=
associated p-value.
Figure S9: Correlation with H3K4me3 profile. Comparison of the log2 expression
fold change (in black) and the H3K4me3 differences profile (in red) along all human
chromosomes. ρ = Spearman correlation; p= associated p-value.
ACKNOWLEDGMENTS
We thank the Swiss National Science Foundation, the European Research Council,
AnEUploidy, the Lejeune foundation and the ChildCare foundation for financial
support. We thank Dr. Sophie Dahoun-Hadorn and Dr. Jean-Louis Blouin for the
discordant twins sample collection.
28
REFERENCES
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
Hunter, A.G.W., Down syndrome, in Management of Genetic Syndromes,
Third Edition, S.B. Cassidy and J.E. Allanson, Editors. 2010, John Wiley &
Sons, Inc. p. 309-332.
Antonarakis, S.E. and C.J. Epstein, The challenge of Down syndrome. Trends
Mol Med, 2006. 12(10): p. 473-9.
Antonarakis, S.E., et al., Chromosome 21 and down syndrome: from
genomics to pathophysiology. Nat Rev Genet, 2004. 5(10): p. 725-38.
Korenberg, J.R., et al., Down syndrome phenotypes: the consequences of
chromosomal imbalance. Proc Natl Acad Sci U S A, 1994. 91(11): p. 49975001.
Pritchard, M.A. and I. Kola, The "gene dosage effect" hypothesis versus the
"amplified developmental instability" hypothesis in Down syndrome. J
Neural Transm Suppl, 1999. 57: p. 293-303.
Rachidi, M. and C. Lopes, Mental retardation in Down syndrome: from gene
dosage imbalance to molecular and cellular mechanisms. Neurosci Res,
2007. 59(4): p. 349-69.
Epstein, C.J., The consequences of chromosome imbalance. Am J Med Genet
Suppl, 1990. 7: p. 31-7.
Shapiro, B.L., Down syndrome--a disruption of homeostasis. Am J Med
Genet, 1983. 14(2): p. 241-69.
Davisson, M.T., et al., Segmental trisomy as a mouse model for Down
syndrome. Prog Clin Biol Res, 1993. 384: p. 117-33.
Reeves, R.H., et al., A mouse model for Down syndrome exhibits learning
and behaviour deficits. Nat Genet, 1995. 11(2): p. 177-84.
Martinez-Cue, C., et al., A murine model for Down syndrome shows reduced
responsiveness to pain. Neuroreport, 1999. 10(5): p. 1119-22.
Costa, A.C., K. Walsh, and M.T. Davisson, Motor dysfunction in a mouse
model for Down syndrome. Physiol Behav, 1999. 68(1-2): p. 211-20.
Demas, G.E., et al., Spatial memory deficits in segmental trisomic Ts65Dn
mice. Behav Brain Res, 1996. 82(1): p. 85-92.
Chrast, R., et al., Mice trisomic for a bacterial artificial chromosome with
the single-minded 2 gene (Sim2) show phenotypes similar to some of those
present in the partial trisomy 16 mouse models of Down syndrome. Hum
Mol Genet, 2000. 9(12): p. 1853-64.
Ema, M., et al., Mild impairment of learning and memory in mice
overexpressing the mSim2 gene located on chromosome 16: an animal
model of Down's syndrome. Hum Mol Genet, 1999. 8(8): p. 1409-15.
Altafaj, X., et al., Neurodevelopmental delay, motor abnormalities and
cognitive deficits in transgenic mice overexpressing Dyrk1A (minibrain), a
murine model of Down's syndrome. Hum Mol Genet, 2001. 10(18): p. 191523.
Ahn, K.J., et al., DYRK1A BAC transgenic mice show altered synaptic
plasticity with learning and memory defects. Neurobiol Dis, 2006. 22(3): p.
463-72.
Conti, A., et al., Altered expression of mitochondrial and extracellular matrix
genes in the heart of human fetuses with chromosome 21 trisomy. BMC
Genomics, 2007. 8: p. 268.
29
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
Costa, V., et al., Massive-scale RNA-Seq analysis of non ribosomal
transcriptome in human trisomy 21. PLoS One. 6(4): p. e18493.
Esposito, G., et al., Genomic and functional profiling of human Down
syndrome neural progenitors implicates S100B and aquaporin 4 in cell
injury. Hum Mol Genet, 2008. 17(3): p. 440-57.
Li, C., et al., Genome-wide expression analysis in Down syndrome: insight
into immunodeficiency. PLoS One. 7(11): p. e49130.
Lockstone, H.E., et al., Gene expression profiling in the adult Down
syndrome brain. Genomics, 2007. 90(6): p. 647-60.
Mao, R., et al., Primary and secondary transcriptional effects in the
developing human Down syndrome brain and heart. Genome Biol, 2005.
6(13): p. R107.
Sommer, C.A., et al., Identification of dysregulated genes in lymphocytes
from children with Down syndrome. Genome, 2008. 51(1): p. 19-29.
Prandini, P., et al., Natural gene-expression variation in Down syndrome
modulates the outcome of gene-dosage imbalance. Am J Hum Genet, 2007.
81(2): p. 252-63.
Sultan, M., et al., Gene expression variation in Down's syndrome mice allows
prioritization of candidate genes. Genome Biol, 2007. 8(5): p. R91.
Deutsch, S., et al., Gene expression variation and expression quantitative
trait mapping of human chromosome 21 genes. Hum Mol Genet, 2005.
14(23): p. 3741-9.
Montgomery, S.B., et al., Rare and common regulatory variation in
population-scale sequenced human genomes. PLoS Genet. 7(7): p.
e1002144.
Dahoun, S., et al., Monozygotic twins discordant for trisomy 21 and
maternal 21q inheritance: a complex series of events. Am J Med Genet A,
2008. 146A(16): p. 2086-93.
Marco-Sola, S., et al., The GEM mapper: fast, accurate and versatile
alignment by filtration. Nat Methods. 9(12): p. 1185-8.
Trapnell, C., L. Pachter, and S.L. Salzberg, TopHat: discovering splice
junctions with RNA-Seq. Bioinformatics, 2009. 25(9): p. 1105-11.
Robinson, M.D., D.J. McCarthy, and G.K. Smyth, edgeR: a Bioconductor
package for differential expression analysis of digital gene expression data.
Bioinformatics. 26(1): p. 139-40.
Guelen, L., et al., Domain organization of human chromosomes revealed by
mapping of nuclear lamina interactions. Nature, 2008. 453(7197): p. 94851.
Peric-Hupkes, D., et al., Molecular maps of the reorganization of genomenuclear lamina interactions during differentiation. Mol Cell. 38(4): p. 60313.
Meuleman, W., et al., Constitutive nuclear lamina-genome interactions are
highly conserved and associated with A/T-rich sequence. Genome Res.
23(2): p. 270-80.
Zullo, J.M., et al., DNA sequence-dependent compartmentalization and
silencing of chromatin at the nuclear lamina. Cell. 149(7): p. 1474-87.
Lanctot, C., et al., Dynamic genome architecture in the nuclear space:
regulation of gene expression in three dimensions. Nat Rev Genet, 2007.
8(2): p. 104-15.
30
38.
39.
40.
41.
42.
43.
44.
45.
46.
47.
48.
49.
50.
51.
52.
53.
54.
55.
56.
Akhtar, A. and S.M. Gasser, The nuclear envelope and transcriptional
control. Nat Rev Genet, 2007. 8(7): p. 507-17.
Vogel, M.J., D. Peric-Hupkes, and B. van Steensel, Detection of in vivo
protein-DNA interactions using DamID in mammalian cells. Nat Protoc,
2007. 2(6): p. 1467-78.
Hansen, R.S., et al., Sequencing newly replicated DNA reveals widespread
plasticity in human replication timing. Proc Natl Acad Sci U S A. 107(1): p.
139-44.
Sadoni, N., et al., Stable chromosomal units determine the spatial and
temporal organization of DNA replication. J Cell Sci, 2004. 117(Pt 22): p.
5353-65.
Ma, H., et al., Spatial and temporal dynamics of DNA replication sites in
mammalian cells. J Cell Biol, 1998. 143(6): p. 1415-25.
Hiratani, I., et al., Global reorganization of replication domains during
embryonic stem cell differentiation. PLoS Biol, 2008. 6(10): p. e245.
DePamphilis, M.L., Review: nuclear structure and DNA replication. J Struct
Biol, 2000. 129(2-3): p. 186-97.
Gilbert, D.M., Replication timing and transcriptional control: beyond cause
and effect. Curr Opin Cell Biol, 2002. 14(3): p. 377-83.
Hiratani, I., et al., Replication timing and transcriptional control: beyond
cause and effect--part II. Curr Opin Genet Dev, 2009. 19(2): p. 142-9.
Weddington, N., et al., ReplicationDomain: a visualization tool and
comparative database for genome-wide replication timing data. BMC
Bioinformatics, 2008. 9: p. 530.
Pope, B.D., et al., Replication-timing boundaries facilitate cell-type and
species-specific regulation of a rearranged human chromosome in mouse.
Hum Mol Genet. 21(19): p. 4162-70.
Barski, A., et al., High-resolution profiling of histone methylations in the
human genome. Cell, 2007. 129(4): p. 823-37.
Casas-Delucchi, C.S., et al., Histone hypoacetylation is required to maintain
late replication timing of constitutive heterochromatin. Nucleic Acids Res.
40(1): p. 159-69.
Singh, M.P., S.S. Wijeratne, and J. Zempleni, Biotinylation of lysine 16 in
histone H4 contributes toward nucleosome condensation. Arch Biochem
Biophys. 529(2): p. 105-11.
Camporeale, G., et al., K12-biotinylated histone H4 marks heterochromatin
in human lymphoblastoma cells. J Nutr Biochem, 2007. 18(11): p. 760-8.
Pestinger, V., et al., Novel histone biotinylation marks are enriched in repeat
regions and participate in repression of transcriptionally competent genes. J
Nutr Biochem. 22(4): p. 328-33.
Chew, Y.C., et al., Biotinylation of histones represses transposable elements
in human and mouse cells and cell lines and in Drosophila melanogaster. J
Nutr, 2008. 138(12): p. 2316-22.
Narang, M.A., et al., Reduced histone biotinylation in multiple carboxylase
deficiency patients: a nuclear role for holocarboxylase synthetase. Hum Mol
Genet, 2004. 13(1): p. 15-23.
Zempleni, J., et al., The role of holocarboxylase synthetase in genome
stability is mediated partly by epigenomic synergies between methylation
and biotinylation events. Epigenetics. 6(7): p. 892-4.
31
57.
58.
59.
60.
61.
62.
63.
64.
65.
66.
67.
68.
69.
70.
71.
72.
73.
Postnikov, Y. and M. Bustin, Regulation of chromatin structure and
function by HMGN proteins. Biochim Biophys Acta. 1799(1-2): p. 62-8.
Lim, J.H., et al., Chromosomal protein HMGN1 enhances the acetylation of
lysine 14 in histone H3. EMBO J, 2005. 24(17): p. 3038-48.
Lim, J.H., et al., Chromosomal protein HMGN1 modulates histone H3
phosphorylation. Mol Cell, 2004. 15(4): p. 573-84.
Zhu, N. and U. Hansen, Transcriptional regulation by HMGN proteins.
Biochim Biophys Acta. 1799(1-2): p. 74-9.
Abuhatzira, L., et al., The chromatin-binding protein HMGN1 regulates the
expression of methyl CpG-binding protein 2 (MECP2) and affects the
behavior of mice. J Biol Chem. 286(49): p. 42051-62.
Liao, H.F., et al., Functions of DNA methyltransferase 3-like in germ cells
and beyond. Biol Cell. 104(10): p. 571-87.
Aapola, U., I. Liiv, and P. Peterson, Imprinting regulator DNMT3L is a
transcriptional repressor associated with histone deacetylase activity.
Nucleic Acids Res, 2002. 30(16): p. 3602-8.
Deplus, R., et al., Dnmt3L is a transcriptional repressor that recruits histone
deacetylase. Nucleic Acids Res, 2002. 30(17): p. 3831-8.
Lepagnol-Bestel, A.M., et al., DYRK1A interacts with the REST/NRSFSWI/SNF chromatin remodelling complex to deregulate gene clusters
involved in the neuronal phenotypic traits of Down syndrome. Hum Mol
Genet, 2009. 18(8): p. 1405-14.
Huang, H., et al., Expression of the Wdr9 gene and protein products during
mouse development. Dev Dyn, 2003. 227(4): p. 608-14.
Bakshi, R., et al., The human SWI/SNF complex associates with RUNX1 to
control transcription of hematopoietic target genes. J Cell Physiol. 225(2):
p. 569-76.
Paton, G.R., M.F. Silver, and A.C. Allison, Comparison of cell cycle time in
normal and trisomic cells. Humangenetik, 1974. 23(3): p. 173-82.
Contestabile, A., et al., Cell cycle alteration and decreased cell proliferation
in the hippocampal dentate gyrus and in the neocortical germinal matrix of
fetuses with Down syndrome and in Ts65Dn mice. Hippocampus, 2007.
17(8): p. 665-78.
Cherukuri, S., et al., Cell cycle-dependent binding of HMGN proteins to
chromatin. Mol Biol Cell, 2008. 19(5): p. 1816-24.
Gu, H., et al., Preparation of reduced representation bisulfite sequencing
libraries for genome-scale DNA methylation profiling. Nat Protoc. 6(4): p.
468-81.
Krueger, F. and S.R. Andrews, Bismark: a flexible aligner and methylation
caller for Bisulfite-Seq applications. Bioinformatics. 27(11): p. 1571-2.
Akalin, A., et al., methylKit: a comprehensive R package for the analysis of
genome-wide DNA methylation profiles. Genome Biol. 13(10): p. R87.
32