news and views
overexpression in human ESCs and iPSCs,
© 2017 Nature America, Inc., part of Springer Nature. All rights reserved.
of Dux in mouse zygotes led to defective have adapted to the activation of repeats in
thus indicating that DUX proteins can activate
a large proportion of the respective EGA genes
in both species.
Mouse ESCs cycle into a ‘2C-like state’ in
which a significant proportion of the transcrip-
tome of two-cell-stage embryos is acti­vated7–9.
2C-like cells represent ~0.4% of the total
mouse ESC population and are identified by
reactivation of MERVL7, whose long terminal
repeat (LTR) is thought to drive transcription
of several EGA genes in embryos10. Both spon-
taneously arising and 2C-like cells induced
through downregulation of CAF1 (ref. 8)
displayed strong Dux upregulation. Indeed,
in both cases, the emergence of 2C-like cells
was strongly reduced upon Dux depletion.
Not withstanding, a large group of two-cell-
stage-related genes remained unaffected in
Dux-knockout cells5, suggesting that, while
DUX seems to be the main transcription fac-
tor regulating MERVL expression, other EGA
genes may be regulated by additional factors.
Using the same MERVL reporter, Hendrickson
et al.2 showed a strong induction of 2C-like
cells, which reached 10 to 74% of the popula-
tion, upon DUX overexpression. Although
DUX overexpression and CAF1 depletion
could induce 2C-like cells in which the most
differentially expressed genes were shared,
a substantial part of the elicited transcriptome
did not overlap. Whether this indicates ‘non-
specific’ changes in gene expression for different
2C-like inducer pathways or whether this points
to additional factors involved in activating the
2C path is unknown. Indeed, the observation
in ChIP–seq data that only ~25% of the DUX-
induced transcriptome is explained by direct
DUX binding would support the latter hypoth-
esis. Importantly, in analogy to the in vitro sys-
tem, De Iaco et al. demonstrated that depletion
Of giraffes’ necks and the inheritance of chromatin
states
Vincenzo Pirrotta
New work reports that both derepressed and hyper-repressed chromatin states in animals can be transmitted to
progeny for many generations. Transmission depends on genomic architecture and histone modifications.
Can acquired phenotypes, not encoded in the
DNA sequence, be inherited and transmitted
across generations? Some studies have reported
Vincenzo Pirrotta is in Molecular Biology and
Biochemistry, Rutgers University, Piscataway,
New Jersey, USA.
e-mail: pirrotta@dls.rutgers.edu
nature genetics | volume 49 | number 6 | june 2017 821
activation of tested EGA genes and impaired
developmental progression3.
Remodeling the chromatin landscape
One of the most striking findings of these
studies relates to changes in global chromatin
organization upon DUX overexpression in
mouse ESCs. Two-cell-stage embryos display
high histone mobility11 and open chromatin
accessibility, as determined by ATAC–seq12.
2C-like cells had previously been reported to
show higher chromatin accessibility at inter-
genic regions and at MERVL-binding sites8,9.
Hendrickson et al. generated ATAC–seq data
in DUX-induced 2C-like cells2. Direct com-
parison of gained peaks—corresponding to
regions that gained chromatin accessibility—
and lost peaks highlighted a beautiful correla-
tion between 2C-like cells and two-cell-stage
embryos. This provides evidence of a chroma-
tin state that is globally remodeled into one
resembling the two-cell-stage embryo by the
action of a single transcription factor. Notably,
almost half of the peaks gained were directly
bound by DUX, suggesting a role for DUX in
chromatin opening, in line with its role as a
pioneer factor13. In addition, this observation
highlights the power of 2C-like cells as an
in vitro model of mammalian embryogenesis.
Finally, Whiddon et al.4 clarified the dif-
ferences between the transcriptional changes
provoked by mouse DUX and human DUX4.
Despite their sequence divergence, both pro-
teins activated orthologous genes in myoblasts
from their respective species. However, when
overexpressed in mouse cells, mouse DUX
could activate EGA-related retrotransposons,
whereas human DUX4 failed to do so. This
difference is due to divergent sequence speci-
ficity for DNA binding, which is proposed to
transmission of epigenetic states for a few
generations—in some cases, through mecha-
nisms mediated by small RNAs. Two new
reports, one using Drosophila melanogaster1
and the other Caenorhabditis elegans2, now
show that epigenetic states can be transmitted
by chromatin mechanisms for many genera-
tions, possibly indefinitely.
cleavage-stage embryos. Whiddon et al. suggest
that the ancestral DUX protein had some capac-
ity to regulate embryonic transcription but the
extent of DUX4 and DUX activity was multiplied
through their acquisition of the ability to regulate
retrotransposons, which became the promoters of
cleavage-related genes. These observations are rel-
evant for understanding FSHD, for which a mouse
model expressing human DUX4 has been gener-
ated, and will help improve FSHD modeling.
DUX proteins thus emerge as critical regu-
lators of mammalian EGA (Fig. 1). However,
key questions remain, such as what activates
Dux expression or, more globally, what initi-
ates the first wave of transcription in mam-
mals. An open chromatin state may create
a window of opportunity for activation of
DUX and other transcription factors to
enable EGA. Alternatively, an upstream pio-
neer activator may exist. The combination of
in vivo approaches with robust in vitro cel-
lular models will be crucial to answer these
questions and advance understanding of
mammalian development.
COMPETING FINANCIAL INTERESTS
The authors declare no competing financial interests.
1. Liang, H.L. et al. Nature 456, 400–403 (2008).
2. Hendrickson, P.G. et al. Nat. Genet. 49, 925–934
(2017).
3. De Iaco, A. et al. Nat. Genet. 49, 941–945 (2017).
4. Whiddon, J.L., Langford, A.T., Wong, C.-J., Zhong, J.W.
& Tapscott, S.J. Nat. Genet. 49, 935–940 (2017).
5. Geng, L.N. et al. Dev. Cell 22, 38–51 (2012).
6. Young, J.M. et al. PLoS Genet. 9, e1003947 (2013).
7. Macfarlan, T.S. et al. Nature 487, 57–63 (2012).
8. Ishiuchi, T. et al. Nat. Struct. Mol. Biol. 22, 662–671
(2015).
9. Eckersley-Maslin, A.A. et al. Cell Rep. 17, 179–192
(2016).
10. Peaston, A.E. et al. Dev. Cell 7, 597–606 (2004).
11. Bošković, A. et al. Genes Dev. 28, 1042–1047 (2014).
12. Wu, J. et al. Nature 534, 652–657 (2016).
13. Choi, S.H. et al. Nucleic Acids Res. 44, 5161–5173
(2016).
Generating epialleles
Klosin et al.2 used C. elegans carrying a trans-
gene driven by the hsp90 promoter to compare
the effects of growth at 25 °C and 20 °C. At 25 °C,
the promoter showed considerable activity that
was retained in progeny raised at 20 °C. This
effect was much stronger when the transgene
was present in extrachromosomal multicopy
 news and views

arrays: growth at 25 °C for a single generation
resulted in increased expression for seven gen-
erations raised at 20 °C. The authors found that
memory could be transmitted by both eggs
and sperm; was inherited with the genome, not
through the cytoplasm; did not depend on small
RNAs; and was associated with loss of methy-
lation at lysine 9 of histone H3 (H3K9) in the
transgene and, in fact, repression of the trans-
gene at 20 °C required the SET-25 histone meth-
yltrasferase responsible for H3K9 trimethylation
in the embryo. Similar transmissible derepres-
sive effects on heterochromatic genes have pre-
viously been reported in Drosophila3,4.
In the Drosophila experiments, Ciabrelli
et al.1 started with flies carrying a transgenic
construct in which the white gene (required
© 2017 Nature America, Inc., part of Springer Nature. All rights reserved.
variations were not normally transmitted to the
progeny. This is not surprising: PREs are set in
their ways in specific cells and specific tissues.
Little is known about the state of Polycomb
repression in the germ line, but, in the embryo,
Polycomb target genes must be again activated
or repressed according to developmental and
tissue-specific cues, not the state of repression
in the parents. But is this entirely true?
How might chromatin configurations be
inherited?
The ‘plastic’ state in the experiments of Ciabrelli
et al. was produced by transient heterozygous
deletion of endogenous Fab-7. This set up a
heritable state in which the transgenic Fab-7
copies were visibly more closely associated with
GAGA factor (GAF) and Mod(mdg4), two
proteins containing a BTB protein–protein
association domain known to mediate loop-
ing interactions. Ciabrelli et al. found that
binding of GAF to Fab-7 was increased in the
plastic state, whereas mutations decreasing the
level of GAF partially suppressed the extreme
phenotypes of the epilines. The transmissible
epigenetic state was dependent on the PRC2
complex, which is responsible for methylation
of histone H3 at lysine 27 (H3K27). In fact,
a similar plastic state could be established by
passaging naive transgenic flies through one
generation in which only one functional copy
of the PRC2 methyltransferase E(z) gene was
present. Once again, progressive selection of
extreme phenotypes resulted in stable inheri-

for red eye pigmentation) was accompanied
by Fab-7, one of the regulatory elements of
the Drosophila Bithorax complex, consisting
of a Polycomb response element (PRE) and a
chromatin insulator. The PRE could recruit
Polycomb complexes PRC1 and PRC2 and
repress the white gene, resulting in eyes con-
taining varying proportions of white- and
red-pigmented cells (Fig. 1). After using fly
genetics to produce a state of ‘increased epige-
netic plasticity’, the authors selected progeny
in which the white transgene was either less
repressed (eyes with more red cells) or more
strongly repressed (eyes with more white cells).
They then bred the more repressed or the less
repressed flies, selecting more extreme pheno-
types each generation. After a few generations,
the extreme phenotype (entirely white or
entirely red eyes) became stable and was trans-
mitted to the progeny, thus establishing stable
epilines, which contain no relevant changes in
DNA sequence. Furthermore, epialleles had
the remarkable property of converting ‘naive’
alleles of the transgene into the same epiallelic
state, an effect that is highly reminiscent of
the well-known paramutation phenomenon
in maize5.
PRE-containing transgenes often show
pairing-dependent Polycomb repression:
when the transgene insertion is homozygous,
repression is much stronger than when a single
copy is present6. This implies a high degree
of interaction between the two allelic copies,
abetted by the high level of interphase pairing
of homologous chromosomes that is common
in dipteran insects. Unusually, however, Fab-7
transgenes also appeared to communicate
with the endogenous Fab-7 element: contact
between them could be detected by in situ
hybridization, and repression was weaker
when endogenous Fab-7 was deleted.
In a population of flies containing the same
Fab-7-white transgene, repression was vari-
able from one individual to another, but these
the remaining endogenous locus. In some way, tance, except that only derepressed states could
this closer association of the transgenic and be transmitted transgenerationally. In contrast,
endogenous Fab-7 loci rendered fluctuations decreasing PRC1 activity for one generation
in repression transmissible through the germ did not create a plastic state and did not result
line, thus allowing the selection of increasingly in transgenerational transmission.
more extreme states, which were then stably How can these remarkable results be incor-
inherited when endogenous Fab-7 was restored porated in a mechanistic account? It is not easy,
to the homozygous state. The interaction of and, with the present state of knowledge, much
Fab-7 elements is likely promoted by the insu- remains conjectural. The plastic state, perhaps
lator element adjacent to the PRE. The Fab-7 better viewed as a state that allows germline
insulator does not bind CTCF but does bind propagation of epigenetic marks, is character-
other insulator-associated factors, including ized by closer association of the two transgene
Naive state
Endogenous Fab-7
Fab-7-white Eye
transgenes pigmentation
Select more repressed
Fab-7-white, more H3K27me3
Stably inherited
epialleles
'Plastic' state Enhanced association
Select less repressed
Fab-7-white, less H3K27me3
Figure 1 A chromosome architecture that might facilitate germline transmission of an epigenetic
state. In the naive state, expression of the Fab-7-white transgenes is variegated. Deletion of one copy
of endogenous Fab-7 (white circle) results in a closer association of the remaining endogenous Fab-7
copy with the two Fab-7-white transgenes. This somehow makes the epigenetic state self-sustaining
in the germ line, even when endogenous Fab-7 is restored to the homozygous state, in subsequent
generations. A similar effect is also obtained by decreasing E(z) function. The close association of
endogenous Fab-7 and the transgenes also causes ‘paramutation’ of a naive Fab-7-white transgene.

822 volume 49 | number 6 | june 2017 | nature genetics


copies with the endogenous locus. Once
established, this association is maintained in
the epilines, both hyper-repressed and dere-
pressed (Fig. 1). In the nematode experiments
of Klosin et al., too, an increased concentration
of transgene copies in the extrachromosomal
arrays was a factor that enhanced the herita-
bility of the derepressed state. In both cases,
a methytransferase was responsible for laying
down the epigenetic mark that determined the
state of repression. In the nematode, the SET-25
methyltransferase forms a complex with a worm
HP1 protein and can bind to its own methyl
mark, trimethylated H3K9. In Drosophila, the
PRC2 complex can bind chromatin contain-
ing trimethylated H3K27 through its Esc/Eed
subunit, which stimulates its catalytic activity.
© 2017 Nature America, Inc., part of Springer Nature. All rights reserved.
In both situations, therefore, a local concen-
tration of the methyl mark enhances the self-
propagation of that mark. This is likely to vary
to a degree depending on the level of methyla-
tion as well as the local chromatin configura-
tion, as has been shown recently for Drosophila
PREs7,8. In C. elegans, the high temperature
would cause loss of methylation and, conse-
quently, a lower ability to restore methylation
in the subsequent generation. In Drosophila,
variation in the level of H3K27 methyla-
tion might occur in the germ line as well as
in the soma. Higher levels of methylation, as
Adenine N6-methylation in diverse fungi
Michael F Seidl
A DNA modification—methylation of cytosines and adenines—has important roles in diverse processes such as
regulation of gene expression and genome stability, yet until recently adenine methylation had been considered to
be only a hallmark of prokaryotes. A new study identifies abundant adenine methylation of transcriptionally
active genes in early-diverging fungi that, together with recent other work, emphasizes the importance of adenine
methylation in eukaryotes.
In many eukaryotes, heritable and reversible
chemical DNA modifications significantly
increase the information content of DNA. DNA
methylation of cytosine, 5-methylcytosine
(5mC), is important for regulation of
gene expression and for genome stability
through its suppression of transposable ele-
ments1–3. In prokaryotes, N6-methyladenine
(6mA) is the most frequent DNA modi-
fication4 and is involved in diverse pro-
cesses such as genome defense against
foreign DNA, DNA repair and replication,
Michael F. Seidl is in the Laboratory of
Phytopathology, Wageningen University and
Research, Wageningen, the Netherlands.
e-mail: michael.seidl@wur.nl
nature genetics | volume 49 | number 6 | june 2017 823
well as local clustering of the Fab-7 elements,
would facilitate propagation of the methyl
mark, resulting in reconstitution of a hyper-
repressed state in the progeny. Lower levels,
further decreased by resulting transcriptional
activity and associated H3K27 acetylation and
methylation of lysine 4 of histone H3 (H3K4),
could cause loss of methylation in the gametes
and in the subsequent generation. The para-
mutagenic ability of the epialleles might simi-
larly result when a naive copy of the transgene
is incorporated into the cluster containing an
epigenetically committed transgene copy and
endogenous Fab-7.
This brief account glides over many potential
difficulties. How is the clustered configuration
recreated in the germ line every generation
when genome architecture is dismantled every
cell cycle and is absent in the early Drosophila
embryo9? Any causative chromatin configura-
tion would have to be reconstructed on the basis
of epigenetic modifications. Clustering based
on Polycomb proteins (Polycomb ‘bodies’) and
on Trithorax (transcription factories) has been
proposed10. GAF binding and GAF-induced
clustering might be modulated by H3K27
methylation. Why, if the closely associated
state is so stably inherited, has it not prevailed
as the natural condition of Fab-7 and similar
elements? Perhaps it has. Normally, flies do not
and the regulation of gene expression5.
In contrast, the occurrence and function of
6mA in eukaryotes has been largely unex-
plored5. Recent studies in animals6,7, pro-
tists8 and green algae9 have begun to uncover
genome-wide patterns of 6mA modification
in eukaryotes. In this issue of Nature Genetics,
Igor Grigoriev and colleagues10 took advantage
of recent technological developments and ana-
lyzed 16 fungal genomes to detect abundant
6mA that is associated with transcriptionally
active genes in early-diverging fungi.
6mA in early-diverging fungi
The fungal kingdom is ~1 billion years old
and unites a fascinating set of diverse species
that inhabit most ecological niches. Fungi are
news and views
carry transgenes containing additional Fab-7
elements nor is there selection for extreme phe-
notypes. The normal condition might be one in
which endogenous Fab-7 associates with sites
that help it to maintain a neutral state, from
which it can shift to a repressed or derepressed
state, depending on the specific tissue. How
common is this kind of epigenetic inheritance,
and how frequently has it gone undetected?
Could mutations in humans cause similar
plastic states leading to transgenerational epi-
genetic inheritance? A lot of work remains to
be done, but the doors to a world of Lamarckian
inheritance are flung wide open.
COMPETING FINANCIAL INTERESTS
The author declares no competing financial interests.
1. Ciabrelli, F. et al. Nat. Genet. 49, 876–886
(2017).
2. Klosin, A., Casas, E., Hidalgo-Carcedo, C., Vavouri, T.
& Lehner, B. Science 356, 320–323 (2017).
3. Seong, K.-H., Li, D., Shimizu, H., Nakamura, R. & Ishii, S.
Cell 145, 1049–1061 (2011).
4. Öst, A. et al. Cell 159, 1352–1364 (2014).
5. Chandler, V.L. Cell 128, 641–645 (2007).
6. Kassis, J.A. Adv. Genet. 46, 421–438 (2002).
7. Laprell, F., Finkl, K. & Müller, J. Science 356, 85–88
(2017).
8. Coleman, R.T. & Struhl, G. Science http://dx.doi.
org/10.1126/science.aai8236 (2017).
9. Hug, C.B., Grimaldi, A.G., Kruse, K. & Vaquerizas, J.M.
Cell 169, 216–228 (2017).
10. Pirrotta, V. &. Li, H.-B. Curr. Opin. Genet. Develop. 22,
101–109 (2012).
taxonomically organized into eight phyla.
Dikarya include two phyla (Ascomycota and
Basidiomycota) that comprise most of the well-
known fungi, while the six remaining phyla are
jointly referred to as early-diverging fungi. In
most studied Dikarya, 5mC predominantly
occurs at transposable elements2. The preva-
lence and function of 6mA in fungal genomes,
however, was not characterized.
Identification and quantification of genome-
wide 6mA requires sensitive detection meth-
ods11. Single-molecule real-time (SMRT)
sequencing is able to determine the genomic
sequence and infer DNA modifications simul-
taneously12. Whereas SMRT sequencing has
been used to determine 6mA patterns in
hundreds of prokaryotes4, in eukaryotes, this