Download as pdf or txt
Download as pdf or txt
You are on page 1of 17

REVIEWS

DISEASE MECHANISMS

Down syndrome and the complexity


of genome dosage imbalance
Stylianos E. Antonarakis
Abstract | Down syndrome (also known as trisomy 21) is the model human phenotype for all
genomic gain dosage imbalances, including microduplications. The functional genomic exploration
of the post-sequencing years of chromosome 21, and the generation of numerous cellular and
mouse models, have provided an unprecedented opportunity to decipher the molecular
consequences of genome dosage imbalance. Studies of Down syndrome could provide knowledge
far beyond the well-known characteristics of intellectual disability and dysmorphic features, as
several other important features, including congenital heart defects, early ageing, Alzheimer
disease and childhood leukaemia, are also part of the Down syndrome phenotypic spectrum. The
elucidation of the molecular mechanisms that cause or modify the risk for different Down syndrome
phenotypes could lead to the introduction of previously unimaginable therapeutic options.

Down syndrome (DS) is the most commonly known causative genetic variation for certain phenotypes, such
disorder of intellectual disability, with a frequency of as congenital heart defects. Importantly, the past decade
one in approximately 700 births worldwide. DS was has generated a cautious enthusiasm for endeavouring
first clinically described in 1866 by Langdon Down in to treat individuals with DS using drugs; this enthu-
London1. Almost a century later, in 1959, trisomy 21 siasm and encouragement originated from studies in
(T21) was reported as the underlying genomic abnor- mouse models.
mality in Down syndrome by Lejeune–Gautier–Turpin All these advances in elucidating the molecular
at Necker Hospital in Paris, France2. The most useful mechanisms that underlie the phenotypes and diseases
mouse model for T21 was generated by Davisson in 1990 associated with DS are discussed in this Review. The pro-
(REF. 3), and the long arm of chromosome 21 (HSA21) posed formation of an international Down Syndrome
was fully sequenced in 2000 through an international Genomes Project is also briefly discussed in order to
collaborative effort4. Owing to its role in DS and its small initiate discussions and debate around this audacious
size, HSA21 is the most-studied human chromosome project, which aims to understand the genomic vari­
(Supplementary information S1 (figure)). ability that contributes to the phenotypic variability of
In the 16 years since HSA21 genome sequencing, DS. Several previous reviews have covered many aspects
there has been considerable progress in understanding of T21 and DS5–15; this Review discusses the most recent
the protein-coding gene content of HSA21, the content developments and prospects for future research. The
of other functional genomic elements and the individ- widespread implementation of non-invasive pre­natal
ual variability of the sequences of HSA21. In addition, testing 16,17 as a screening test for the fetal diagnosis
the development of several different mouse models of of T21 has been reviewed elsewhere18 and is thus not
T21 has facilitated many studies that have enhanced our discussed here.
understanding of this syndrome. There has also been
substantial progress in understanding the effect of T21 HSA21 functional elements and variation
Department of Genetic
Medicine and Development, on the cellular transcriptome and on cellular differenti- Since the complete DNA sequence of the long arm of
University of Geneva Medical ation, which has been studied using induced pluri­potent HSA21 was reported4, considerable progress has been
School and University stem cells (iPSCs). Important advances have been made made in understanding the functional genomic units of
Hospitals of Geneva, 1 rue in understanding the phenotypic impact of the over​ the chromosome (see FIG. 1 and Supplementary infor-
Michel-Servet, 1211 Geneva,
Switzerland.
expression of some HSA21 genes, the molecular basis mation S2 (table)). This has been achieved by large
Stylianos.Antonarakis@ of the early onset Alzheimer disease (EOAD) that is seen international projects such as the Encyclopedia  of
unige.ch in DS, the molecular basis of the leukaemias that fre- DNA Elements (ENCODE)19 and through the work
doi:10.1038/nrg.2016.154 quently occur in DS and the identification of genomic of individual laboratories that are interested in spe-
Published online 28 Dec 2016 regions of HSA21 that harbour functional elements or cific research questions. A complete understanding of

NATURE REVIEWS | GENETICS VOLUME 18 | MARCH 2017 | 147


©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

a
Exons 3.6%
Protein coding (2.4%)
lncRNA, miRNA, antisense (0.7%)
Pseudogene (0.3%)
Other (0.2%)
Regulatory 12.4%
Promoter (2.4%)
Enhancer (5.7%)
Insulator (1.3%)
Open chromatin (3%)
Repeats 36%
DNA (2.4%)
LINE (14.7%)
SINE (9.4%)
LTR (7.7%)
Other (1.8%)
Not annotated (48%)

b p13 p12 p11.2 q11.2 q21.1 q21.2 q21.3 q21.11 q22.12 q22.2 q22.3

GRC37/hg19
chr21:26,651,798-27,651,798 (1Mb)

26,700,000 26,800,000 26,900,000 27,000,000 27,100,000 27,200,000 27,300,000 27,400,000 27,500,000 27,600,000

Chromatin–chromatin interactions (HiC)

Gene
RPL13AP7 MIR155HG JAM2 APP MARCKSP1
LINC00158 MIR155 RNGTTP1 ATP5J
LINC00515 GABPA
FDX1P2
MRPL39
Promoter (ENCODE CHMM, 7 cell lines)

Enhancer (ENCODE CHMM, 7 cell lines)

Insulator (ENCODE CHMM, 7 cell lines)

DNase I DGF (UW 7 cell lines)

Linkage disequilibrium (Phased Genotypes, CEU)

eQTL (GTEx, 44 tissues)

Nature Reviews | Genetics

148 | MARCH 2017 | VOLUME 18 www.nature.com/nrg


©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

Expression quantitative the functional nucleotides of HSA21 is of paramount transcription factors that may influence the expression
trait loci importance for the elucidation of the molecular basis of other genes in the genome. Studying the overexpres-
(eQTLs). Genetic variation and ­pathophysiology of the various phenotypes of DS. sion of transcription factor genes on HSA21 will pro-
that is associated with gene HSA21 is the smallest human chromosome, and vide important information about the proportion of the
expression variation.
its overall gene density per megabase is average for entire transcriptome that is influenced by each of these
Copy number variants the human genome at approximately 15 genes per Mb genes, and by their combined action.
(CNVs). One form of genetic (REFS 4,9). Gene density for protein-coding and pseudo­ In one such recent study, which utilized the power of
variation in which sections of genes per Mb is also average; however, HSA21 is among mouse cellular models, chromatin immunoprecipi­tation
the genome are repeated and
the richest chromosomes for long non-coding RNA sequencing (ChIP-seq) analysis detected the binding of
the number of repeats varies
between individuals.
(lncRNA)-encoding genes but, by contrast, among the HSA21‑encoded transcription factor single-minded 2
the poorest for microRNA (miRNA)-encoding genes (SIM2) to more than 1,000 high-confidence DNA sites in
Indels per Mb (REF. 9). Some of these lncRNAs and mi­RNAs mouse embryonic stem cells following the overexpres-
A short name for insertions have emerged as important post-transcriptional regu- sion of the SIM2 gene. These sites overlap with regions
and deletions in the genome.
lators in stem cells20,21. HSA21 is also relatively poor in of occupancy of master transcription factors, which
Acrocentric other non-coding RNAs (ncRNAs), as well as in long define a particular sub-category of enhancers known as
Mammalian chromosomes in interspersed nuclear elements (LINEs) and short inter- the superenhancers23, which provides an explanation for
which the centromere is close spersed nuclear elements (SINEs) (see Supplementary how SIM2 can regulate its gene network and can result
to one end. Chromosomes 13,
information S3 (figure) for numbers and types of tran- in some of the DS phenotypes.
14, 15, 21 and 22 are
acrocentric in humans.
scripts per functional category). Overall, and as per the HSA21 encodes several classes of ncRNAs, the
ENCODE project results, HSA21 is among the poorest overexpression of some of which may contribute to
chromosomes in terms of functional DNA elements per different phenotypes of DS. A recent study of the non-­
Mb. It is easy to speculate that this is one of the reasons ribosomal transcriptome of endothelial progenitor cells
that T21 is viable in the postnatal period. Alternatively, in T21 provided initial evidence for the dysregulation
the viability of T21 could be due to the absence of any of the non-coding transcriptome in DS24. For example,
genes on HSA21 for which trisomy is not compatible HSA21 miR-155 was found to be overexpressed, which
with life after birth. Another alternative explanation is may be associated with the lower blood pressure that is
that the presence of an extra HSA21 does not substan- usually observed in DS. miR-155 downregulates angio­
tially disturb the mitotic mechanism and thus somatic tensin receptor 1 (AGTR1) in an allele-specific man-
cells with T21 (and without additional abnormalities) ner by binding the 1166A allele of AGTR1 but not the
could be continuously produced in vivo. ­hypertension-associated 1166C allele25.
Interestingly, HSA21 shows significant enrichment The genomic variability of HSA21 is of considerable
for encoded proteins that are located in cytoskeletal clinical importance, as a fraction of this chromosome is
structures compared with the rest of the genome, as functional and could contribute to the phenotypic vari­
can be seen when using the tool gostat (see Further ability of DS. Some single nucleotide variants (SNVs)
information). These cytoskeletal proteins are known alter codons in important functional parts of certain
to be involved in neurological disorders, in particu- proteins, whereas other SNVs occur in transcription
lar, Alzheimer neuropathology 22. Among the pro- factor-binding sites and could alter levels of gene expres-
teins known to be encoded by HSA21, 23 proteins are sion. Some SNVs have been identified as expression
involved in signal transduction and 31 proteins are quantitative trait loci (eQTLs) for gene expression26, and
as controlling differential methylation, splicing or chro-
matin marks27–30. Furthermore, there are close to 1,000
◀ Figure 1 | Functional elements of HSA21. a | Hierarchical annotation of the functional copy number variants (CNVs), approximately 500 indels
fraction of the long arm of HSA21 (21q). The functional fraction of 21q (comprising gene, and 25 inversions that have been mapped to HSA21 in
regulatory regions and open chromatin) includes approximately 16% of the total different individuals. All of these variants are candi-
sequence. The non-annotated portion is likely to also contain functional elements. dates for contributing to the phenotypic variability of
This fraction includes all introns and intergenic sequences (data from build hg19 and
T21. I argue that the sequence of HSA21 and the rest
REFS 156–159). The short arm of HSA21 is not included here because its sequences are
shared with those of the short arms of all other acrocentric chromosomes, and the exact of the exome or genome should be determined in every
sequences have not yet been completely elucidated. b | A representative 1 Mb segment individual with DS in order to define the genomic vari­
of HSA21 sequence. The chromatin interaction domains (topologically associating ants that are associated with or that cause the different
domains; TADs) as determined by Hi‑C are shown below the sequence. Genes, phenotypic features of the syndrome.
promoters, enhancers, insulators, DNase hypersensitivity (all likely to be functional
elements) are listed in the tracks below. The linkage disequilibrium domains (from Gene expression dysregulation domains
genomes of European ancestry) and the expression quantitative trait loci (eQTLs; T21 is probably a disorder of gene expression, as the
genomic variation associated with the gene expression variation) in 44 tissues identified basic genetic abnormality is the presence of an additional
by the GTEx (genotype tissue expression) project are shown at the bottom. The figure has copy of the 45 Mb acrocentric HSA21. The overexpres-
been created with the data from the genome browsers USCS and ENSEMBL. The seven
sion of the protein-coding genes and ncRNAs is likely
Encyclopedia of DNA Elements (ENCODE) cell lines are GM12878, H1hESC, HepG2,
HSMM, HUVEC, K562 and NHLF. CEU, samples of European ancestry from the Centre to disturb several cellular functions and developmental
d’Etude du Polymorphisme Humain; CHMM, continuous hidden Markov model; DGF, processes. In addition, the supernumerary HSA21 may
digital genomic footprinting; LINE, long interspersed nuclear element; lncRNA, long result in a global disturbance of the transcriptome owing
non-coding RNA; LTR, long terminal repeat; miRNA, microRNA; SINE, short interspersed to the additional chromosomal material, regardless of its
nuclear element; UW, University of Washington. genic content or regulatory repertoire.

NATURE REVIEWS | GENETICS VOLUME 18 | MARCH 2017 | 149


©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

Thus, the study of the transcriptomic differences the van Steensel laboratory 35, and replication domains in
between normal cells and T21 cells and tissues is of mammalian cells. The overall position of LADs was not
fundamental importance in understanding the molec- altered in trisomic cells; however, the active promoter
ular basis of the different phenotypic features of T21. chromatin mark profile of H3K4me3 (REF. 19) in T21
Studies using microarray and RNA-Seq (see REFS 8,31 fibroblasts was modified and accurately followed the
for additional references) have confirmed the increased GEDD pattern.
expression of the majority of the HSA21‑encoded genes These results indicate that the nuclear compartments
in DS individuals and have also indicated transcript of T21 cells undergo chromatin modifications that influ-
differences for genes on other chromosomes. There is, ence the overall transcriptome and that GEDDs may,
however, a substantial problem regarding these tran- therefore, contribute to some phenotypes of T21. The
scriptome comparisons, as the gene expression variation relation of GEDDs to specific phenotypes has, however,
among individuals introduces considerable ‘noise’. The not yet been studied, and it would be challenging to
abundance of genomic variants identified as eQTLs in prove which phenotypes are linked to GEDDs. These
the genome substantially alters allelic expression and GEDDs could not be observed in transcriptome compar-
thus for most genes there is wide variation in transcript isons of unrelated T21 fibroblasts and normal fibroblasts
levels across individuals. In an early study by Prandini (as would be expected due to variation in gene expres-
et al.32 on HSA21 gene expression in cell lines from sion); however, when only genes with low coefficients of
unrelated individuals, only 39% and 62% of the studied variation of expression are considered a highly statistical
HSA21 genes in lymphoblastoid and fibroblast cell lines difference is observed in the fold-change in expression
respectively showed a significant difference between T21 between T21 samples and normal samples for genes at
and normal samples. More than 50% of the transcripts the overexpressed GEDDs compared with those at the
studied showed substantial overlap of the distribution underexpressed GEDDs.
of expression levels between T21 and normal samples in These results provide preliminary evidence that T21
both cell types. This problem is even more pronounced could be considered a disorder that affects chromatin
in transcriptomic comparisons for genes outside of function and that the resulting phenotypes may be due
HSA21. Therefore, the natural variation of gene expres- to two reasons: first, the overexpression of HSA21 genes
sion masks the effect of the extra chromosome on the (coding and non-coding); and second, the dysregula-
overall transcriptome. tion of the expression of genes that are not on HSA21.
One possible solution to this problem is to study the In addition, the chromatin alterations observed in T21
differences in gene expression between samples from may not be T21 specific but may also occur in other
monozygotic twins discordant for T21; in this case, the aneuploidies or in normal cells under specific types
whole genome sequence is identical and thus all of stress.
the regu­latory variation is shared between the samples, The hypothesis that could be immediately tested
with the only differences originating from the super- is that the overexpression of certain genes on HSA21
numerary HSA21. An alternative solution is to sepa- causes GEDDs: such candidate genes include HMGN1
rate the cells from mosaic T21 tissues and to establish (high mobility group nucleosome binding domain 1),
pure populations of cells; in this case, the genomes are the product of which influences the structure and func-
also identical except for the presence of an extra copy tion of chromatin through histone modifications and
of HSA21 in one of the two cell populations. Our lab­ is considered to be an important modulator of gene
oratory was fortunate to identify a pair of monozygotic expression36. Other candidates are genes such as HLCS
twin fetuses that were discordant for T21 (REF. 33). The (holocarboxylase synthetase), DYRKIA (dual specificity
extra HSA21 in the T21 fetus was maternal in origin tyrosine-phosphorylation-regulated kinase 1A), BRWD1
and had a recombination 4 Mb from the telomere of (bromodomain and WD repeat-containing protein 1)
21q. RNA-Seq analysis revealed the expected over- and RUNX1 (Runt-related transcription factor 1),
all increase in gene expression of HSA21 genes; there which encode proteins that are involved in epigenetic
were also numerous differentially expressed genes on mechanisms and which may also influence chromatin
other chromosomes. Surprisingly, however, the over- structure. The competing hypothesis is that any extra
all genome-wide differential expression in fibroblasts chromosomal material of sufficient length may lead to
between the normal and T21 monozygotic twin sam- GEDDs; in this scenario, several trisomies, including
ples was organized along all chromosomes into domains T13 and T18, may result in a similar general dysfunc-
that were either upregulated or downregulated in T21. tion of gene expression. This could also be extended to
These were termed gene expression dysregulation ­cancer cells with different aneuploidies. Current mouse
domains (GEDDs)34. The GEDDs were well conserved models of T21 will provide excellent experimental bases
Lamina-associated domains
in iPSCs derived from the twins’ fibroblasts. The com- to test these alternative hypotheses.
(LADs). Regions of the
chromatin that interact with parison of the transcriptome of the fibroblasts from the A recent study compared the transcriptomes of 58
the nuclear lamina. Ts65Dn mouse model of DS and normal littermates also paired tissue samples from high-quality post-mortem
showed GEDDs along the mouse chromosomal regions human brains from individuals with DS and euploid
Replication domains that were syntenic in humans. Overexpression domains controls. The transcriptomes were assessed using an
Regions of the genome that are
copied (replicated) at the same
in humans were also overexpressed in the syntenic chro- exome array platform. The results of this study once
time during the S phase of cell mosomal regions in the mouse. The GEDDs correlated again showed that most of the dysregulated protein-­
division. with lamina-associated domains (LADs), as defined by coding genes do not map on HSA21, and that a

150 | MARCH 2017 | VOLUME 18 www.nature.com/nrg


©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

substantial fraction of these differentially expressed may not be reliable, as these correlations are based on
genes are associ­ated with oligodendrocyte differenti­ation expression in skin fibroblasts and blood rather than in
and myelination31. The involvement of the oligodendro- brain. However, the GTEx project has revealed several
glia was also observed in several iPSC differentiation genetic variants (eQTLs) that control the expression of
studies34,37,38. Another recent transcriptome study using APP in different tissues, including several brain regions;
fibroblasts and blood cells from T21 and normal controls these eQTLs could be used in well-powered studies as
showed the increased induction of interferon-stimulated risk biomarkers for APP expression levels and Aβ depo­
genes39, confirming an old idea that the interferon path- sition. The expression variation of APP (and the other
way is a major signalling cascade that is dysregulated in genes involved in this pathway) due to cis-regulatory
T21 (REF. 40). variation could explain the observation that levels of
the soluble Aβ42 peptide (an Aβ isoform that is 42 amino
DS and Alzheimer disease acids in length) are increased in approximately 50% of
Almost all patients with T21 develop neuropathology T21 fetal brains49. Other HSA21 genes that are involved
that is similar to that of Alzheimer disease (AD), and in the processing of APP include DYRK1A, SUMO3
the majority of them develop dementia by the age of (small ubiquitin-like modifier 3), BACE2, ETS2 and
60 years10,41. Individuals with DS now live longer than miR155 (REF. 17).
they used to thanks to improvements in health care Increased APP expression in the Ts65Dn mouse
and in care for people with disabilities; a recent study model has been shown to disrupt transport of nerve
in England and Wales estimated that the median life growth factor (NGF) and may thus cause the degener-
expectancy of people with DS who were born in 2011 ation of basal forebrain cholinergic neurons (BFCNs)50
is 58 years42. Thus, a considerable number of individ- that occurs in T21 and AD51–53. These observations were
uals with DS now reach the age of onset of AD. In fact, linked to abnormal endocytosis, a process that is essen-
T21 individuals constitute the largest fraction of people tial for neuronal function and the traffic of neurotrophic
with EOAD. factors54. It has recently been shown that APP increases
The proteins and cleavage peptides encoded by APP the activation of RAB5 to cause the enlargement of
(amyloid beta precursor protein) on HSA21 have been endosomes and that it disrupts the retrograde axonal
extensively studied as important components of the trafficking of NGF signals55; thus, dysregulation of RAB5
neuropathology of AD43,44. The deposition of the cleav- in response to APP overexpression may contribute to
age peptide amyloid beta (Aβ) in extracellular plaques EOAD in DS. HSA21 genes other than APP that affect
occurs decades earlier in DS than in non‑DS individu- the secretory-endosomal function, and which thus may
als. In DS–AD there are more and earlier plaques and also be implicated in EOAD in DS, include DOPEY2,
tangles in the hippocampus compared with EOAD cystatin B (CSTB) and synaptojanin 1 (SYNJ1)56,57.
and late-onset Alzheimer disease (LOAD) in non‑DS;
most of the deposition in non-DS LOAD is in the cor- DS and leukaemia
tex 11. Therefore, it is important to determine whether Children with T21 are at an increased risk of two dif-
the presence of three copies of the APP gene in T21 con- ferent types of leukaemia. The presence of a super-
tributes to EOAD. Several lines of evidence support this numerary HSA21 confers a 20‑fold increased risk of
hypothesis: first, rare duplications of the APP locus (in B‑lymphoblastic acute leukaemia58,59, and a 500‑fold
non‑DS individuals) causes autosomal dominant EOAD increased risk of acute megakaryoblastic leukaemia
with cerebral amyloid angiopathy 11,45, which is not (AMKL) 60 compared with the aged-matched gen-
observed in individuals with full T21; second, genomic eral population. Thus, T21 provides the opportunity
variation within the promoter of APP that increases its to better understand the molecular mechanisms of
expression also increases the risk for AD46; and third, a ­leukaemo​genesis58,61,62 (FIG. 3).
78‑year-old individual with partial T21 but with only two
copies of APP did not show the symptoms and neuro- Acute B‑lymphoblastic leukaemia in DS. Children with
pathology of AD47. Conversely, in all the genome-wide acute B-lymphoblastic leukaemia in DS (DS‑ALL) have
association studies reported so far no signal that has a poorer survival rates than non‑DS children with ALL;
significant association with EOAD and that maps near this is predominantly due to a much higher incidence
the APP locus has been detected. The expression of of relapses63. The current understanding of tumori­
Partial T21
A genomic alteration in which
APP (as for most genes) is variable across individuals genesis in DS‑ALL is that the primary somatic events
a portion of chromosome 21 and different tissues (FIG. 2a), and thus the expression of are mutations or rearrangements of CRLF2 (cytokine
is triplicated. APP in different tissues per individual should be con- receptor-like factor 2) or genes encoding other lympho-
sidered32. In addition, this expression variation results cyte differentiation factors, which are detected in the vast
Driver mutations
in an extensive overlap between trisomy and normal majority of DS-ALLs64–66. The progression of the disease
A mutation in a gene or other
functional genomic element tissues (FIG. 2b). Recent publicly available data from is associated with secondary activating driver ­mutations
that provides a selective the GTEx project (see Further information)48 on the in either JAK2 or RAS (NRAS and KRAS) genes, which
advantage to a clone of cells. gene expression in different tissues from a considera- coexist with those in CRLF2 and other primary driver
ble number of subjects have shown that APP expres- genes but which are frequently characterized by a
Variant allele frequency
The frequency of a variant (not
sion is poorly correlated among blood, skin and brain smaller variant allele frequency. Mutations in JAK2 or RAS
reference) allele in each human (FIG. 2c–d); therefore, the reported correlations between are present in roughly one-third of DS‑ALLs, and are
population. APP expression and AD in T21 and non‑T21 patients almost completely mutually exclusive66. Preliminary data

NATURE REVIEWS | GENETICS VOLUME 18 | MARCH 2017 | 151


©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

a 500

400

300

b 4
APP
RPKM

Normalized expression values


200 3

(qPCR)
2

100

0
0
Normal T21 Normal T21
m

d
Ca ortetal

lia l

he llar

ng en s

lia l
pu
ng sa

ng sa

(n=13) (n=15) (n=17) (n=16)


se

oo
ga mb leu
x

re

lia s
llu
c on

ga a

ga ba

po
am
isp e

bl
be
eb

cu uc
m eb
fr

ex
en

le
oc
re

ac N
at

he er
n

ho
m
Ce
ai

in
pp
ud

Lymphoblastoid Fibroblasts
ta

W
Br

Sk
Hi

Pu

cell lines
Nucleus accumbens

Putamen basal
Caudate basal

Hippocampus

Skin exposed

Whole blood
Brain frontal

hemisphere
Cerebellum
Cerebellar
ganglia

ganglia

ganglia
cortex

c d
1
Brain frontal 1 0.57 0.52 0.13 0.2 0.65 0.42 ‑0.08 ‑0.13
cortex 0.8
APP expression, hippocampus (RPKM)

Caudate basal 1 0.52 0.44 0.55 0.61 0.76 ‑0.15 ‑0.04 300
ganglia 0.6
Cerebellar 1 0.49 0.29 0.48 0.46 ‑0.14 ‑0.27
hemisphere 0.4

Cerebellum 1 0.38 0.13 0.32 0.16 ‑0.08 0.2

1 0.46 0.45 ‑0.04 0.17 0 200


Hippocampus
Nucleus accumbens 1 0.57 ‑0.15 ‑0.05 −0.2
ganglia
Putamen basal −0.4
1 ‑0.06 0.02
ganglia
−0.6 100
Skin exposed 1 0.17
−0.8
Whole blood 1
−1
80 100 120 140 160
APP expression, skin (RPKM)
Figure 2 | Variation in expression of amyloid precursor protein. (Spearman correlation values are shown). Note the low correlation of
a | Example of expression variation of the amyloid precursor protein (APP) Nature
expression between the blood, skin and regions of Reviews
the brain. | Genetics
APP expression
gene in different tissues from samples collected and analysed in the GTEx is also poorly correlated among the different regions of the brain. Thus, for
project (REF. 48 and GTEx project). b | Example of gene expression variation APP transcription studies, it is not possible to use one tissue as a proxy
for the APP gene on HSA21. The expression levels of several samples from for another, and it is necessary to study the appropriate tissue for the
T21 and normal cells in lymphoblastoid cell lines and fibroblasts are shown. hypothesis that is being tested. d | The poor correlation of APP expression
Note the extensive overlap of gene expression variation among the 62 between the skin and hippocampus. Data from the GTEx project. qPCR,
samples tested. Data from REF. 32. c | Correlation of expression of the APP quantitative polymerase chain reaction; RPKM, reads per kilobase of
gene in different tissues from the GTEx (genotype tissue expression) project transcript per million mapped reads.

152 | MARCH 2017 | VOLUME 18 www.nature.com/nrg


©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

JAK2

B-lymphocyte • CRLF2
B-lymphocyte • Chromatin remodellers
proliferation • Haemato lineage TFs RAS DS-ALL
• Tumour suppressors
HMGN1 • Cohesin complex
Lymphoid
stem cell

Haematopoietic
stem cell

GATA1 (X-chr)
+

AMKL
Myeloid +
stem cell Additional
driver mutations
GATA1 (X-chr)
+
TMD
Megakaryoblast

Disappearance
of clones
Figure 3 | Leukaemogenesis in Down syndrome. Mechanisms of leukaemogenesis in the two types of leukaemia in
Down syndrome, acute B-lymphoblastic leukaemia (DS-ALL) and acute megakaryoblastic leukaemia Nature Reviews
(AMKL), Genetics
are| depicted.
In DS‑ALL (top part of the figure) the overexpression of the HMGN1 gene on HSA21 contributes to B‑lymphocyte
proliferation, and subsequent somatic driver mutations in either JAK2 or RAS genes cause the progression to DS‑ALL.
Additional somatic mutations that contribute to the leukaemogenesis are shown and include cytokine receptor-like
factor 2 (CRLF2), chromatin remodellers, haematopoietic-related transcription factors (TFs), tumour suppressors and
proteins of the cohesin complex. In transient myeloproliferative disorder (TMD) and AMKL (bottom part of figure) the
somatic mutation of GATA1 on the X‑chromosome (X-chr) together with the overexpression of genes on HSA21 leads to
TMD. Additional driver somatic mutations convert some of the TMD clones to AMKL. Alternatively, the TMD clones may be
eliminated, resulting in spontaneous remission.

suggest that the secondary driver mutation in DS‑ALL and splenomegaly 68,69. The clonal expansion of mega­
may switch between the primary tumour and relapse, karyoblasts is exclusively associated with somatic trun-
for example, from JAK2 or PTPN11 to RAS but that the cating mutations in the GATA1 transcription factor gene
primary driver mutation remains the same66. that maps on the X‑chromosome. These mutations elimi­
Notably, none of the above-mentioned driver muta- nate the full-length GATA1 protein and leave only the
tions maps to HSA21. What then is the importance of the short isoform, GATA1s. Interestingly, GATA1 mutations
extra HSA21 in DS‑ALL? The study of transgenic mice are very rarely observed in leukaemic cells lacking T21
(mouse model Ts1Rhr) revealed that a triplication of a (REF. 70), further suggesting the necessity of ­interaction
genomic interval orthologous to HSA21 that contains 31 between T21 and GATA1s in AMKL in DS.
genes confers self-renewal of mouse progenitor B cells. The contribution of T21 to AMKL may be related to
Among these genes, overexpression of HMGN1, which the triplicate overexpression of the RUNX1 (REF. 71) and
encodes a nucleosome remodelling protein, is associated ERG genes72, as the somatic triplication (focal amplifica-
with altered chromatin marks, such as the suppression tion) of the HSA21 region that contains these genes in
of the repressive mark H3K27me3, and the proliferation of non‑DS individuals is associated with myeloblastic leu-
B lymphocytes. HMGN1 overexpression remarkably links kaemia73,74. Recent work using iPSCs from T21 with or
the DS‑ALL and the GEDD observations discussed above, without GATA1 mutations and with or without deletion
providing a rationale for further ­investigations of the role of a 4‑Mb region of HSA21 showed that T21 resulted
of this gene in phenotypes of DS67. in the accelerated production of aberrantly differenti-
ated cells through the increased production of early
Transient myeloproliferative disorder and AMKL. haematopoietic progenitors and the upregulation of
The occurrence of transient myeloproliferative dis­order mutated GATA1 (REF. 75). These effects were mediated by
(TMD) in newborns with DS and the occurrence of ­dosage alterations of RUNX1, ETS2 and ERG genes that
Megakaryoblasts AMKL in children with DS shed light on a leukaemic are encoded within the 4 Mb deleted region on HSA21.
Precursor cell to a process that begins in the fetal period and that evolves Remarkably, most infants with DS‑TMD spontan­
promegakaryocyte,
which in turn becomes a
for several years. Approximately 10–15% of newborns eously regress to complete remission. However, 20–30%
megakaryocyte during with DS present with TMD, that is, with circulating of the cases with regressed DS‑TMD develop AMKL
haematopoiesis. ­ egakaryoblasts, leukocytosis, anaemia, hepatomegaly
m some years later, usually between 1 and 4 years of age76.

NATURE REVIEWS | GENETICS VOLUME 18 | MARCH 2017 | 153


©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

The leukaemic clones in these DS‑AMKL cases contain of approximately 5.27 Mb on 21q22.3 (REF. 81) that was
the initial GATA1‑truncating mutations, but have also later narrowed down to 1.77 Mb (REF. 82) (between genes
acquired, through continuous mutagenesis, additional DSCAM and ZBTB21). In a similar study with differ-
driver mutations in known cancer genes70,77, including ent partial T21 samples, the CHD candidate region was
cohesin component gene products and epigenetic regu- mapped to an overlapping region of 15.4 Mb (REF. 83).
lators. These third-hit driver gene mutations are similar The mouse models Ts65Dn, Ts1Cje and Dp(16)2Yey,
to those observed in non‑DS leukaemia genes (such as which have partial T21 and which include CHDs in
FLT3, EZH2, APC, JAK and RAS). their phenotypic manifestations, provided a candidate
Open questions on the fascinating molecular mech- interval that is compatible with the conclusions from the
anism underlying the TMD to AMKL progression in human partial T21 (REFS 84–87). Recently, studies using
DS include the exact mechanism of the cooperation the mouse model Dp3Tyb further narrowed down the
between the GATA1 and HSA21 genes, the mutation CHD interval to a 5‑Mb region88. The use of a combined
rate in cellular clones with GATA1 mutations, and the approach that included Drosophila melanogaster, mouse
sequential genomic and epigenomic changes that drive transgenesis and a cardiac myoblast H9C2 cell model
leukaemogenesis78. suggested that the combined overexpression of DSCAM
and COL6A2 contributes to ASD89.
Congenital heart defects and T21 The example of CHDs in DS illustrates the necessity
Among the most notable and relatively easily scor­ of a multidisciplinary approach to identify the tripli-
able phenotypes of DS are the various congenital heart cated functional genomic elements that are necessary
defects (CHDs). However, all of these display incomplete for each phenotypic feature of DS. A detailed analysis
penetrance, as only approximately 40% of DS individ- of the variation in HSA21 and the rest of the genome
uals present with some form of CHD79. The most fre- that modifies the risk for CHDs was previously lack-
quent forms of CHDs in DS are the atrioventricular ing. However, recently, such a study was carried out
septal defects (AVSDs) constituting 43% of CHD cases, using cohorts of approximately 250 DS individuals
whereas ventricular septal defects (VSDs), atrial septal with or without CHDs. This study identified SNPs and
defects (ASDs) and tetralogy of Fallot constitute 32%, CNVs on HSA21 that significantly modify the risk for
19% and 6%, respectively. In fact, AVSDs are almost CHDs and, in particular, AVSDs through genotyping
exclusively seen in DS79,80. Attempts have been made and comparative genomic hybridization using a tiling
to map the genomic determinants of the CHDs in DS array for the entire HSA21 (REF. 90). These findings
on HSA21. By analysing rare partial T21 cases with were not confirmed by a separate study, although the
CHDs, investigators have identified minimal CHD can- genotyping methodology was different between the
didate regions. One study identified a candidate region two studies91. The search for the genomic variability
underlying CHDs in DS provides a proof‑of‑principle
for the large-scale sequence analysis of a sufficiently
large number of DS individuals that is needed in order
Box 1 | The Down Syndrome Genomes Project
to identify the genomic determinants of the DS pheno-
The current working hypothesis in the field is that the phenotypic heterogeneity and types. The results of these analyses will advance not only
variability in Down syndrome (DS) is primarily due to: first, the three copies of understanding of the molecular pathophysiology of DS,
functional genomic elements on HSA21; second, the genetic variation of HSA21; and but also understanding of similar phenotypes (includ-
third, the genetic variation on non‑HSA21 loci. Thus, the variable phenotypes are likely ing AD, CHDs and intellectual disability) in non‑DS
to be multifactorial, caused by variation at multiple loci and interactions between them
­individuals (see BOX 1).
and with non-genetic factors.
I propose the formation of a new international collaborative effort, the Down
Syndrome Genomes Project, to collect samples, to establish and monitor the phenotype, Candidate genes by haploinsufficiency
and to identify the genomic variation of a large number of DS samples from all geoethnic Some genes in the genome are categorized as haplo­
groups in order to unravel genomic contribution to the phenotypic variability. The insufficient, as loss‑of‑function variants of one allele
genome exploration may include the generation of a high depth sequence of the entire result in a recognizable phenotype. An illustrative exam-
HSA21 (more than 100x coverage), an exome or genome sequence of more than 50x ple is the DYRK1A gene on HSA21; truncating mutations
coverage, and at a later stage a long-read sequence of HSA21. in one allele cause one form of dominant intellectual
The objective will be to identify not only the single nucleotide variants (SNVs) but disability 92,93. It has been argued that haplo­insufficient
also the copy number variants (CNVs) and any other variability that may contribute to genes are also sensitive to three copies and are thus good
the phenotypes of DS. The required number of samples needs to be determined but a
candidates for contributing to some of the phenotypes
rough estimate of 5,000 would probably suffice to study the common phenotypic
features of DS. For example, in the case of the atrioventricular septal defects (AVSDs)
in full or partial trisomies94. Indeed, several studies have
in DS with a prevalence of 16%, a genotype relative risk of a causative variant of 2, shown that the overexpression of DYRK1A and other
and a variant allele frequency (VAF) of 0.3, a sample size of 300 cases (DS with genes contributes to the phenotypes of DS94,95. Recent
AVSDs) and 300 controls (DS without any congenital heart defects) provides 90% studies have resulted in haploinsufficiency prediction
statistical power to detect the risk allele154,155. Similarly, for the early onset Alzheimer scores for most protein-coding genes96,97. The curated
disease (EOAD) in DS a sample size of 1,000 cases and an equal number of controls and publicly available data from exome sequencing of
(DS without EOAD) would be required. Of paramount importance is the accurate approximately 60,000 samples from the Broad Institute
phenotypic characterization of the participating subjects and appropriate follow-up. (ExAC; see Further information)98, provide valuable
Cohorts of different ages will also be important in order to capture the entire information on the tolerance of loss‑of‑function vari-
spectrum of the phenotypic variability.
ants in each gene; genes intolerant to loss‑of-function

154 | MARCH 2017 | VOLUME 18 www.nature.com/nrg


©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

7.5

GRIK1−AS2

7.0
S100B

KCNE1 SMIM11
TFF1
LRRC3 DSCR4
KCNE2 CLDN8 PCP4
CSTB SUMO3
TFF3 OLIG1
PIGP C21orf62
CLDN14 OLIG2 SOD1
ATP5J
6.5 CBR3 B3GALT5
YBEY HMGN1
N6AMT1 MRPS6
SH3BGR TFF2 CYYR1 CRYAA
−log 10 (P value)

TCP10L CBR1
AP000695.1 C21orf128 KCNJ15 WRB TMEM50B
CLDN17 MIS18A
SPATC1L
FAM207A PTTG1IP RCAN1
MAP3K7CL RIPPLY3 MRAP
ATP5O
C21orf91 CHODL BTG3
IL10RB C21orf33 HSF2BP RWDD2B UBE2G2 DSCR3
C21orf2 KCNJ6
FAM3B
RBM11 HSPA13 PSMG1 PDXK U2AF1
SLC5A3
LIPI ICOSLG EVA1C
RSPH1 IGSF5 IFNAR2 CXADR BACE2
C21orf59 IFNGR2 ETS2
6.0 DNAJC28 CLIC6 SAMSN1 RUNX1 ERG
C21orf58 AGPAT3
LCA5L WDR4 PKNOX1
POFUT2 SLC19A1 CHAF1B PCBP3
NDUFV3 CRYZL1 PRMT2
TMPRSS3 CBS JAM2 ADARB1 CCT8
SETD4 POTED
DNMT3L BACH1 MRPL39 IFNAR1 GABPA HUNK
HLCS ABCG1
FTCD
RRP1 TMPRSS2 SIM2 ZBTB21
SIK1
DONSON RIPK4 C2CD2
AIRE ADAMTS5 NRIP1
DYRK1A
MX1 TSPEAR ITGB2
ADAMTS1 PWP2
UBASH3A MX2 RRP1B GART NCAM2
USP16 MORC3
PFKL
GRIK1 PDE9A SLC37A1 APP
LSS TRAPPC10
TPTE PAXBP1
5.5 TMPRSS15 SCAF4
COL6A2 DIP2A COL6A1
UMODL1 PRDM15 USP25
TRPM2 TIAM1
COL18A1 MCM3AP
SON
LTN1
DOPEY2 SYNJ1
DSCAM
TTC3 ITSN1
BRWD1

PCNT

5.0
0 0.25 0.50 0.75 1.00
Haploinsufficiency (pLI)
Figure 4 | Haploinsufficiency of genes on HSA21. Plot of the haploinsufficiency (intolerability of loss‑of‑function
variants) index of HSA21 genes. The x‑axis contains the pLI metric from the EXaC database (see FurtherNature Reviews | Genetics
information),
which is a metric of the number of observed loss‑of‑function variants over that expected. The y‑axis contains the negative
log 10 (P value) for the estimated metric. Genes on the right of the plot, close to pLI of 1 (in the darker orange area) are
more likely to be haploinsufficient, whereas genes on the left, close to pLI of 0 (in the lighter orange area) are less likely to
be haploinsufficient. In red are genes that are causative for recessive disorders in which carriers with non-functional
alleles are unaffected. As expected, these genes lie mostly within the non-haploinsufficient area. Genes with the highest
haploinsufficiency index are most likely to contribute to a Down syndrome (DS) phenotype, whereas those with the
lowest haploinsufficiency index are least likely to contribute to a DS phenotype. More details on the calculation of the pLI
are provided in the EXaC FAQ page (see Further information).

variants in heterozygosity are likely to be haploinsuffi­ DS phenotypes. Notably, when HSA21 protein-coding
cient, whereas genes that tolerate such variants in genes are ranked according to haploinsufficiency (that is,
hetero­zygosity are more likely to be involved in reces- intolerability to loss‑of‑function variants), several of the
sive Mendelian phenotypes. HSA21 protein-­coding genes that rank highly (such as DYRK1A) have already
genes with high haploinsufficiency scores could then been associated with the pathophysiology of certain
be considered as good candidates for contribution to ­clinical features of DS (FIG. 4).

NATURE REVIEWS | GENETICS VOLUME 18 | MARCH 2017 | 155


©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

Mouse model studies of DS The continuation of studies in laboratory mice of


The elucidation and comparative analysis of the partial T21 is crucial for understanding DS, and cross-
genomes of human and laboratory mice revealed that the ing existing mice could provide additional models.
long arm of human HSA21 is homologous to portions For example, crossing mice in which single genes on
of three mouse chromosomes: MMU16, MMU17 and HSA21 have been disrupted could be used to system-
MMU10. It thus became clear that it would be difficult atically delete the extra copy of specific genes from the
to create a mouse model for trisomy of the entire HSA21. trisomic genome, thus enabling the contribution of
Mouse models for T21 that have been introduced over each gene to be assessed. In addition, the introduction
the years3,85,86,88,99–106 are described in BOX 2. of newer gene-editing methods could, in theory, make

Box 2 | Mouse models of T21


A mouse trisomic for mouse chromosome 16 (MMU16) was created in 1985, the T21 scientific community provides an enthusiasm for many discoveries
but it was embryonic lethal. Myriel Davisson at the Jackson Laboratory in in the future.
1990 made a viable partial trisomy 16 mouse3 equivalent to a large segment In parallel, a transchromosomic mouse has been generated by introducing
of HSA21. This mouse, Ts65Dn, has been extensively used for the a copy of the human HSA21 in mice105. Human HSA21 in the mouse has
understanding of the molecular pathophysiology of Down syndrome as it several small deletions, duplications and rearrangements owing to previous
had abnormal cognitive phenotypes and neurophysiology findings. Ts65Dn cell irradiation, but this Tc1 mouse continues to serve as a useful model,
also carries a substantial segment of MMU17 not syntenic to HSA21. particularly in the study of Alzheimer neuropathology that is seen in T21.
Another mouse trisomic for a smaller segment of MMU16 (and a partial In general, mice with partial triplication of MMU16 show cognitive
monosomy for MMU12) was subsequently generated86 and the comparison phenotypes similar to those in T21 (using the Morris water maze, T‑maze,
of its phenotypes was instructive in order to begin to dissect the genomic fear condition and tests of synaptic plasticity), although partial trisomy
segments that are involved in different phenotypes. The advancement of MMU10 mice do not show cognitive impairment. Partial MMU17 trisomic
techniques for the manipulation of the mouse genome has since resulted in mice show abnormal hippocampal long-term potentiation. Detailed
a substantial number of additional mouse models of partial ‘human T21’. descriptions of the phenotypes in each particular mouse model can be
The figure schematically shows the different partial trisomy mouse models found in the original literature.
(for genomic regions syntenic to HSA21) published to date. A similar See the figure for a schematic representation of the mouse models of T21.
number of partial monosomies for regions syntenic to HSA21 have also The syntenic fraction of MMU16 is show in blue, that of MMU17 in green,
been made. and that of MMU10 in red. The extent of the partial triplication per mouse
A whole ‘human T21’ mouse model was produced in 2010 through model is shown below. The bordering genes for these triplicated regions are
appropriate crosses of partial trisomy mice for all syntenic regions of also shown. The human HSA21 inserted in mouse, Tc1 (REFS 107,146), is
HSA21 (REF. 85) (combination of models Dp(16)1Yey, Dp(17)1Yey and shown at the top of the figure. Duplicated regions are shown with thick
Dp(10)1Yey; see the figure). There is currently an effort to create mouse lines, and deletions are shown with thin lines. The references for these
models for smaller regions of partial trisomy and to study the resulting mouse models are: Ts65Dn3, Ts1Cje86, Ts1Rhr101, Dp1Yah102, Dp2Yah103,
phenotypes. Most of these newer mouse models have not yet been Dp3Yah108, Dp1Tyb–Dp9Tyb88, Dp(16)1Yey104, Dp(16)2Yey105, Dp(16)3Yey and
extensively studied but their availability and the collaborative spirit within Dp(16)4Yey106, and Dp(17)1Yey and Dp(10)1Yey85.

Tc1

HSA21

MMU16 MMU17 MMU10

Hspa13 Il10rb mir802 Dyrk1a B3galt5 Zbtb21 U2af1 Cstb Prmt2


Lipi mir155 App Sod1 mir18a Cbr1 Kcnj6 Fam3b Rrp1b Pdxk
Hunk Ifnar1 Runx1 Dscr3 Igsf5 Abcg1
Tiam1 Umodl1
Dp9Tyb
Dp1Tyb
Dp3Yah
Dp(16)1Yey
Dp(10)1Yey
Dp2Yah
Dp(17)1Yey
Ts65Dn Dp1Yah
Dp(16)2Yey
Dp(16)3Yey
Ts1Cje
Dp7Tyb
Dp2Tyb
Dp8Tyb
Dp(16)4Yey
Dp4Tyb
Dp3Tyb
Ts1Rhr
Dp5Tyb
Dp6Tyb

Nature Reviews | Genetics

156 | MARCH 2017 | VOLUME 18 www.nature.com/nrg


©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

Table 1 | iPSCs from trisomy 21


Samples Differentiation Conclusions Refs
DS and control 109
DS and control Neurons Cortical neurons derived from T21 iPSCs exhibited greater secretion 110
of amyloid peptides, tau protein phosphorylation and cell death
DS (free T21 and 111
translocation)
DS and control Neurons • Reduced number of neurons 112
• NPCs produced more APP
DS (isogenic, the extra Haematopoietic Higher colony-forming potential for erythroid, myeloid and 113
HSA21 was lost in one progenitors megakaryocytic lineages
iPSC)
DS and control Haematopoietic Increased propensity for erythropoiesis 114
progenitors
DS (isogenic) Neurons Impairment of neurogenesis 115
DS and control Neurons Defective early glial development 38
Mosaic DS (isogenic) Neurons Significant synaptic deficit of excitatory glutamatergic synapses 116
and inhibitory GABAergic synapses
DS Silencing of one HSA21 through insertion of XIST gene 117
Monozygotic twins DS Neurons • Reduced neurogenesis, more astroglia and oligodendroglia 34
and normal (isogenic) • DYRK1A inhibition corrected these defects 37
DS and normal Neurons • Astrocytes had higher ROS and lower synaptic molecules; oxidative 118
(isogenic and stress-mediated cell death and abnormal maturation of neurons
non-isogenic) • Partial correction with minocycline
Mosaic DS (isogenic) • Neurons • Reduced neurogenesis, mitochondrial dysfunction, increased 119
• Haematopoietic secretion of amyloid peptides
progenitors • Reduced and faster expansion of haematopoietic lineage
Full T21, partial T21 Haematopoietic Model for TMD; T21 perturbs haematopoietic development through 75
(del 4 Mb) and normal progenitors increased production of early haematopoietic progenitors and
with and without upregulation of the mutant GATA1; this in turn results in accelerated
GATA1 mutations production of aberrantly differentiated cells
APP, amyloid precursor protein; DS, Down syndrome; iPSCs, induced pluripotent stem cells; NPCs, neuronal progenitor cells;
ROS, reactive oxygen species; T21, trisomy 21; TMD, transient myeloproliferative disorder.

it possible to silence one allele of the triplicated genes (EGCG) inhibitor of DYRK1A107 and the GABAA antag-
in these mouse models. onist picrotoxin108 were used for the treatment of mouse
When studying genotype–phenotype correlations, phenotypes due to partial trisomy 16.
the specific regulation of gene expression in different
mouse strains and the polymorphic variability of extra iPSCs with T21
chromosomal material should be taken into account. An emerging cellular model approach for the study of
Mouse model studies may also be limited by the differ- the developmental effects of T21 involves the generation
ential expression of mouse and human genes that results of iPSCs and subsequent differentiation to various cell
from differences in the regulatory landscapes between types and tissues. This experimental approach could
the two species. provide information on the impact of the supernumer-
Current mouse models are useful for studying not ary HSA21 in neural, haematopoietic and cardiac devel-
only the cognitive defects of DS, but also additional opment, and in other processes, such as trophoblast
phenotypes, including CHDs and dysmorphic features. formation. In addition, iPSC lines upon differentiation
For example, the most recent results of studies using could be used to test drugs or other chemicals that could
the mouse models in BOX 2 have established that the correct the developmental defects of T21. All T21 iPSC
region between miR-802 and Zbtb21 that contains 39 lines thus far were generated using fibroblasts from
genes is sufficient to cause CHDs when present in three patients, or amniotic fluid cells from pregnancies, and
copies88; further analyses of these models suggests that appropriate controls were used in most studies. The first
three copies of at least two causative genes in this inter- such T21 iPSC line was reported in 2008; since then, 12
val are required for the appearance of this developmental additional iPSC lines have been published34,37,38,75,109–119
defect88.
Finally, mouse models can be used for the initial (see TABLE 1 for a list of published iPSC lines and the
evaluation of treatment modalities before clinical trials main observations per study). The continuation of
in humans; for example, the epigallocatechin‑3‑gallate development of T21 iPSCs, central banking of these

NATURE REVIEWS | GENETICS VOLUME 18 | MARCH 2017 | 157


©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

cells and their distribution to researchers with appro- were to reduce neurodegeneration using antioxidants,
priate projects will certainly enhance our understanding interfere with dysfunctional signalling pathways, and
of some of the developmental processes that are affected to reduce the level or function of proteins encoded by
in T21. Because of the variability of iPSC lines produced HSA21 genes.
even from a single sample, more than one such iPSC As postnatal therapies are of more practical impor-
line per sample should be produced in order to partially tance than prenatal therapies, I briefly discuss below five
account for this limitation. different approaches for cognitive impairment therapy
that have generated considerable interest in the past
Therapy for DS? couple of years.
Pharmacotherapy
The prospect of therapy for DS was unimaginable a dec- The inhibitory GABAergic system. Ts65Dn mice dis-
ade ago. However, the initial knowledge of the protein-­ play excessive GABA-mediated inhibition, so drugs
coding gene content of HSA21, the continuous efforts that affect the GABAA receptors are a logical choice for
to understand the signalling and other metabolic and the therapy of cognitive impairment in DS. Negative
developmental pathways that are dysregulated in DS, modu­lators (antagonists) of GABAA receptors (picro-
the availability of mouse models, the generation of T21 toxin (PTX) and phenyltetrazole) were first used in
iPSCs and their differentiation to various cell types 2007 in the Ts65Dn DS mouse model, and resulted
and tissues, and the understanding of the numerous in the recovery of cognition and long-term potenti-
neurodevelopmental alterations, now provide some ation108. These compounds were known to have pro-
enthusiasm for potential therapeutic modalities. In convulsant and anxio­genic effects, and attention has
addition, the EOAD that develops in the majority of subsequently focused instead on selective negative
DS individuals (who survive long enough) could serve allosteric regulators of the GABAAa5 receptor 122, such
as a model to understand non‑T21 AD, and provides as compound RO4938581, which reduces GABAAa5
a strong rationale for the pharmaceutical industry receptor-mediated inhibition and rescues functional
to focus some efforts on the therapy of the different and neuromorphological deficits in Ts65Dn mice123.
­neurodevelopmental manifestations of DS. A related compound, RO5186582 (Basmisanil; Roche)
The Ts65Dn mouse model has been extensively used has now completed phase II clinical studies in humans.
for preclinical studies of experimental pharmaceutical The preliminary results of these ­trials were recently
therapies. It is likely that more recently described novel announced (Statement on CLEMATIS trial; see Further
mouse models will be used in the future, as the Ts65Dn information), but unfortunately there was no differ-
model carries an additional partial trisomy for MMU17, ence observed in primary and secondary end points on
and also has the problem of male sterility, which makes improvement in cognition between adults (18–30 years
it somewhat difficult to maintain a colony. of age) and adolescents (12–17  years of age). The
The current focus of the pharmacological treatment ­ongoing clinical trial in children (6–11 years of age) has
of DS is on the improvement of the cognitive impairment been discontinued. More detailed information is needed
that is probably due to neurodevelopmental alterations, to evaluate this clinical trial and to plan for the future of
neurotransmitter alterations and neuro­degeneration, this type of approach.
and is also targeted to the overexpression of selected
genes on HSA21. Kinase inhibitors of DYRK1A. The DYRK1A gene on
Ideally, future therapeutic efforts should target the HSA21 has been implicated in several phenotypic fea-
embryonic and/or fetal period, as the neurogenesis and tures of DS and Ts65Dn mice. DYRK1A is a tyrosine
neuronal migration in the neocortex takes place dur- phosphorylation-regulated kinase, the overexpression
ing this period. Importantly, however, neuro­genesis of which is involved in neuronal deficits, atrophy of
and neuronal migration in the hippocampal dentate dendritic spines, spine dysgenesis, precocious AD-like
gyrus and the cerebellum continue to occur after birth; neurodegeneration and cognitive deficits. It has thus
remarkably, neurogenesis in the hippocampus120 contin- become an attractive target for therapy, as several com-
ues at a slow rate throughout life. Thus, treatment may be pounds could inhibit its kinase activity with differ-
beneficial not only after birth, but also later in life. This ent specificity 95. The kinase inhibitors that have been
is particularly relevant to hippocampal neuro­genesis, as ­studied so far in Ts65Dn mice include the polyphenols
most of the mouse therapies have so far been focused in green tea124 and the purified polyphenol epigallo-
on types of learning and memory that are related to the catechin‑3‑gallate (EGCG), both of which have shown
hippocampal circuit physiology. A detailed description promising results125. Furthermore, EGCG reverses the
of the mouse preclinical t­ herapeutic studies in general is developmental abnormalities of iPSC differentiation in
provided in REF. 121. T21 (REF. 126). Two clinical trials using EGCG in par-
Some of these therapies were directed at antagoniz- ticipants 14–29 (REF. 127) and 16–34 (REF. 128) years of
ing GABAergic transmission to diminish the excessive age have been completed. The results of the latter trial,
inhibition of neurotransmission that is seen in DS, which was carried out in Catalonia, Spain, have recently
whereas others were aimed at enhancing the function of been published129, and reported that EGCG combined
the NMDA receptor, serotonergic signalling, the endo- with 12 months of cognitive training was significantly
cannabinoid system, noradrenergic transmission and more effective than placebo and cognitive training
cholinergic transmission. Other therapeutic objectives at improving visual recognition memory, inhibitory

158 | MARCH 2017 | VOLUME 18 www.nature.com/nrg


©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

control and adaptive behaviour; there were positive non‑T21 adults with AD using formoterol, a long-acting
results in two of the 15 tests in the cognitive evaluation β2 adrenergic agonist based on these preclinical results
battery of tests and one of nine adaptive skills in the is under way 141.
Adaptive Behaviour Assessment System II129. It is not yet In conclusion, there are now several theoretical and
clear whether these results are over and above the null practical therapeutic scenarios, and undoubtedly addi-
expectation. As the authors suggest, phase III trials will tional ones will be proposed in the future. A scientifically
be needed to assess and confirm the long-term efficacy rigorous review of the existing data, and cautious enthu-
of EGCG and cognitive training. siasm will probably lead to well thought out clinical
trials with defined outcomes and objectives. The meas-
Activators of the Sonic Hedgehog pathway. The cere­ urable and objective outcomes need to be standardized
bellum in DS and Ts65Dn mice is smaller than in nor- and debated for the different studies to be comparable.
mal controls and contains a significantly decreased International collaborations among scientists and physi-
number of granule and Purkinje cell neurons. This cians, as well as the participation of patient groups, and
pheno­typic characteristic has been functionally linked coordination among the funding agencies and industry
to reduced mitogenic response to Sonic Hedgehog will benefit the patients and families on this exciting
(Shh) signalling 130. Treatment of the Ts65Dn trisomic road ahead.
mice with SAG 1.1, a synthetic activator of the Shh path-
way restored the number of granule cell precursors and Silencing of the extra chromosome 21: the XIST effect
mitotic cells to the levels found in euploid littermates130. The therapeutic dream in DS is to be able to silence, in all
A follow-up study 131 using a single treatment of new- cells, the expression of all genes on the supernumerary
born mice with SAG 1.1 resulted in normal cerebellar HSA21 and to eliminate the trans-regulatory influence
morphology in adults; remarkably, this treatment also of this extra chromosome. One extraordinary labora-
resulted in behavioural improvements and normal- tory step towards this audacious goal was the silencing
ized performance in tasks for learning and memory. of most of the protein-coding genes on one of the three
However, despite the restored cerebellar morphology, copies of chromosome 21 in T21 pluripotent stem cells,
this treatment did not rescue the cerebellum-dependent which was achieved by the introduction of the XIST
motor learning deficits. Currently, there are no known gene into one HSA21 in the cultured trisomic cells.
clinical trials in DS using Shh activators. XIST, the X‑inactivation non-coding gene, is located
on the X‑chromosome and is exclusively expressed on
Serotonergic system, neurogenesis and dendritic spines. the inactive X‑chromosome in female cells. The RNA
The neurogenesis defect and reduced dendritic spine product of XIST accumulates across the interphase
size in DS and Ts65Dn mice132 have also been linked to inactive X‑chromosome and, through numerous hetero­
the serotonergic system and reduced expression of the chromatin modifications, silences the transcription of
serotonin 5‑HT1A receptor (HTR1A)133. Treatment of most of the genes located there142. Remarkably, the
these mice with fluoxetine (a selective serotonin reup- introduction of a single copy of XIST into the DYRK1A
take inhibitor) resulted in the complete rescue of defec- locus of one of the three copies of chromosome 21 in T21
tive proliferation in the hippocampal dentate gyrus, pluripotent stem cells resulted in the silencing of most
subventricular zone, striatum and neocortex, and a protein-­coding genes on the XIST-bearing HSA21 with-
complete recovery of memory performance134. Several out affecting the expression of genes elsewhere on the
additional anatomical and physio­logical deficiencies genome117. These results provide the technical possibility
were fully rescued 1–3 months after treatment 135–137. of correcting the consequences of T21 (and potentially
These results provide a substantial body of prelimi- other full-chromosome trisomies) by embryo manipu-
nary evidence for the initiation of clinical trials in DS. lation. It is not known whether the introduction of XIST
As fluoxetine is an approved drug for the treatment of could rescue the phenotypic effects in mouse models of
major depressive disorder, obsessive-­compulsive dis­ DS as no such studies have yet been published. Even if
order and panic disorder, it would be faster and easier such mouse experiments are successful, and are with-
to plan and perform such clinical trials in DS compared out considerable side effects, genome editing in human
with trials of unapproved or novel therapies. embryos using the current CRISPR‑Cas9 technology
is far from acceptable, and a societal debate needs to
Cholinergic system degeneration. BFCN degenera- focus on this potential therapeutic approach that could
tion is a common finding in DS and AD, as well as in ­theoretically be used in preimplantation embryos.
the Ts65Dn mouse model; this degeneration is linked
to defects in explicit memory function and attention Deep brain stimulation?
and working memory. Protection of BFCN degenera- As mentioned above, studies on Ts65Dn mice revealed
tion by a high concentration of choline in pregnancy that one of the problems of DS lies with defects in
has been proposed as a potential therapeutic inter- hippocampus-dependent learning and memory 143–147.
vention. Prenatal treatment with choline in pregnant Experimental studies in humans with AD and rats with
mice resulted in the partial enhancement of cognitive amnesia have shown that deep brain stimulation (DBS)
function and hippocampal neurogenesis in Ts65Dn off- to the fimbria-fornix could improve this phenotype148,149.
spring 138–140. The mechanism underlying these obser- In a recent study 150, forniceal DBS rescued spatial learn-
vations remains unclear. However, a clinical trial in ing and memory, and restored in vivo hippocampal

NATURE REVIEWS | GENETICS VOLUME 18 | MARCH 2017 | 159


©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

initiated a vast field of research, and created a new


medical specialty, but the search for understanding
Environment Chromosome and treatment is far from completion. The advances in
21 DNA the functional exploration of the genome, the reduc-
tion of the cost of sequencing DNA and RNA, the new
methods to study the brain, the ability to study the
Functional
Stochastic elements genomics of single cells, and the interest of the pharma­
events of Chr21 ceutical industry, provide renewed excitement for a
Down renaissance of translational and fundamental research
syndrome
phenotypes in DS.
Epigenetic Future research objectives that reflect the opinion
modifications Variability
(non-genetically of Chr21
of the author are presented below (note that several
controlled) other research objectives could be added, and thus the
list is not exhaustive). These objectives include: first,
Variability to understand the function of all functional elements
Chromatin of other
structure of HSA21, and to identify which functional genomic
chromosomes
elements in three copies have a phenotypic impact.
Second, to identify the genomic determinants of the
phenotypic variability in DS. The proposed Down
Figure 5 | Determinants contributing to the Down Syndrome Genomes Project (see BOX 1) aims to accom-
syndrome phenotypes. The relative Nature Reviews
weight | Genetics
of these plish this task. Third, to identify the influence of T21
determinants, and their interactions are not shown as on the different cell types and tissues. Mouse models are
these are not fully known. Chromosome 21 DNA: the
the most appropriate proxies for such purposes. Fourth,
presence of additional chromosomal material in the
nucleus that is approximately 45 Mb long. Functional to understand the impact of genome dosage imbalance
elements of Chr21: all the genomic regions that contain at the proteome level, as there is considerable ‘buffering’
a function, including protein-coding genes, non-coding from the transcriptome to proteome. Fifth, to intensify
RNAs and regulatory elements. Variability of Chr21: the research for drug therapies; the molecular targets for DS
nucleotide variability of this chromosome that could therapies are expanding and single drug trials, or com-
influence the function of each element. Variability of binations of drugs that act on different targets, could
other chromosomes: the nucleotide variability be tested. Sixth, to develop in vitro tissues (organoids
throughout the genome that has functional such as those described in REF. 153) from T21, and to
importance. Chromatin structure: the particular explore them as experimental models. Seventh, to use
chromatin topology and composition. Epigenetic
iPSCs from T21 and to differentiate them to different
modifications: particularly those that are not determined
by the primary DNA sequence. cell types. A considerable number of iPSC lines ori­
ginating from different individuals are needed to also
study the impact of genetic variation and epigenetic
modifications on the phenotypic variability. The iPSC
long-term potentiation and hippo­campal neuro­genesis lines complement the mouse models, as they could help
in a mouse model of Rett syndrome (a genetic dis­order us to understand phenotypic consequences that cannot
with intellectual disability due to loss-of-function vari­ be studied in mice. Eighth, to use all the genomic and
ants in the MECP2 gene on the X-chromosome151). chromatin biochemical and epigenetic marks and to
This successful, albeit preliminary, non-gene-specific study the chromatin structure and topology in order
intervention in the Rett mouse model, now raises the to understand the genome regulation in the different
question: could forniceal DBS be successful in treat- cells or tissues. This genomic regulation is likely to be
ing similar hippocampal-related defects in the mouse determined by the disomic genome-wide variation,
­models of DS? and the trisomic genomic variation on HSA21. Ninth,
the continuous study of the brain development and
A vaccine against the Aβ peptides of APP? function is important to understand the intellectual
An immunotherapy to reduce the levels of the Aβ amy- disability of DS, and the frequent EOAD. Tenth, the
loid peptides has been tested in the Ts65Dn mouse metabolomes of different tissues (that is, the chemicals
model. The anti‑Aβ vaccine was given at 5 months of with qualitative and quantitative alterations) in DS need
age, and resulted in a modest, nonsignificant reduction to be studied not only to identify biomarkers for differ-
in brain Aβ levels152; however, there was an improve- ent phenotypes but also to understand the pathophysi­
ment of memory deficits, and reduction of cholinergic ology of the various phenotypes of DS. An unbiased
neuron atrophy. These initial results provide yet another approach may be necessary, as the tissue of primary
­therapeutic option that could be further explored. interest is not always obvious. The prime example of
phenylketonuria in which the resulting intellectual dis-
Perspectives ability is due to the abnormality of an enzyme deficiency
The phenotypic expression and variability of T21 is in the liver and not in the brain is compelling; thus,
complex and complicated (as shown schematically in out‑of‑the-box hypotheses may provide unexpected
FIG. 5). The discovery of the extra HSA21 in DS has breakthroughs in DS.

160 | MARCH 2017 | VOLUME 18 www.nature.com/nrg


©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

1. Langdon-Down, J. Observations on an ethnic 26. Lappalainen, T. et al. Transcriptome and genome and causes cholinergic neuron degeneration. Neuron
classification of idiots. London Hospital Reports 3, sequencing uncovers functional variation in humans. 51, 29–42 (2006).
259–262 (1866). Nature 501, 506–511 (2013). 51. Granholm, A. C., Sanders, L. A. & Crnic, L. S. Loss of
First description of the phenotype of DS. 27. Gutierrez-Arcelus, M. et al. Passive and active DNA cholinergic phenotype in basal forebrain coincides
2. LeJeune, J., Gautier, M. & Turpin, R. Etudes des methylation and the interplay with genetic variation with cognitive decline in a mouse model of Down’s
chromosomes somatiques de neufs enfants in gene regulation. eLife 2, e00523 (2013). syndrome. Exp. Neurol. 161, 647–663 (2000).
mongoliens. C. R. Hebd Seances Acad. Sci. 248, 28. Gutierrez-Arcelus, M. et al. Correction: passive and 52. Whitehouse, P. J. et al. Alzheimer’s disease and senile
1721–1722 (in French) (1959). active DNA methylation and the interplay with genetic dementia: loss of neurons in the basal forebrain.
Identification of the extra chromosome as the variation in gene regulation. eLife 2, e01045 (2013). Science 215, 1237–1239 (1982).
genomic cause of DS. 29. Kilpinen, H. et al. Coordinated effects of sequence 53. Yates, C. M., Simpson, J., Maloney, A. F., Gordon, A.
3. Davisson, M. T., Schmidt, C. & Akeson, E. C. variation on DNA binding, chromatin structure, and & Reid, A. H. Alzheimer-like cholinergic deficiency in
Segmental trisomy of murine chromosome 16: transcription. Science 342, 744–747 (2013). Down syndrome. Lancet 2, 979 (1980).
a new model system for studying Down syndrome. 30. Waszak, S. M. et al. Population variation and genetic 54. Cataldo, A. M. et al. Endocytic pathway abnormalities
Prog. Clin. Biol. Res. 360, 263–280 (1990). control of modular chromatin architecture in humans. precede amyloid β deposition in sporadic Alzheimer’s
Description of the most widely used mouse model Cell 162, 1039–1050 (2015). disease and Down syndrome: differential effects of
of DS. 31. Olmos-Serrano, J. L. et al. Down syndrome APOE genotype and presenilin mutations.
4. Hattori, M. et al. The DNA sequence of human developmental brain transcriptome reveals defective Am. J. Pathol. 157, 277–286 (2000).
chromosome 21. Nature 405, 311–319 (2000). oligodendrocyte differentiation and myelination. 55. Xu, W. et al. Amyloid precursor protein-mediated
The landmark paper on the sequence of the long Neuron 89, 1208–1222 (2016). endocytic pathway disruption induces axonal
arm of chromosome 21. 32. Prandini, P. et al. Natural gene-expression variation dysfunction and neurodegeneration. J. Clin. Invest.
5. Antonarakis, S. E. 10 years of genomics, chromosome in Down syndrome modulates the outcome of gene- 126, 1815–1833 (2016).
21, and Down syndrome. Genomics 51, 1–16 (1998). dosage imbalance. Am. J. Hum. Genet. 81, 252–263 56. Cossec, J. C. et al. Trisomy for synaptojanin1 in Down
6. Antonarakis, S. E. Chromosome 21: from sequence to (2007). syndrome is functionally linked to the enlargement of
applications. Curr. Opin. Genet. Dev. 11, 241–246 33. Dahoun, S. et al. Monozygotic twins discordant for early endosomes. Hum. Mol. Genet. 21, 3156–3172
(2001). trisomy 21 and maternal 21q inheritance: a complex (2012).
7. Antonarakis, S. E. & Epstein, C. J. The challenge of series of events. Am. J. Med. Genet. A 146A, 57. Zhu, L. et al. Reduction of synaptojanin 1 accelerates
Down syndrome. Trends Mol. Med. 12, 473–479 2086–2093 (2008). Aβ clearance and attenuates cognitive deterioration
(2006). 34. Letourneau, A. et al. Domains of genome-wide gene in an Alzheimer mouse model. J. Biol. Chem. 288,
8. Antonarakis, S. E., Lyle, R., Dermitzakis, E. T., expression dysregulation in Down’s syndrome. Nature 32050–32063 (2013).
Reymond, A. & Deutsch, S. Chromosome 21 and 508, 345–350 (2014). 58. Hasle, H., Clemmensen, I. H. & Mikkelsen, M.
down syndrome: from genomics to pathophysiology. The study that established the genome-wide Risks of leukaemia and solid tumours in individuals
Nat. Rev. Genet. 5, 725–738 (2004). transcriptome dysregulation due to trisomy 21. with Down’s syndrome. Lancet 355, 165–169
9. Letourneau, A. & Antonarakis, S. E. Genomic 35. Guelen, L. et al. Domain organization of human (2000).
determinants in the phenotypic variability of Down chromosomes revealed by mapping of nuclear lamina 59. Zipursky, A., Poon, A. & Doyle, J. Leukemia in Down
syndrome. Prog. Brain Res. 197, 15–28 (2012). interactions. Nature 453, 948–951 (2008). syndrome: a review. Pediatr. Hematol. Oncol. 9,
10. Hartley, D. et al. Down syndrome and Alzheimer’s 36. Pope, B. D. & Gilbert, D. M. Genetics: up and down 139–149 (1992).
disease: common pathways, common goals. in Down’s syndrome. Nature 508, 323–324 (2014). 60. Al‑Kasim, F. et al. Incidence and treatment of
Alzheimers Dement. 11, 700–709 (2015). 37. Hibaoui, Y. & Feki, A. Human pluripotent stem cells: potentially lethal diseases in transient leukemia of
11. Wiseman, F. K. et al. A genetic cause of Alzheimer applications and challenges in neurological diseases. Down syndrome: Pediatric Oncology Group Study.
disease: mechanistic insights from Down syndrome. Front. Physiol. 3, 267 (2012). J. Pediatr. Hematol. Oncol. 24, 9–13 (2002).
Nat. Rev. Neurosci. 16, 564–574 (2015). 38. Briggs, J. A. et al. Integration-free induced pluripotent 61. Hitzler, J. K. & Zipursky, A. Origins of leukaemia in
12. Dierssen, M. Down syndrome: the brain in trisomic stem cells model genetic and neural developmental children with Down syndrome. Nat. Rev. Cancer 5,
mode. Nat. Rev. Neurosci. 13, 844–858 (2012). features of down syndrome etiology. Stem Cells 31, 11–20 (2005).
13. Reeves, R. H., Baxter, L. L. & Richtsmeier, J. T. Too 467–478 (2013). 62. Mateos, M. K., Barbaric, D., Byatt, S. A., Sutton, R.
much of a good thing: mechanisms of gene action in 39. Sullivan, K. D. et al. Trisomy 21 consistently & Marshall, G. M. Down syndrome and leukemia:
Down syndrome. Trends Genet. 17, 83–88 (2001). activates the interferon response. eLife 5, e16220 insights into leukemogenesis and translational
14. Reeves, R. H. & Garner, C. C. A year of unprecedented (2016). targets. Transl Pediatr. 4, 76–92 (2015).
progress in Down syndrome basic research. 40. Tan, Y. H., Schneider, E. L., Tischfield, J., Epstein, C. J. 63. Buitenkamp, T. D. et al. Acute lymphoblastic leukemia
Ment. Retard. Dev. Disabil. Res. Rev. 13, 215–220 & Ruddle, F. H. Human chromosome 21 dosage: effect in children with Down syndrome: a retrospective
(2007). on the expression of the interferon induced antiviral analysis from the Ponte di Legno study group. Blood
15. Salehi, A., Faizi, M., Belichenko, P. V. & Mobley, W. C. state. Science 186, 61–63 (1974). 123, 70–77 (2014).
Using mouse models to explore genotype–phenotype 41. Wisniewski, K. E., Wisniewski, H. M. & Wen, G. Y. 64. Mullighan, C. G. et al. Rearrangement of CRLF2 in
relationship in Down syndrome. Ment. Retard. Dev. Occurrence of neuropathological changes and B‑progenitor- and Down syndrome-associated acute
Disabil. Res. Rev. 13, 207–214 (2007). dementia of Alzheimer’s disease in Down’s syndrome. lymphoblastic leukemia. Nat. Genet. 41, 1243–1246
16. Lo, Y. M. et al. Presence of fetal DNA in maternal Ann. Neurol. 17, 278–282 (1985). (2009).
plasma and serum. Lancet 350, 485–487 (1997). One of the first reports of neuropathology similar 65. Hertzberg, L. et al. Down syndrome acute
17. Chiu, R. W. et al. Noninvasive prenatal diagnosis of to that of Alzheimer dementia in DS. lymphoblastic leukemia, a highly heterogeneous
fetal chromosomal aneuploidy by massively parallel 42. Wu, J. & Morris, J. K. The population prevalence of disease in which aberrant expression of CRLF2 is
genomic sequencing of DNA in maternal plasma. Down’s syndrome in England and Wales in 2011. associated with mutated JAK2: a report from the
Proc. Natl Acad. Sci. USA 105, 20458–20463 Eur. J. Hum. Genet. 21, 1016–1019 (2013). International BFM Study Group. Blood 115,
(2008). 43. Rumble, B. et al. Amyloid A4 protein and its precursor 1006–1017 (2010).
18. Wong, F. C. & Lo, Y. M. Prenatal diagnosis innovation: in Down’s syndrome and Alzheimer’s disease. 66. Nikolaev, S. I. et al. Frequent cases of RAS-mutated
genome sequencing of maternal plasma. Annu. Rev. N. Engl. J. Med. 320, 1446–1452 (1989). Down syndrome acute lymphoblastic leukaemia lack
Med. 67, 419–432 (2016). One of the first studies of amyloid deposition in JAK2 mutations. Nat. Commun. 5, 4654 (2014).
19. Consortium, E. P. An integrated encyclopedia of DNA the brains of AD and DS. 67. Lane, A. A. et al. Triplication of a 21q22 region
elements in the human genome. Nature 489, 57–74 44. Lemere, C. A. et al. Sequence of deposition of contributes to B cell transformation through HMGN1
(2012). heterogeneous amyloid β-peptides and APO E in overexpression and loss of histone H3 Lys27
20. Emmrich, S. et al. miR‑99a/100~125b tricistrons Down syndrome: implications for initial events in trimethylation. Nat. Genet. 46, 618–623 (2014).
regulate hematopoietic stem and progenitor cell amyloid plaque formation. Neurobiol. Dis. 3, 16–32 68. Roy, A., Roberts, I., Norton, A. & Vyas, P. Acute
homeostasis by shifting the balance between TGFβ (1996). megakaryoblastic leukaemia (AMKL) and transient
and Wnt signaling. Genes Dev. 28, 858–874 45. Hooli, B. V. et al. Role of common and rare APP DNA myeloproliferative disorder (TMD) in Down syndrome:
(2014). sequence variants in Alzheimer disease. Neurology a multi-step model of myeloid leukaemogenesis.
21. O’Loghlen, A. et al. MicroRNA regulation of Cbx7 78, 1250–1257 (2012). Br. J. Haematol. 147, 3–12 (2009).
mediates a switch of Polycomb orthologs during ESC 46. Theuns, J. et al. Promoter mutations that increase 69. Roberts, I. et al. GATA1‑mutant clones are frequent
differentiation. Cell Stem Cell 10, 33–46 (2012). amyloid precursor-protein expression are associated and often unsuspected in babies with Down
22. Bamburg, J. R. & Bloom, G. S. Cytoskeletal with Alzheimer disease. Am. J. Hum. Genet. 78, syndrome: identification of a population at risk of
pathologies of Alzheimer disease. Cell Motil. 936–946 (2006). leukemia. Blood 122, 3908–3917 (2013).
Cytoskeleton 66, 635–649 (2009). 47. Prasher, V. P. et al. Molecular mapping of Alzheimer- 70. Yoshida, K. et al. The landscape of somatic mutations
23. Letourneau, A. et al. HSA21 single-minded 2 (Sim2) type dementia in Down’s syndrome. Ann. Neurol. 43, in Down syndrome-related myeloid disorders.
binding sites co‑localize with super-enhancers and 380–383 (1998). Nat. Genet. 45, 1293–1299 (2013).
pioneer transcription factors in pluripotent mouse 48. GTEx Consortium. Human genomics. the Genotype- 71. Elagib, K. E. et al. RUNX1 and GATA‑1 coexpression
ES cells. PLoS ONE 10, e0126475 (2015). Tissue Expression (GTEx) pilot analysis: multitissue and cooperation in megakaryocytic differentiation.
24. Costa, V. et al. Massive-scale RNA-Seq analysis of gene regulation in humans. Science 348, 648–660 Blood 101, 4333–4341 (2003).
non ribosomal transcriptome in human trisomy 21. (2015). 72. Stankiewicz, M. J. & Crispino, J. D. AKT collaborates
PLoS ONE 6, e18493 (2011). The GTEx study revealing gene expression with ERG and Gata1s to dysregulate
25. Sethupathy, P. et al. Human microRNA‑155 on regulatory variation in numerous human tissues. megakaryopoiesis and promote AMKL. Leukemia 27,
chromosome 21 differentially interacts with its 49. Teller, J. K. et al. Presence of soluble amyloid 1339–1347 (2013).
polymorphic target in the AGTR1 3’ untranslated β-peptide precedes amyloid plaque formation in 73. Harewood, L. et al. Amplification of AML1 on a
region: a mechanism for functional single-nucleotide Down’s syndrome. Nat. Med. 2, 93–95 (1996). duplicated chromosome 21 in acute lymphoblastic
polymorphisms related to phenotypes. Am. J. Hum. 50. Salehi, A. et al. Increased App expression in a mouse leukemia: a study of 20 cases. Leukemia 17,
Genet. 81, 405–413 (2007). model of Down’s syndrome disrupts NGF transport 547–553 (2003).

NATURE REVIEWS | GENETICS VOLUME 18 | MARCH 2017 | 161


©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

74. Weber, S. et al. Gain of chromosome 21 or 98. Lek, M. et al. Analysis of protein-coding genetic 120. Zhao, C., Deng, W. & Gage, F. H. Mechanisms and
amplification of chromosome arm 21q is one variation in 60,706 humans. Nature 536, 285–291 functional implications of adult neurogenesis. Cell
mechanism for increased ERG expression in acute (2016). 132, 645–660 (2008).
myeloid leukemia. Genes Chromosomes Cancer 55, The current reference study for the frequency of 121. Stagni, F., Giacomini, A., Guidi, S., Ciani, E. &
148–157 (2016). genomic exonic variants. Bartesaghi, R. Timing of therapies for Down
75. Banno, K. et al. Systematic cellular disease models 99. Olson, L. E., Richtsmeier, J. T., Leszl, J. & syndrome: the sooner, the better. Front. Behav.
reveal synergistic interaction of trisomy 21 and Reeves, R. H. A chromosome 21 critical region does Neurosci. 9, 265 (2015).
GATA1 mutations in hematopoietic abnormalities. not cause specific Down syndrome phenotypes. 122. Braudeau, J. et al. Chronic treatment with a
Cell Rep. 15, 1228–1241 (2016). Science 306, 687–690 (2004). promnesiant GABA‑A α5‑selective inverse agonist
76. Malinge, S., Izraeli, S. & Crispino, J. D. Insights into 100. Pereira, P. L. et al. A new mouse model for the increases immediate early genes expression during
the manifestations, outcomes, and mechanisms of trisomy of the Abcg1–U2af1 region reveals the memory processing in mice and rectifies their
leukemogenesis in Down syndrome. Blood 113, complexity of the combinatorial genetic code of down expression levels in a Down syndrome mouse model.
2619–2628 (2009). syndrome. Hum. Mol. Genet. 18, 4756–4769 Adv. Pharmacol. Sci. 2011, 153218 (2011).
77. Nikolaev, S. I. et al. Exome sequencing identifies (2009). 123. Martinez-Cue, C. et al. Reducing GABAA α5 receptor-
putative drivers of progression of transient 101. Brault, V. et al. Opposite phenotypes of muscle mediated inhibition rescues functional and
myeloproliferative disorder to AMKL in infants with strength and locomotor function in mouse models of neuromorphological deficits in a mouse model of
Down syndrome. Blood 122, 554–561 (2013). partial trisomy and monosomy 21 for the proximal down syndrome. J. Neurosci. 33, 3953–3966
78. Malinge, S. et al. Development of acute Hspa13–App region. PLoS Genet. 11, e1005062 (2013).
megakaryoblastic leukemia in Down syndrome is (2015). 124. Guedj, F. et al. Green tea polyphenols rescue of brain
associated with sequential epigenetic changes. Blood 102. Li, Z. et al. Duplication of the entire 22.9 Mb human defects induced by overexpression of DYRK1A.
122, e33–e43 (2013). chromosome 21 syntenic region on mouse PLoS ONE 4, e4606 (2009).
79. Ferencz, C. et al. Congenital cardiovascular chromosome 16 causes cardiovascular and 125. Stagni, F. et al. Treatment with epigallocatechin
malformations associated with chromosome gastrointestinal abnormalities. Hum. Mol. Genet. 16, gallate rescues neurogenesis and neuron maturation
abnormalities: an epidemiologic study. J. Pediatr. 1359–1366 (2007). in the Ts65Dn mouse model of Down syndrome.
114, 79–86 (1989). 103. Liu, C. et al. Genetic analysis of Down syndrome- (Symposia abstract). Neuropsychol. Trends 16, 120
80. Roizen, N. J. & Patterson, D. Down’s syndrome. associated heart defects in mice. Hum. Genet. 130, (2014).
Lancet 361, 1281–1289 (2003). 623–632 (2011). 126. Hibaoui, Y. et al. Modelling and rescuing
81. Barlow, G. M. et al. Down syndrome congenital heart 104. Liu, C. et al. Engineered chromosome-based genetic neurodevelopmental defect of Down syndrome using
disease: a narrowed region and a candidate gene. mapping establishes a 3.7 Mb critical genomic region induced pluripotent stem cells from monozygotic
Genet. Med. 3, 91–101 (2001). for Down syndrome-associated heart defects in mice. twins discordant for trisomy 21. EMBO Mol. Med. 6,
82. Korbel, J. O. et al. The genetic architecture of Down Hum. Genet. 133, 743–753 (2014). 259–277 (2014).
syndrome phenotypes revealed by high-resolution 105. Hernandez, D., Mee, P. J., Martin, J. E., 127. US National Library of Medicine. ClinicalTrials.gov.
analysis of human segmental trisomies. Proc. Natl Tybulewicz, V. L. & Fisher, E. M. Transchromosomal https://clinicaltrials.gov/ct2/show/NCT01394796
Acad. Sci. USA 106, 12031–12036 (2009). mouse embryonic stem cell lines and chimeric mice (2010).
83. Lyle, R. et al. Genotype–phenotype correlations that contain freely segregating segments of human 128. US National Library of Medicine. ClinicalTrials.gov.
in Down syndrome identified by array CGH in chromosome 21. Hum. Mol. Genet. 8, 923–933 https://clinicaltrials.gov/ct2/show/NCT01699711
30 cases of partial trisomy and partial monosomy (1999). (2012).
chromosome 21. Eur. J. Hum. Genet. 17, 454–466 106. Duchon, A., Besson, V., Pereira, P. L., Magnol, L. & 129. de la Torre, R. et al. Safety and efficacy of cognitive
(2009). Herault, Y. Inducing segmental aneuploid mosaicism training plus epigallocatechin‑3‑gallate in young
84. Dunlevy, L. et al. Down’s syndrome-like cardiac in the mouse through targeted asymmetric sister adults with Down’s syndrome (TESDAD): a double-
developmental defects in embryos of the chromatid event of recombination. Genetics 180, blind, randomised, placebo-controlled, phase 2 trial.
transchromosomic Tc1 mouse. Cardiovasc. Res. 88, 51–59 (2008). Lancet Neurol. 15, 801–810 (2016).
287–295 (2010). 107. De la Torre, R. et al. Epigallocatechin‑3‑gallate, a 130. Roper, R. J. et al. Defective cerebellar response to
85. Yu, T. et al. A mouse model of Down syndrome DYRK1A inhibitor, rescues cognitive deficits in Down mitogenic Hedgehog signaling in Down’s syndrome
trisomic for all human chromosome 21 syntenic syndrome mouse models and in humans. Mol. Nutr. mice. Proc. Natl Acad. Sci. USA 103, 1452–1456
regions. Hum. Mol. Genet. 19, 2780–2791 (2010). Food Res. 58, 278–288 (2014). (2006).
The first mouse model for the human entire 108. Fernandez, F. et al. Pharmacotherapy for cognitive 131. Das, I. et al. Hedgehog agonist therapy corrects
trisomy 21. impairment in a mouse model of Down syndrome. structural and cognitive deficits in a Down syndrome
86. Sago, H. et al. Ts1Cje, a partial trisomy 16 mouse Nat. Neurosci. 10, 411–413 (2007). mouse model. Sci. Transl Med. 5, 201ra120 (2013).
model for Down syndrome, exhibits learning and The mouse study that provided hope for 132. Belichenko, P. V. et al. Synaptic structural
behavioral abnormalities. Proc. Natl Acad. Sci. USA pharmacotherapy in DS. abnormalities in the Ts65Dn mouse model of Down
95, 6256–6261 (1998). 109. Park, I. H. et al. Disease-specific induced pluripotent syndrome. J. Comp. Neurol. 480, 281–298 (2004).
87. Shinohara, T. et al. Mice containing a human stem cells. Cell 134, 877–886 (2008). 133. Clark, S., Schwalbe, J., Stasko, M. R., Yarowsky, P. J.
chromosome 21 model behavioral impairment and 110. Shi, Y. et al. A human stem cell model of early & Costa, A. C. Fluoxetine rescues deficient
cardiac anomalies of Down’s syndrome. Hum. Mol. Alzheimer’s disease pathology in Down syndrome. neurogenesis in hippocampus of the Ts65Dn mouse
Genet. 10, 1163–1175 (2001). Sci. Transl Med. 4, 124ra29 (2012). model for Down syndrome. Exp. Neurol. 200,
88. Lana-Elola, E. et al. Genetic dissection of Down 111. Mou, X. et al. Generation of disease-specific induced 256–261 (2006).
syndrome-associated congenital heart defects using pluripotent stem cells from patients with different 134. Bianchi, P. et al. Early pharmacotherapy restores
a new mouse mapping panel. eLife 5, e11614 karyotypes of Down syndrome. Stem Cell Res. Ther. 3, neurogenesis and cognitive performance in the
(2016). 14 (2012). Ts65Dn mouse model for Down syndrome.
89. Grossman, T. R. et al. Over-expression of DSCAM and 112. Lu, H. E. et al. Modeling neurogenesis impairment in J. Neurosci. 30, 8769–8779 (2010).
COL6A2 cooperatively generates congenital heart Down syndrome with induced pluripotent stem cells 135. Guidi, S. et al. Early pharmacotherapy with fluoxetine
defects. PLoS Genet. 7, e1002344 (2011). from trisomy 21 amniotic fluid cells. Exp. Cell Res. rescues dendritic pathology in the Ts65Dn mouse
90. Sailani, M. R. et al. The complex SNP and CNV 319, 498–505 (2013). model of down syndrome. Brain Pathol. 23, 129–143
genetic architecture of the increased risk of congenital 113. Maclean, G. A. et al. Altered hematopoiesis in trisomy (2013).
heart defects in Down syndrome. Genome Res. 23, 21 as revealed through in vitro differentiation of 136. Stagni, F. et al. Pharmacotherapy with fluoxetine
1410–1421 (2013). isogenic human pluripotent cells. Proc. Natl Acad. Sci. restores functional connectivity from the dentate
91. Ramachandran, D. et al. Genome-wide association USA 109, 17567–17572 (2012). gyrus to field CA3 in the Ts65Dn mouse model of
study of Down syndrome-associated atrioventricular 114. Chou, S. T. et al. Trisomy 21‑associated defects in down syndrome. PLoS ONE 8, e61689 (2013).
septal defects. G3 5, 1961–1971 (2015). human primitive hematopoiesis revealed through 137. Stagni, F. et al. Long-term effects of neonatal
92. van Bon, B. W. et al. Intragenic deletion in DYRK1A induced pluripotent stem cells. Proc. Natl Acad. Sci. treatment with fluoxetine on cognitive performance in
leads to mental retardation and primary USA 109, 17573–17578 (2012). Ts65Dn mice. Neurobiol. Dis. 74, 204–218 (2015).
microcephaly. Clin. Genet. 79, 296–299 (2011). 115. Li, L. B. et al. Trisomy correction in Down syndrome 138. Moon, J. et al. Perinatal choline supplementation
93. Ji, J. et al. DYRK1A haploinsufficiency causes a new induced pluripotent stem cells. Cell Stem Cell 11, improves cognitive functioning and emotion
recognizable syndrome with microcephaly, intellectual 615–619 (2012). regulation in the Ts65Dn mouse model of Down
disability, speech impairment, and distinct facies. 116. Weick, J. P. et al. Deficits in human trisomy 21 syndrome. Behav. Neurosci. 124, 346–361 (2010).
Eur. J. Hum. Genet. 23, 1473–1481 (2015). iPSCs and neurons. Proc. Natl Acad. Sci. USA 110, 139. Velazquez, R. et al. Maternal choline supplementation
94. Conrad, B. & Antonarakis, S. E. Gene duplication: 9962–9967 (2013). improves spatial learning and adult hippocampal
a drive for phenotypic diversity and cause of human 117. Jiang, J. et al. Translating dosage compensation to neurogenesis in the Ts65Dn mouse model of Down
disease. Annu. Rev. Genomics Hum. Genet. 8, 17–35 trisomy 21. Nature 500, 296–300 (2013). syndrome. Neurobiol. Dis. 58, 92–101 (2013).
(2007). The amazing cis silencing effect of XIST gene 140. Ash, J. A. et al. Maternal choline supplementation
95. Becker, W., Soppa, U. & Tejedor, F. J. DYRK1A: insertion in an autosome. improves spatial mapping and increases basal
a potential drug target for multiple Down syndrome 118. Chen, C. et al. Role of astroglia in Down’s syndrome forebrain cholinergic neuron number and size in aged
neuropathologies. CNS Neurol. Disord. Drug Targets revealed by patient-derived human-induced Ts65Dn mice. Neurobiol. Dis. 70, 32–42 (2014).
13, 26–33 (2014). pluripotent stem cells. Nat. Commun. 5, 4430 141. US National Library of Medicine. ClinicalTrials.gov.
96. Huang, N., Lee, I., Marcotte, E. M. & Hurles, M. E. (2014). https://clinicaltrials.gov/ct2/show/NCT02500784
Characterising and predicting haploinsufficiency in the 119. Murray, A. et al. Isogenic induced pluripotent stem (2015).
human genome. PLoS Genet. 6, e1001154 (2010). cell lines from an adult with mosaic Down syndrome 142. Brown, C. J. et al. The human XIST gene: analysis of a
97. Steinberg, J., Honti, F., Meader, S. & Webber, C. model accelerated neuronal ageing and 17 kb inactive X‑specific RNA that contains conserved
Haploinsufficiency predictions without study bias. neurodegeneration. Stem Cells 33, 2077–2084 repeats and is highly localized within the nucleus. Cell
Nucleic Acids Res. 43, e101 (2015). (2015). 71, 527–542 (1992).

162 | MARCH 2017 | VOLUME 18 www.nature.com/nrg


©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

143. Chakrabarti, L., Galdzicki, Z. & Haydar, T. F. Defects in 151. Amir, R. E. et al. Rett syndrome is caused by Acknowledgements
embryonic neurogenesis and initial synapse formation mutations in X‑linked MECP2, encoding methyl- The author thanks A. Fort, Y. Hibaoui, S. Nikolaev, M.
in the forebrain of the Ts65Dn mouse model of Down CpG-binding protein 2. Nat. Genet. 23, 185–188 Garieri, C. Borel, F. Santoni, Y. Herault, E. Yu, V. Tybulewicz,
syndrome. J. Neurosci. 27, 11483–11495 (2007). (1999). X. L. d’Ardhuy for information and assistance in the prepara-
144. Lott, I. T. Neurological phenotypes for Down syndrome 152. Belichenko, P. V. et al. An anti-β-amyloid vaccine tion of this manuscript. The author also thanks the GTEx
across the life span. Prog. Brain Res. 197, 101–121 for treating cognitive deficits in a mouse model of investigators for data access, and the funding agencies SNF,
(2012). Down syndrome. PLoS ONE 11, e0152471 EU, ERC, Lejeune, and ChildCare for the support of research
145. Hyde, L. A., Frisone, D. F. & Crnic, L. S. Ts65Dn mice, (2016). in our laboratory. Furthermore, the author thanks all past
a model for Down syndrome, have deficits in context 153. Lancaster, M. A. et al. Cerebral organoids model and present members of the laboratory, collaborators, and
discrimination learning suggesting impaired human brain development and microcephaly. Nature the patients and their families for their inspiration
hippocampal function. Behav. Brain Res. 118, 53–60 501, 373–379 (2013). and support.
(2001). 154. Purcell, S., Cherny, S. S. & Sham, P. C. Genetic Power
146. Kleschevnikov, A. M. et al. Hippocampal long-term Calculator: design of linkage and association genetic Competing interests statement
potentiation suppressed by increased inhibition in the mapping studies of complex traits. Bioinformatics 19, The author declares no competing interests.
Ts65Dn mouse, a genetic model of Down syndrome. 149–150 (2003).
J. Neurosci. 24, 8153–8160 (2004). 155. Altshuler, D., Daly, M. J. & Lander, E. S. Genetic
147. Galdzicki, Z. & Siarey, R. J. Understanding mental mapping in human disease. Science 322, 881–888
DATABASES
retardation in Down’s syndrome using trisomy 16 mouse (2008).
ExAC: http://exac.broadinstitute.org/
models. Genes Brain Behav. 2, 167–178 (2003). 156. Harrow, J. et al. GENCODE: the reference human
GTEx project: http://www.gtexportal.org
148. Hamani, C. et al. Memory enhancement induced by genome annotation for the ENCODE Project. Genome
gostat: http://gostat.wehi.edu.au/
hypothalamic/fornix deep brain stimulation. Res. 22, 1760–1774 (2012).
Ann. Neurol. 63, 119–123 (2008). 157. Ernst, J. et al. Mapping and analysis of chromatin FURTHER INFORMATION
149. Shirvalkar, P. R., Rapp, P. R. & Shapiro, M. L. state dynamics in nine human cell types. Nature 473, Broad Institute: http://exac.broadinstitute.org/faq
Bidirectional changes to hippocampal theta-gamma 43–49 (2011). Statement on CLEMATIS trial: http://www.roche.com/media/
comodulation predict memory for recent spatial 158. Neph, S. et al. An expansive human regulatory store/statements.htm
episodes. Proc. Natl Acad. Sci. USA 107, 7054–7059 lexicon encoded in transcription factor footprints.
(2010). Nature 489, 83–90 (2012). SUPPLEMENTARY INFORMATION
150. Hao, S. et al. Forniceal deep brain stimulation rescues 159. Jurka, J. et al. Repbase Update, a database of See online article: S1 (figure) | S2 (table) | S3 (figure)
hippocampal memory in Rett syndrome mice. Nature eukaryotic repetitive elements. Cytogenet. Genome ALL LINKS ARE ACTIVE IN THE ONLINE PDF
526, 430–434 (2015). Res. 110, 462–467 (2005).

NATURE REVIEWS | GENETICS VOLUME 18 | MARCH 2017 | 163


©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.

You might also like