Reviews

Functions and mechanisms of

epigenetic inheritance in animals
Ksenia Skvortsova1, Nicola Iovino2* and Ozren Bogdanović1,3*
Abstract | The idea that epigenetic determinants such as DNA methylation, histone modifications
or RNA can be passed to the next generation through meiotic products (gametes) is long standing.
Such meiotic epigenetic inheritance (MEI) is fairly common in yeast, plants and nematodes, but its
extent in mammals has been much debated. Advances in genomics techniques are now driving
the profiling of germline and zygotic epigenomes, thereby improving our understanding of MEI
in diverse species. Whereas the role of DNA methylation in MEI remains unclear, insights from
genome-​wide studies suggest that a previously underappreciated fraction of mammalian
genomes bypass epigenetic reprogramming during development. Notably , intergenerational
inheritance of histone modifications, tRNA fragments and microRNAs can affect gene regulation
in the offspring. It is important to note that MEI in mammals rarely constitutes transgenerational
epigenetic inheritance (TEI), which spans multiple generations. In this Review , we discuss
the examples of MEI in mammals, including mammalian epigenome reprogramming, and the
molecular mechanisms of MEI in vertebrates in general. We also discuss the implications of
the inheritance of histone modifications and small RNA for embryogenesis in metazoans, with
a particular focus on insights gained from genome-​wide studies.

Transgenerational
Genetic information transmitted from parents to their become genetically assimilated; the induced phenotypic
Spanning more than two offspring underpins the inheritance of traits across gen­ traits had eventually manifested at high frequency in
generations, from F0 to F2 erations. Nevertheless, a number of examples of herit­ the selected population even without the exposure to the
and beyond. able phenotypic variation in different organisms as well initial environmental agent. Such genetic assimila­
as clinical cases in humans cannot be fully explained by tion of environmentally induced phenotypes could
Intergenerational epigenetic
inheritance Mendelian genetics, which encompasses DNA-​sequence- therefore act as a driver of evolutionary change9 and pos­
Inheritance of an epigenetic based inheritance1–3. Apart from DNA sequence, gene sibly contribute to the origin of new species. Although
trait across two generations, regulation determinants that can be transmitted through now we know that the phenomena observed by
from F0 to F1.
mitosis and meiosis, such as covalent chemical modi­ Waddington are underpinned by a genetic basis rather
fications to the DNA, histone post-​translational modifi­ than by transgenerational propagation of an acquired trait,
cations (PTMs) and diverse RNA species, can also be the precise mechanisms of such genetic assimilation
1
Genomics and Epigenetics transmitted from parental gametes to the zygote (Fig. 1). remain unknown. Such experiments, however, raise the
Division, Garvan Institute of Such factors, which comprise the layer of regulatory question of which molecular components are implicated
Medical Research, Sydney, information that is superimposed on the DNA sequence in the complex interactions between the genome and the
New South Wales, Australia.
and imparts cell-​t ype-specific function, are often environment and in their potential heritability. A num­
2
Department of Chromatin referred to as ‘epigenetic’ factors4. These epigenetic fac­ ber of studies published in the past couple of decades
Regulation, Max Planck
Institute of Immunobiology
tors are frequently invoked when discussing potentially have postulated that traumatic experiences, exposure to
and Epigenetics, Freiburg, heritable cellular responses to environmental signals in chemicals and deficiency in nutrients can alter cellular
Germany. the absence of detectable DNA sequence alterations. epigenetic states10–15 and that, in some cases, such altered
3
St Vincent’s Clinical School, The idea that particular characteristics acquired states might be transmitted to the offspring2,16–21. In this
University of New South in response to environmental exposure can be trans­ Review, we explore the principles of meiotic epigenetic
Wales–Sydney, Sydney, ferred from parents to progeny has its roots in the doc­ inheritance (MEI) in animals, which we define as epi­
New South Wales, Australia.
trines of Hippocrates5, which were later propagated by genetic inheritance through meiotic products that spans at
*e-​mail: iovino@
Jean-​Baptiste Lamarck6 and others. Work from Conrad least two generations (that is, from F0 to F1 and beyond).
ie-​freiburg.mpg.de;
o.bogdanovic@garvan.org.au Waddington 7,8 demonstrated that the exposure of MEI includes both intergenerational epigenetic inheritance
https://doi.org/10.1038/ Drosophila melanogaster pupae to ether vapour or heat and transgenerational epigenetic inheritance (TEI).
s41580-018-0074-2 shock resulted in phenotypes that through selection had Comprehensive reviews of TEI in microorganisms,

774 | december 2018 | volume 19 www.nature.com/nrm

 Reviews

a DNA methylation
DNMT1 Unmethylated CpG
Methylated CpG
DNMT3A DNMT3B

Maternal allele

Paternal allele
CpG island Gene body Enhancer Repeat ICR Imprinted
gene
b Histone modifications SETD1A, KMT2A, KMT2C,
SETD1B KMT2AB KMT2AD
PRC2 (COMPASS) (COMPASS- (COMPASS-
like) like)
SUZ12 EED
SETDB1 SUV39H1/2 EZH2
HP1 H3K9me3 H3K27me3
H3K4me3 H3K4me2 H3K4me1
MTF2

PRE
Constitutive heterochromatin Facultative heterochromatin Transcriptionally permissive chromatin

c Small non-coding RNAs

Exogenous and piRNA genes tRNA genes
miRNA genes endogenous dsRNAs (C. elegans)
DNMT2
pri-miRNA dsRNA
tRNAs
Dicer
Drosha AGO PRG-1
siRNA piRNA
pre-miRNA Target RNA

Dicer AGO RDE-1

siRNA ERGO-1 tRFs
miRNA Target RNA

RdRP

Secondary
siRNAs
?
AGO WAGO AGO

Transcriptional and post-transcriptional regulation

Fig. 1 | DNA methylation, histone post-​translational modifications and (H3K4me1), H3K4me2 and H3K4me3 characterize transcriptionally
small non-​coding RNAs. a | DNA methylation in vertebrates occurs permissive chromatin202–204 and are catalysed by SET domain-​containing
primarily at CpG dinucleotides and is catalysed by DNA methyltransferases protein 1A (SET1A)–COMPASS (complex proteins associated with Set1) or
(DNMTs)89,90. DNMT1 is involved in the maintenance of DNA methylation SET1B–COMPASS, by histone-​lysine N-​methyltransferase 2A (KMT2A)–
whereas DNMT3A and DNMT3B catalyse de novo DNA methylation186. COMPASS-​like or KMT2B–COMPASS-​like and by KMT2C–COMPASS-​like
Although the majority of CpG sites in vertebrates are methylated, or KMT2D–COMPASS-​like complexes205. c | Primary microRNAs (pri-​miRNAs)
promoter-​a ssociated CpG-​rich regions (CpG islands) are generally are cleaved by the endoribonuclease Drosha to yield precursor miRNAs
unmethylated187. Active enhancers display reduced DNA methylation132. (pre-​miRNAs)206. Pre-​miRNAs and double-​stranded RNAs (dsRNAs) are
Imprinted genes — genes expressed from only one of the parental cleaved by the endoribonuclease Dicer to produce mature miRNAs and
alleles — display allele-​specific methylation at imprinting control regions small interfering RNAs (siRNAs)206. In Caenorhabditis elegans, Argonaute
(ICRs)58. b | Trimethylation of histone H3 lysine 9 (H3K9me3) characterizes (AGO) proteins such as RNAi-​defective 1 (RDE-1) and ERGO-1 (ref.33) bind
constitutive heterochromatin 188. H3K9me is catalysed by histone to siRNAs, which leads to the recruitment of RNA-​d ependent RNA
methyltransferases SETDB1, SU(VAR)3-9 homologue 1 (SUV39H1) and polymerases (RdRPs) and their subsequent amplification to secondary
SUV39H2 (refs189,190) and is bound by heterochromatin protein 1 (HP1) siRNA34,35. Similarly , in the germ line, PIWI-​interacting RNAs (piRNAs) are
proteins191,192. H3K27me3 characterizes facultative heterochromatin193 and bound by Piwi proteins such as PRG-1 to recruit RdRPs, which results in the
is catalysed by EZH2 (enhancer of zeste homologue 2)194,195, part of the production of secondary siRNAs42,43. DNMT2 methylates tRNAs and alters
Polycomb repressive complex 2 (PRC2), which also includes SUZ12 their stability and the formation of tRNA fragments (tRFs)161,207. Processed
(refs196,197) and EED198. PRC2 is recruited by Polycomb response elements small RNAs are loaded onto AGO and worm-​specific AGO (WAGO)
(PREs) in flies 199,200 or by MTF2 (metal-​response element-​b inding proteins to exert gene silencing208. It is not yet clear how tRFs exercise
transcription factor 2) in vertebrates 201. H3K4 mono-​m ethylation their function.

NATure RevIeWs | MOLeCuLAR CeLL BIOLOGy volume 19 | december 2018 | 775

Reviews

yeast, plants and in the nematode Caenorhabditis elegans mammals) WD repeat-​containing protein 5 (WDR-5),
have been published previously22–29. Here, we discuss ASH-2 and SET-2 decreased levels of H3K4 trimethyl­
examples of MEI and epigenome reprogramming in ation (H3K4me3; a positive regulator of gene expression)
vertebrates and elaborate on the mechanisms of MEI, and resulted in approximately 20% extension of the
namely through DNA methylation, histone modifica­ C. elegans lifespan, which was inherited across several
tions, tRNA fragments (tRFs) and other small RNAs. generations51. Depletion of the H3K4me3 demethylase
Finally, we discuss the implications of parental trans­ RBR-2 abolished the transmission of extended lifespan
mission of histone modifications and small RNAs for in WDR-5-deficient worms, suggesting that TEI of pro­
embryonic development in metazoans. longed lifespan due to deficiency in H3K4 trimethyl­
ation depends on histone demethylation. Similarly,
MEI in Caenorhabditis elegans knockouts of spr-5, which codes for an orthologue of the
One of the best-​studied mechanisms of epigenetic inheri­ H3K4 dimethylation (H3K4me2) lysine-​specific histone
tance in animals is RNAi, which is thoroughly described demethylase 1A LSD1 (also known as KDM1A), result in
in the nematode C. elegans. Apart from RNAi responses a phenotype characterized by an increased incidence of
that are heritable through many generations, C. elegans sterility across generations (germline mortality)52. This
also displays robust transgenerational inheritance of phenotype is driven by the accumulation of H3K4me2,
both active and repressive histone PTMs. causing the misregulation of genes expressed during
spermatogenesis, and is indicative of TEI of H3K4me2
RNAi pathways. In C. elegans, double-​stranded RNA patterns. Repressive histone PTMs such as H3K27me3
(dsRNA) triggers RNAi, which promotes systemic mRNA and H3K9me3 can also be inherited in C. elegans.
degradation and is heritable through the germ line30. The X chromosome inactivation through Polycomb repres­
RNAi response in C. elegans is dependent on dsRNA sive complex 2 (PRC2)-mediated H3K27 trimethyl­
cleavage by the RNase III family nuclease Dicer31,32, ation can be intergenerationally transmitted from both
which results in the production of small interfering RNAs oocytes and sperm to the embryos53. Also, a recent study
(siRNAs) (Fig. 1c). siRNAs are subsequently bound by the demonstrated TEI of exposure to high temperature,
Argonaute (AGO) proteins, such as RNAi-​defective 1 which directly inhibited the H3K9 methyltransferase
(RDE-1)33. This results in the recruitment of RNA-​ SET-25; this resulted in derepression of SET-25 target
dependent RNA polymerase (RdRP) and the production loci in the germ line and was inherited for multiple
of secondary siRNAs34,35. Secondary siRNAs are loaded generations of worms raised at normal temperature54
onto worm-​specific AGO (WAGO) proteins36 that local­ (Table 1). This effect persisted for up to 14 generations on
ize to the nucleus and initiate silencing33,37. The transgen­ an integrated transgene array, whereas for endogenous
erational effect of RNAi in C. elegans is dependent on the repetitive elements, considerable changes in expression
inheritance of small RNA molecules and their amplifi­ levels were observed for up to 6 generations. How small
cation, as demonstrated by virus-​derived small RNAs38. RNAs cooperate with repressive histone PTMs in epi­
The heritable maintenance of silencing is dependent on genetic inheritance is currently unclear44,46,48,50. Although
nuclear RNAi pathways39. Additionally, very recent work the inactivation of RNAi components such as HRDE-1
has identified a highly conserved RNA helicase, ZNFX-1, or NRDE-2 did not affect the transmission of this
that together with WAGO proteins forms phase-​separated H3K9me3-mediated epigenetic memory54, very recent
nuage granules in the cytoplasm of germ cells and that data argue that H3K9me3 might be required for siRNA-​
has a role in the propagation of heritable RNAi40,41. dependent TEI that specifically targets newly evolved
Endogenously derived PIWI-​interacting RNAs (piRNAs), genes and is particularly sensitive to transgenes55.
which suppress the expression and activity of transpo­
sons in the germ line, can also trigger heritable RNAi42,43. MEI in mammals
Together with PIWI-​like protein PRG-1, piRNAs exert Although MEI is well-​d ocumented in plants
their function through the generation of secondary (Supplementary Box 1) and in C. elegans, in mammals
siRNAs42,43 (Fig. 1c). Such piRNA-​initiated silencing is such phenomena remain more of an exception than
Facultative heterochromatin heritable across generations and is dependent on nuclear the rule. This is partly because in mammals, the epi­
Condensed, transcriptionally RNAi pathways44–46. Chromatin remodelling activity and genome inherited from the gametes is heavily repro­
silent chromatin that retains repressive-​chromatin pathways, including the histone grammed upon fertilization and during the formation
the ability to decondense and
H3 lysine 9 (H3K9) methyltransferase SET-25, also par­ of primordial germ cells (PGCs)56,57 (see below). However,
license transcription within
temporal and spatial contexts.
ticipate in the maintenance of long-​term silencing44,46–48. although these epigenome remodelling processes
Interestingly, transgenerational inheritance of RNAi can reduce DNA methylation to its lowest levels during the
PIWI-​interacting RNAs continue in the absence of H3K9 trimethylation49,50 and mammalian life cycle, methylation is not erased and
(piRNAs). A class of the H3K9 methyltransferase MET-2 was recently shown re-​established with the same efficiency at all sequences,
endogenous small non-​coding
RNAs that interact with Piwi-​
to inhibit the biogenesis of heritable siRNAs, thereby thereby constituting a potential route for MEI. For
domain-containing proteins constraining transgenerational transmission of RNAi50. clarity, we use the term ‘intergenerational’ epigenetic
and have a role in inheritance if the potentially heritable effect experi­
retrotransposon silencing Histone PTM inheritance in Caenorhabditis elegans. enced by the pregnant female (F0) was observed in F1
in the germ line.
TEI of both active and repressive histone PTMs occurs and F2, and we use the term ‘transgenerational’ epi­
Primordial germ cells in the germ line of the nematode C. elegans. Mutations genetic inheritance if the effect persisted to F3 or beyond.
(PGCs). Primary germ cells that in proteins of the Trithorax methyltransferase complex Accordingly, we use ‘intergenerational’ if the effect expe­
give rise to gametes. (myeloid/lymphoid or mixed-​lineage leukaemia, in rienced by the male (F0) or a non-​pregnant female (F0)

776 | december 2018 | volume 19 www.nature.com/nrm

 Reviews

Table 1 | Inheritance of histone post-​translational modifications and small RNAs

Histone Species Maternal Function Loss-​of-function Offspring phenotype Refs
post-​ (M) or mechanism
translational paternal
modification (P)
or small inheritance
non-​coding
RNA
H3K27me3 Drosophila M Regulation of embryonic • E(z) knockdown Aberrant accumulation of H3K27ac in 137

melanogaster development and (maternal) the embryo; homeotic phenotype

establishment of proper • E(z) mutation
H3K27ac patterns at ZGA
Drosophila M, P Modulation of stably E(z) mutation Gradual establishment and inheritance 136

melanogaster (preference inherited 3D chromatin of derepressed but not repressed

for M) interactions epialleles
Caenorhabditis M, P Memory of gamete-of- Mutation of Gradual dilution of H3K27me3 from 53

elegans origin-specific repression Polycomb protein gamete-​of-origin chromosomes

MES-3
Mus musculus M Gene-​distal chromatin NA NA 122

structure?
Mus musculus M DNA-​methylation- KDM6B Biallelic expression at 18% of 66

independent maternal overexpression imprinted regions in the zygote

imprinting in the embryo
H2A.Z and Danio rerio P Deterring DNA methylation • srcap morpholino Reduction of nuclear H2A.Z(FV) levels, 128

H3K4me1 in early embryos and • Recombinant coinciding with acquisition of DNA

establishment of poised, Anp32e protein methylation
pre-​ZGA chromatin
H3K4me2 Caenorhabditis M, P Epigenetic memory? SPR-5 (KDM1) Progressive derepression of genes that 52

elegans mutation regulate spermatogenesis, defects in

oogenesis and spermatogenesis and
sterility
Mus musculus P Epigenetic memory? KDM1A H3K4me2 reduction coinciding with 150

overexpression in deregulation of genes in the early

developing sperm embryo and developmental defects
H3K4me3 Caenorhabditis M, P Lifespan regulation wdr-5, ash-2 and Reduction of H3K4me3, coinciding 51

elegans set-2 mutations with lifespan extension inherited for

several generations
Mus musculus M Broad H3K4me3 KDM5A and KDM5B Aberrant reactivation of 120,121

domains associated with overexpression H3K4me3-marked loci

transcriptional silencing
H3K9me3 Drosophila M, P Heterochromatin High-temperature- Reduction of H3K9me2, disruption of 175

melanogaster organization induced heterochromatin formation and gene

phosphorylation silencing
of Atf-2
Caenorhabditis M, P Temperature-​sensitive set-25 mutation Reactivation of SET-25-silenced 54

elegans transcriptional repression transposons

Homo sapiens P Heterochromatin build-​up NA NA 145

in early embryos
Small non-​ Mus musculus P Intergenerational Low-​protein diet Increase in the abundance of 156

coding RNAs transmission of diet-​induced sperm tRNA-​derived small RNAs

metabolic phenotypes? and decrease in let-7 miRNAs,
downregulation of genes regulated by
the retroelement MERVL in the early
embryo and altered metabolism in the
offspring
Mus musculus P Intergenerational High-​fat diet Changes in the abundance of sperm 154

transmission of diet-​induced tRNA-​derived small RNAs and increase

metabolic phenotypes? in 5mC and m2G tRNA modifications
coinciding with the deregulation of
metabolic pathway genes in the F1
offspring, glucose intolerance and
insulin resistance
Mus musculus P Intergenerational and High-​fat diet Changes in F1 sperm RNA content 155

transgenerational (tRNA fragments and miRNA) and

transmission of diet-​induced inheritance of latent metabolic
metabolic phenotypes phenotypes in F1 and F2

NATure RevIeWs | MOLeCuLAR CeLL BIOLOGy volume 19 | december 2018 | 777

Reviews

Table 1 (cont.) | Inheritance of histone post-​translational modifications and small RNAs

Histone Species Maternal Function Loss-​of-function Offspring phenotype Refs
post-​ (M) or mechanism
translational paternal
modification (P)
or small inheritance
non-​coding
RNA
Small Mus musculus M, P Execution of the first zygotic Chemical inhibition Reduced DNA synthesis and 157,163

noncoding division of sperm-​derived suppression of the first zygotic division

RNAs (cont.) miR-34c and
deletion of maternal
Dicer
Mus musculus M, P NA Mutation of KIT Increased levels of aberrant Kit RNA 153

tyrosine kinase species coinciding with white tail tip

receptor phenotype in the offspring
m2G, N2-methylguanosine; 5mC, 5-methylcytosine; Anp32e, acidic (leucine-​rich) nuclear phosphoprotein 32 family member E; Atf-2, activating transcription factor 2;
E(z), enhancer of zeste; H3K27ac, histone 3 lysine 27 acetylation; H3K4me1, H3K4 mono-​methylation; H3K4me2, H3K4 dimethylation; H3K4me3, H3K4 trimethylation;
KDM, lysine-​specific histone demethylase; MERVL , mouse endogenous retrovirus with leucine tRNA primer ; miRNA , microRNA ; ZGA , zygotic genome activation.

was observed in F1 and ‘transgenerational’ if the effect DNA methylation state of the IAP inserted in the agouti
persisted to F2 and beyond (Box 1). locus68,69. While the phenotype of the silent, methyl­
ated allele (pseudoagouti) is wild-​type dark-​brown coat
Genomic imprinting. In mammals, a small number of colour, the hypomethylated, active allele gives rise to a
genes (~125 in mice and ~100 in humans; Geneimprint) yellow coat. The spectrum of agouti phenotypes in the
are expressed in a parent-​of-origin-​specific manner, a offspring is dependent upon the phenotype of the dam.
phenomenon termed genomic imprinting58. This dif­ For example, a dam with the yellow coat phenotype is
ferential allelic expression is the result of epigenetic more likely to have yellow coat offspring68. Interestingly,
silencing of the inactive allele, a process that involves dietary supplementation of methyl donors during
DNA methylation59–62. Such imprinted genes are regu­ mid-​gestation results in the shift of agouti phenotypes
Imprinting control regions lated through imprinting control regions (Fig. 1a), which towards pseudoagouti in both F1 and F2, suggestive
cis-​regulatory regions for which are resistant to post-​fertilization DNA methylation of increased IAP methylation in the germ line and the
the allele-​specific epigenetic reprogramming, thereby forming a platform for inter­ intergenerational retention of this altered epigenetic
state mediates differential
expression of the parental
generational epigenetic inheritance. Many imprinted state16. Nevertheless, bisulfite sequencing data suggest
alleles. genes are expressed during prenatal development and that both the maternally contributed and paternally con­
are required for proper placental function63,64. Errors tributed Avy-​associated IAPs are reprogrammed during
Prader–Willi syndrome occurring during the establishment of imprinting in the early embryogenesis, albeit with different dynamics70.
A complex disorder
parental germ line were shown to account for a num­ This suggestion implies that DNA methylation might
characterized by
developmental and neural ber of behavioural and neurodevelopmental disorders, not be the primary inherited mark controlling the spec­
phenotypes that can be such as Prader–Willi syndrome and Angelman syndrome65. trum of agouti phenotypes in F1. It is currently not clear
caused by a paternal Interestingly, recent work in mouse embryos provided whether histone modifications or RNA-​based mecha­
imprinting defect. evidence for imprinting that occurs independently nisms have a role in the inheritance of the Avy epialleles;
Angelman syndrome
of DNA methylation and is associated with maternal however, recent studies have demonstrated robust inter­
A neurological disorder inheritance of the repressive modification H3K27me3 generational inheritance of H3K4me3 and H3K27me3
characterized by intellectual (see below)66. The number of genes (n = 76) regulated by histone modifications through the maternal germ line71.
disability that can be caused such non-​canonical imprinting mechanisms appears to Methyl-​donor supplementation during mid-​gestation
by a maternal imprinting
be similar to the number of genes regulated by canonical, affects the percentage of agouti phenotypes only when
defect.
DNA-​methylation-dependent imprinting. the Avy allele is contributed through the paternal germ
Bisulfite sequencing line. This finding suggests that the paternal allele might
Treatment of DNA with sodium Epigenetic reprogramming of intracisternal A par- undergo more extensive epigenetic reprogramming and
bisulfite followed by ticles. Intracisternal A particles (IAPs) are retrotrans­ is thus affected by environmental cues during the period
sequencing; determines
cytosine methylation status,
posons present in many copies (~1,000) in the mouse of PGC methylome reprogramming (embryonic day 8.5
as unmethylated cytosine is genome67. To suppress their capacity for retrotrans­ (E8.5)–E.15.5)16. Therefore, it is possible that the inheri­
converted into uracil upon this position in the germ line, IAPs are predominantly tance of the maternal Avy locus is more tightly regulated,
treatment. methylated and silenced in germline cells. Similarly to perhaps through the coordinated action of diverse gene-​
genomic imprints, IAPs can largely resist embryonic regulatory mechanisms, including histone modifications,
Epialleles
Genetically identical alleles DNA methylation reprogramming, which makes them and is not limited only to DNA methylation.
displaying distinct epigenetic potentially attractive targets for MEI studies67. One of the
modifications. best examples of MEI in mammals is the agouti viable Contribution of MEI to human diseases. Considerable
yellow (Avy) locus in mice, which harbours an insertion efforts have been made in identifying the contribution of
Epimutations
Heritable changes in gene
of the IAP retrotransposon upstream of the agouti gene epigenetic modifications to the heritability of complex
expression that are caused by that determines coat colour68. Avy expression is highly diseases. In particular, the possibility that epimutations
changes in the epigenetic state. variable among littermates and is dependent on the confer predisposition to cancer has garnered much

778 | december 2018 | volume 19 www.nature.com/nrm

 Reviews

Box 1 | Interpreting meiotic epigenetic inheritance studies in animals reported in which the methylated state was inherited
from the affected mother to one of her three sons; how­
Although many studies of meiotic epigenetic inheritance (MEI) in animals have been ever, it was then erased in his germ line74. This obser­
published to date, the field is still facing substantial challenges. Whereas MEI in plants vation demonstrates the possibility that epimutations
and nematodes encompasses a series of well-​studied phenomena24–28, MEI studies in might be passed on from parents to their offspring,
mammals have proved especially challenging to set up and to interpret owing to many
albeit infrequently, likely owing to the DNA methyl­
external confounding factors, which can easily result in the misinterpretation of acquired
data88. First, studies dealing with MEI should, where possible, incorporate robust DNA ation reprogramming that occurs post-​fertilization and
sequencing controls to exclude potential genetic contributions to the studied in PGCs. Although additional data support the (rare)
phenomena84. Another important consideration in interpreting MEI studies is the nature occurrence of intergenerational epigenetic inheritance at
and persistence of the phenotypic effect under investigation. For example, the MLH1 locus76,77, evidence for this type of heritability
if stress is applied to the paternal lineage (F0), the germ cells (sperm) that would in humans remains very limited. Finally, it should also
participate in the formation of the F1 generation likely would have also been affected by be stressed that, on the basis of currently available data,
the stress. This likelihood would make the F0–F1 transmission of phenotypic effects it is very difficult to unequivocally rule out the involve­
intergenerational, whereas for transgenerational epigenetic inheritance the phenotypic ment of other cis-​acting or even trans-​acting factors
effect would have to be preserved to F2 and beyond (see the figure). In mammals, if the in the intergenerational heritability of disease-​causing
stress is applied to a pregnant female (F0), then the fetus (F1) as well as its primordial
epimutations in humans.
germ cells, which have the potential to form the F2 generation, would be affected. In
such cases, only effects that would be visible in and beyond F3 would be perceived as
truly transgenerational (see the figure). Parental care and shared environment can also MEI of environmental cues in humans. Understanding
considerably affect offspring epigenomes185, making it difficult to distinguish between how the environment affects the human epigenome
MEI and de novo acquisition of epigenetic modifications in response to particular and how potentially acquired traits can be propagated
environmental cues. Additionally, the tissue of origin can be a major confounding factor through generations remains an area of intense investi­
in epigenetic inheritance studies. Most assays used for quantification of epigenetic gation78–81. One of the most noted studies on this topic
modifications do not discriminate between different cell types in the interrogated is the Dutch Hunger Winter study79,82,83. DNA methyl­
sample; therefore, more subtle epigenetic changes or changes present in just a small ation profiling of blood samples extracted from indi­
subpopulation of cells could easily be missed. Altogether, MEI studies in animals, and viduals who were conceived during the Dutch Hunger
particularly in mammals, frequently suffer from these methodological shortcomings, and
Winter (1944–1945) revealed small DNA methylation
it is important to be aware of these confounding factors when designing studies aimed at
demonstrating intergenerational or transgenerational inheritance of epigenetic traits. differences in the differentially methylated region (DMR)
associated with the insulin-​like growth factor 2 (IGF2)
gene82. Five CpG dinucleotides located within the IGF2
Exposed male or Intergenerational Transgenerational
non-pregnant female effects effects DMR were assessed by quantitative mass spectrometry
to reveal on average 5.2% lower DNA methylation in
the affected siblings. These results were successfully
validated (5.6% lower in the affected individuals)
by locus-​specific bisulfite sequencing approaches.
Parent Parent F1 F2 F3 Notably, only the cohort exposed to maternal malnutri­
germline (F0) progeny progeny progeny tion during early gestation but not during later stages
of pregnancy displayed these changes in DNA methyl­
Exposed pregnant Intergenerational Transgenerational ation at the IGF2 locus82. Although this understand­
female effects effects
ably invites speculation about epigenetic mechanisms
of adaptation to the environment, one has to keep in
Fetus (F1) mind that MEI studies in humans frequently suffer
from serious methodological flaws. A major limitation,
Fetus even in the presence of a testable hypothesis, is the fact
Parent F1 F2 F3
germline (F2) (F0) progeny progeny progeny
that such analyses are often performed on a complex
tissue such as blood, which consists of different cell
types. Therefore, it is difficult to exclude the possibil­
attention over the past few decades72. One of the best ity that the observed differences in DNA methylation
examples of such an epimutation is the silencing of merely reflect the differences in cell-​type composition
the mutL homologue 1 (MLH1) gene, which encodes between samples. Furthermore, even if those differ­
a DNA mismatch repair factor, through hypermethyl­ ences were truly caused by the proposed environmental
ation of its promoter. The MLH1 epimutation is wide­ stress, it remains challenging to prove that such modest
Microsatellite instability spread in somatic tissues derived from all three major changes could result in an adaptive phenotype. Finally,
Genetic instability in short embryonic lineages (ectoderm, mesoderm and endo­ a major confounding factor that is often disregarded
tandem repeats due to derm), suggesting that it has an embryonic or germline in MEI studies is the difference in DNA sequence
impaired DNA mismatch
origin 73,74. Such MLH1 epimutations contribute to composition between individuals, which is well
repair.
Lynch syndrome, which is characterized by early onset known to be a major contributor to the inheritance of
Differentially methylated of multiple cancers, including colorectal cancer and epigenetic states84–87.
region endometrial cancer, and in general cancers displaying Whereas the Dutch Hunger Winter study aimed to
(DMR). A genomic region microsatellite instability75. In the majority of cases, MLH1 assess the potential impacts of early intrauterine expo­
displaying statistically
significant change in DNA
epimutations arise de novo in affected individuals and sure to famine on DNA methylation during adult life,
methylation between at least are reprogrammed to the unmethylated state in their a more recent study tried to address whether intergener­
two samples. offspring72. Nevertheless, to date at least one case was ational inheritance of stress effects, such as those caused

NATure RevIeWs | MOLeCuLAR CeLL BIOLOGy volume 19 | december 2018 | 779

Reviews

to survivors of the Holocaust, can be demonstrated in MEI through DNA methylation

humans81. The offspring of Holocaust survivors dis­ DNA methylation has often been proposed as a MEI
played diminished DNA methylation of the FKBP5 carrier in animals owing to its relative stability and well-​
gene, which was previously linked to post-​traumatic defined mechanisms of de novo deposition and main­
stress disorder, whereas this gene was marginally tenance89,90 (Fig. 1a). Although DNA methylation is present
more methylated in the Holocaust survivors them­ in fungi, plant and animal kingdoms, its genomic abun­
selves than in the non-​affected controls. The concept dance and distribution vary considerably between spe­
of non-​genetic trauma inheritance has previously been cies91,92. For example, species with easily demonstrable
postulated in mice20; however, drawing any definite MEI, such as D. melanogaster and C. elegans, are virtu­
conclusions regarding the potential intergenerational ally devoid of DNA methylation93,94. Thus, it is unclear
epigenetic inheritance of trauma in humans would be to what extent DNA methylation contributes to MEI in
very challenging (Box 1). Apart from the methodologi­ animals and in what biological contexts. However, recent
cal issues discussed above, many of which apply also to insights from genome-​wide methylome studies suggest
this study, it is very difficult to distinguish between the that a considerable fraction of mammalian genomes
effects of parental upbringing on a growing child and might potentially bypass the removal of DNA methyl­
epigenetic inheritance per se. Owing to the genetically ation that occurs during preimplantation and PGC
hetero­geneous nature of the human population and long reprogramming (Table 2).
human lifespan, longitudinal studies that would span
multiple generations and that would be able to properly Reprogramming of the preimplantation methylome.
distinguish genotype from phenotype remain very dif­ DNA methylation is mostly static in adult somatic tis­
ficult to set up88. In the absence of those, the published sues; however, global DNA methylation patterns are
research on the topic of MEI in humans requires highly reprogrammed twice during the mammalian life
circumspect treatment. cycle56. The first reprogramming event takes place

Table 2 | Reprogramming of embryonic DNA methylation in vertebrates

Species Occurrence of Genomic regions possibly escaping post-​ Occurrence Genomic regions possibly Refs
post-​fertilization fertilization global mC reprogramming of global mC escaping reprogramming in
global 5mC reprogramming primordial germ cells
reprogramming in primordial
germ cells
Homo sapiens + ICRs, gene bodies and LINE-1 + NA 104

+ ERVK , ICRs and LINEs + NA 105

+ NA + Repeat-​free regions including 115

promoters, enhancers, gene

bodies, CGIs (n = 1,426) and
repeats (n > 100,000)
+ Exons, 3ʹ UTRs, promoters, splice sites, L1HS + Exons, 3ʹ UTRs, promoters, 113

and HERVK transposons (n = 1,471)a splice sites, L1HS and HERVK

transposons (n = 1,471)a
+ NA + LINE-1, SINEs (Alu) and ERVK 114

Mus musculus + IAPs + NA 67

+ Single-​copy germline-​expressed genes and + NA 103

somatically expressed genes (n < 20)

+ Oocyte-​specific CGIs, LINEs, LTRs and ICRs + NA 98

+ ICRs + NA 100

+ NA + IAPs, LTR–ERV1 and single-​copy 109

genes (n = 23)
+ NA + IAPs and non-IAP-related CGIs 112

(n = 265)
+ NA + Promoter CGIs (n = 11) and IAPs 110

Xenopus laevis – Global methylome inheritance NA NA 124,125,143

and Xenopus
tropicalis
Danio rerio – Global methylome inheritance (paternal NA NA 126,127

methylome configuration inherited)

5mC, 5-methylcytosine; AGCs, advanced germ cells; CGIs, CpG islands; ERVK , endogenous retrovirus K; HERVK, human endogenous retrovirus K; IAP,
intracisternal A particle; ICM, inner cell mass; ICRs, imprinting control regions; L1HS, L1 Homo sapiens-​specific; LINE, long interspersed nuclear element; LTR , long
terminal repeat; LTR–ERV1, LTR–endogenous retrovirus sequence 1; NA, not applicable; PGCs, primordial germ cells; SINEs, short interspersed nuclear elements;
UTRs, untranslated regions; +, occurrence; –, no occurrence. aGenomic regions found to commonly escape 5mC reprogramming in the blastocyst (ICM) and germ
cells (PGCs and AGCs).

780 | december 2018 | volume 19 www.nature.com/nrm

 Reviews

post-​f ertilization and involves widespread DNA patterns to those observed in mice 113–115. However,
demethylation of the paternal pronucleus, followed more than 116,000 genomic regions were identified
by a progressive drop in global DNA methylation that remained significantly hypermethylated following
in the zygote, which reaches its lowest point in the methylome reprogramming in human PGCs115. Of those,
blastocyst 95–98. The precise mechanisms of embry­ more than 7,000 regions were repeat-​poor and mostly
onic DNA methylome remodelling remain a topic coincided with CpG-​rich regions known as CpG islands,
of debate99–101; however, it is likely that both active promoters, enhancers and gene bodies. Many of these
(enzymatic) and passive (replication-​coupled dilu­ regions were associated with obesity, multiple sclerosis
tion) demethylation mechanisms have a role in this and schizophrenia. Comparisons with mouse bisulfite
process. Active DNA demethylation is mediated by sequencing data revealed 794 such escapees that were
ten-​e leven translocation (TET) enzymes, which oxidize shared between mice and humans (Table 2). Of note,
5-methylcytosine, whereas passive reduction in DNA many of these loci were linked to neural and metabolic
methylation levels is dependent on the inhibition of functions and could thus be of potential interest for fur­
DNA (cytosine-5)-methyltransferase 1 (DNMT1) activ­ ther exploration within the context of disease-​related
ity in the nucleus56. Whereas rapid post-​fertilization risk inheritance.
demethylation of the paternal pronucleus was observed In mice, metabolic deficiencies coinciding with per­
in mouse, rat, pig, bovine and human embryos, sheep turbed DNA methylation in the germ line of adults can
and rabbit preimplantation development involves less be inherited intergenerationally116. Undernourishment
demethylation of the paternal genome, indicative of of pregnant F0 dames during a critical phase of embry­
species-​specific differences in mammalian embryonic onic PGC reprogramming resulted in metabolic pheno­
epigenome remodelling102. The embryonic genome is types and modest DNA hypomethylation at discrete loci
remethylated following implantation, coinciding with in adult F1 sperm (F1 DMRs). Whereas these altered
the loss of cellular pluripotency97,98,103. Nevertheless, methylation patterns were not detectable in brain and
not all genomic loci are reprogrammed with the same liver tissues obtained from E16.5 F2 embryos, genes
efficiency. Apart from imprinted loci and IAPs, single-​ found in the vicinity of a small number of F1 DMRs had
base-resolution studies in mouse embryos revealed significantly altered transcriptional profiles in those tis­
the persistence of DNA methylation on oocyte-​ sues in F2. Indeed, several of these genes were attributed
hypermethylated DMRs and a number of long inter­ with metabolic functions, and pancreatic islets isolated
spersed nuclear element transposons during the initial from 4-month-​old F2 mice displayed impaired insu­
wave of global hypomethylation98. Similarly, a number lin secretion. The altered transcriptional state of those
of DMRs identified as hypermethylated in the oocytes genes in E16.5 embryos, in the absence of correspond­
were found to be intermediately methylated during ing DNA methylation changes, suggests that F1 sperm
human preimplantation development, suggestive of DMRs might exercise an effect during early F2 embry­
maternal DNA methylome inheritance104. Another ogenesis. This effect would not be detectable in terms
important feature of human preimplantation devel­ of altered DNA methylation; however, it would be con­
opment is the persistence of DNA methylation within veyed into an altered transcriptional output detectable
gene bodies104,105. As gene bodies are often enriched in at E16.5. Alternatively, DNA methylation might not be
enhancer elements106,107, such persistence of DNA methyl­ the primary regulatory mechanism responsible for the
ation might allow for the intergenerational inheritance formation of F1 sperm DMRs as well as for the observed
of cis-​regulatory states. In summary, these observations transcriptional changes in F2. In mammalian sperm, the
provide evidence for locus-​specific retention of DNA majority of nucleosomes are replaced with protamines.
methylation (Table  2) during a regulatory phase in However, certain loci of developmental importance,
which DNA methylation maintenance is largely inac­ such as Hox genes, promoters of developmental genes,
tive. This evidence suggests that a previously underap­ imprinted genes and microRNA (miRNA) clusters, can
preciated fraction of mammalian genomes, including retain nucleosomes and associated histone PTMs, such
promoters, gene bodies and repeats, confer resistance as H3K27me3, H3K4me2 and H3K4me3 (refs117,118).
to early DNA methylation reprogramming and thus Interestingly, 21% of the identified F1 DMRs (n = 23)
Ten-​eleven translocation constitute a potential platform for MEI. were found within such nucleosomal regions116. This
(TET) enzymes finding raises the possibility that the observed differ­
Enzymes required for active DNA methylation reprogramming in PGCs. The second ences in F1 sperm DNA methylation might merely be
DNA demethylation, which
catalyse a series of iterative
wave of mammalian DNA methylation reprogramming secondary effects of altered chromatin structure, poten­
oxidations of 5-methylcytosine takes place in PGCs and also involves a combination tially caused by defects in regulatory mechanisms other
to 5-hydroxymethylcytosine of active and passive mechanisms108–111. The hallmark than DNA methylation, such as nucleosome position­
and further to 5-formylcytosine of this process is erasure of imprinting followed by ing or histone PTMs, both of which have potential for
and 5-carboxylcytosine.
gender-​specific re-​establishment of imprinting in the intergenerational heritability119–122. In light of these
CpG islands gonads. As in the post-​fertilization demethylation findings, it is worthwhile mentioning that in many
Genomic regions of high GC process, not all genomic sequences are reprogrammed cases epigenetic inheritance might be limited only to
content and high frequency with the same efficiency109,112. Studies in mice revealed certain genomic contexts. For example, targeting tri­
of CpG sites relative to the that the feature most resilient to reprogramming are methylation of both H3K27 and H3K9 requires input
genome average; often
associated with gene
IAPs, which remain predominantly hypermethylated from specific DNA sequences85–87. Thus, it is plausible
promoters and maintained in throughout PGC demethylation109,112 (Table 2). Human that sequences that escape various rounds of epigenetic
an unmethylated state. PGCs exhibit similar genomic DNA methylation erasure reprogramming might be genetically defined.

NATure RevIeWs | MOLeCuLAR CeLL BIOLOGy volume 19 | december 2018 | 781

Reviews

Methylome reprogramming in anamniotes. Unlike Histone PTM inheritance in anamniotes. The degree of
mammals, anamniote vertebrates, such as fish and frogs, nucleosome retention in vertebrate sperm varies greatly.
do not undergo global DNA methylation remodelling Whereas zebrafish retain their entire nucleosomal con­
during embryogenesis123–127. Instead, zebrafish remodel tent138, in Xenopus laevis, only ~10% of the nucleosomes
their methylomes before zygotic genome activation (ZGA) are retained in the mature sperm139. This level is similar
in a fascinating fashion: the maternal methylome is to that in humans, whereas in mice the degree of nucleo­
reconfigured to match the paternal methylome pattern, some retention is close to 1%118. The sperm chromatin
thus making the early embryonic methylome adopt a of both zebrafish and mammals is characterized by
paternal-​like configuration at ZGA126,127. This harmoniza­ the presence of active (H3K4me2 and H3K4me3) and
tion of paternally and maternally inherited epigenetic repressive (H3K27me3) histone PTMs, thereby allow­
states before ZGA is achieved through the usage of ing for the possibility of intergenerational transmission
‘placeholder’ nucleosomes that repel DNA methyl­ation of these marks117,118,138. In X. laevis, sperm maturation
but, on the other hand, are incompatible with transcrip­ entails a global loss of H3K4me2 and H3K4me3 and the
tional activation128 (see below). Other notable differ­ retention of H3K27me3, and this process appears to be
ences in early embryonic reprogramming also exist important for the execution of correct gene expression
between zebrafish and mammals, including the absence patterns during embryogenesis139. In zebrafish, the pres­
of early TET expression and of active demethylation in ence of both H3K27me3 and H3K4me3 at a number of
zebrafish 127,129,130 . The lack of global embryonic developmental genes pre-​marked in sperm by the same
demethyl­ation and remethylation makes zebrafish histone PTMs was observed before the ZGA140. Those
a potentially useful model organism for MEI stud­ observations notwithstanding, the major chromatin
ies in vertebrates131. However, whether such princi­ signatures of embryonic development are established at
ples of embryonic reprogramming are common to or after the ZGA141,142. Moreover, the deposition of his­
teleosts and other anamniotes is currently unclear. tone marks at the ZGA seems to be largely maternally
Similarily, it is currently unknown whether the erasure defined, as recently described in Xenopus tropicalis143.
and re-​establishment of DNA methyl­ation observed in Currently, very little is known about the fate of
mammalian PGCs is conserved more broadly. germline-​inherited chromatin modifications during
embryogenesis. A plausible explanation of how paren­
MEI through histone PTMs tal chromatin patterns might be maintained in early
Despite the increasingly precise classification of regions zebrafish embryos was recently proposed to involve
that might escape DNA methylome reprogramming placeholder nucleosomes128. These nucleosomes are
in mammals (Table 2), direct evidence for the role of specific to the paternal germ line, are enriched in H3K4
DNA methylation in meiotic inheritance in animals is mono-​methylation (H3K4me1) and the histone variant
largely lacking70,116. This is perhaps not surprising con­ H2A.Z(FV) and occupy hypomethylated regions (Fig. 2).
sidering that in many cases DNA methylation merely Interestingly, the genomic locations of placeholder nucleo­
anti-​correlates with the activity of gene-​regulatory somes in sperm and early embryos correlate well, sugges­
regions, suggestive of its role as a secondary, reinforcing tive of a role for placeholder nucleosomes in establishing
regulatory mark132,133. Instead, an increasing number of the pre-​ZGA chromatin structure. The maternal genomic
recent studies have demonstrated intergenerational epi­ contribution, which initially lacks the placeholder con­
genetic inheritance and TEI of histone PTMs (Fig. 1b) in formation, accumulates placeholder nucleosomes during
diverse animal species. We have already discussed the pre-​ZGA development and is progressively remodelled
inheritance of histone PTMs in C. elegans above. to match its paternal counterpart, in line with previous
notions of maternal–paternal DNA methylome repro­
Histone PTM inheritance in Drosophila melanogaster. gramming126,127. Coexistence of H3K4me1 and H2A.Z
D. melanogaster has a long history of exploration of MEI on placeholder nucleosomes deters DNMT activity
phenomena134,135. The role of histone PTMs in TEI in while keeping the chromatin in a transcriptionally inac­
this organism was recently demonstrated by the forma­ tive state. This provides a plausible mechanism for DNA
tion of long-​range chromatin interactions accompanied methylation remodelling in the absence of TET protein
by PRC2-mediated H3K27me3 deposition, which deli­n­ activity in the early zebrafish embryo130. At the onset
eated the inheritance of repressed or derepressed epi­ of ZGA, placeholder-​marked loci become either active
alleles through multiple generations136. Interestingly, (H3K4me3-marked) or poised and silent (H3K4me3–
the establishment of these epigenetic states was shown H3K27me3-marked) (Fig. 2). However, it is important to
to be sensitive to environmental stimuli such as tem­ note that placeholder nucleosomes are established even
perature, thereby demonstrating the interplay between in parthenogenetic embryos (maternal haploids that
the environment and the epigenome. In line with these lack sperm DNA)128, suggestive of a considerable mater­
observations of stable Polycomb TEI, the maternal inheri­ nal contribution in the establishment of placeholder
tance of H3K27me3 and enhancer of zeste (E(z)) protein chromatin. It remains to be determined to what extent
in D. melanogaster was recently shown to have instruc­ placeholder nucleosomes are conserved beyond zebrafish.
tive roles for gene expression at ZGA137 (see below).
Taken together, these examples demonstrate the require­ Histone PTM inheritance in mammals. In mammal­
Zygotic genome activation
(ZGA). Activation of zygotic
ment of meiotic histone PTM inheritance for responses ian sperm, the majority of nucleosomes are replaced
genome transcription for the to environmental stimuli and the supervision of early with protamines to facilitate the compaction of the
first time after fertilization. embryogenesis in insects. paternal genome144. Nevertheless, a small percentage

782 | december 2018 | volume 19 www.nature.com/nrm

 Reviews

DNMTs

Sperm H3K4me1 H3K4me1

mC Placeholder
nucleosome
(H2A.Z(FV)) Reprogramming of maternally contributed
Fertilization
Pre-ZGA DNA methylation patterns
embryo

ZGA
Post-ZGA
embryo H3K4me3 H3K4me3

H3K27me3 Active state

(housekeeping genes)
Poised, silent state
(developmental genes)
Fig. 2 | The role of placeholder nucleosomes in germline-​to-embryo epigenetic reprogramming in zebrafish.
In zebrafish sperm and embryos before zygotic genome activation (ZGA), DNA hypomethylated regions are occupied by
‘placeholder’ nucleosomes bearing the histone variant H2A.Z(FV)128 and mono-​methylated histone H3 lysine 4 (H3K4me1).
The presence of placeholder nucleosomes mediates the maintenance of paternal hypomethylated DNA patterns
through the transcriptionally quiescent cleavage stages by deterring de novo DNA methylation without activating
transcription. The maternal contribution to the embryonic epigenome, which initially lacks the placeholder
conformation, is remodelled through the accumulation of placeholder nucleosomes to match the sperm epigenome in
pre-​ZGA embryos. Upon ZGA , placeholder nucleosomes at DNA hypomethylated regions resolve into either active
(H3K4 trimethylated (H3K4me3)-marked) or poised and inactive (H3K27me3-marked and H3K4me3-marked)
nucleosomes128. DNMTs, DNA (cytosine-5)-methyltransferases; mC, methylcytosine.

of nucleosomes and their associated histone PTMs are findings, overexpression of human LSD1 in the develop­
retained, thereby forming a potential platform for the ing mouse sperm resulted in the reduction of H3K4me2
intergenerational transmission of regulatory states117,118. at promoters of genes regulating developmental and meta­
Both active (H3K4me2 and H3K4me3) and repressive bolic processes and was accompanied by deregulation of
(H3K27me3, H3K9me3 and H4K20me3) histone PTMs gene expression in early F1 embryos150. Notably, these
have been detected in mammalian sperm; however, their changes promoted developmental defects in the offspring
distribution and dynamics vary considerably between and were transmitted across three generations, indica­
species. For example, in humans, H3K9me3-marked tive of TEI. Thus, both active and repressive histone
and H4K20me3-marked nucleosomes are transmitted PTMs can be transferred from the gametes to embryos
through the sperm into the zygote, where they partici­ in mammals (Table 1), with potential implications for the
pate in the build-​up of constitutive heterochromatin145. In regulation of embryonic development (see below).
mice, where only ~1% of the nucleosomes are retained
in sperm118, the build-​up of paternal heterochromatin in MEI through small RNAs
the zygote is initiated through maternally contributed The meiotic inheritance of silencing through small non-
PRC1 (ref.146). Until recently, genome-​wide studies of coding RNAs has been widely studied in nematodes30,38,39,44–48
MEI in mammals were very challenging to orchestrate and flies151,152 but much less so in vertebrates153–156 (Table 1).
owing to the limited amounts of embryonic material that A wealth of data suggest that distinct RNA species pres­
can be obtained for investigation. This limitation is now ent in sperm and oocytes can be carried into the zygote
being overcome by the advent of novel genomics tech­ upon fertilization20,153–158. Nonetheless, the molecular
nologies, which allow for low-​input transcriptome and mechanisms by which these RNA species might act in the
Constitutive epigenome profiling, often at single-​cell resolution147–149. embryo to potentially regulate early development remain
heterochromatin Recent genome-​wide studies revealed the existence of largely unknown. Different external cues, such as viral
Highly condensed,
permanently transcriptionally
robust inheritance of H3K4me3 and H3K27me3 patterns exposure38, nutrition154–156,159, psychological stress20,21 and
silent late-​replicating through oocytes in mice and their role in the regulation high temperature160, can alter the abundance and com­
chromatin. of embryonic development66,71,120–122. In line with these position of RNAs in the germ line, affect embryogenesis

NATure RevIeWs | MOLeCuLAR CeLL BIOLOGy volume 19 | december 2018 | 783

Reviews

and induce phenotypic changes in diverse species (see PTM H3K27 acetylation (H3K27ac) at enhancers during
below). Recent work also suggests that RNA modifica­ ZGA, thereby linking maternally inherited, Polycomb-​
tions might have an important role in the heritability mediated gene silencing with proper enhancer usage in
of metabolic phenotypes. DNMT2, which is a tRNA the early embryo137 (Fig. 3a).
methyltransferase, might have a major role in the trans­ In mammalian embryos, H3K27me3 appears to have
mission of paternally acquired metabolic disorders in an important role in DNA-​methylation-independent
mice161. Specifically, injection of total RNA from sperm maternal imprinting 66. Mapping open chromatin
of obese (high-​fat diet (HFD)-fed) males into zygotes regi­ons using the DNase I-​hypersensitivity assay in
resulted in a metabolic phenotype that was not observed mouse zygotes revealed several hundred parental-​allele-
upon injection of RNA extracted from Dnmt2−/− HFD-​ specific (imprinted) DNase I-​hypersensitive sites (DHSs),
fed males. It is currently unknown whether the loss of which are indicative of gene expression, including novel
tRNA methylation can have a broader, indirect effect on imprinted regions. Of note, only ~20% of paternal-​allele-
other RNA species present in the sperm161. specific DHSs showed DNA methylation at the mater­
piRNAs (Fig. 1c) are found in a broad range of organ­ nal allele, implying the existence of an alternative
isms, such as C. elegans44–46, D. melanogaster151,152, zebra­ mechanism for reducing chromatin accessibility at the
fish162 and mammals, in which they have major roles in maternal allele (Fig. 3b). One-​fifth of these maternal-​
maintaining the stability of germline genomes. However, allele imprints were lost upon overexpression of lysine-​
unlike in C. elegans and D. melanogaster45–47,151, where specific demethylase 6B (KDM6B), which in the zygote
piRNAs also have a role in germline-​inherited epi­ is H3K27me3-specific and forms open chromatin at
genetic silencing (Fig. 1c), in mammals such a function the maternal alleles. H3K27me3-regulated maternal
for piRNAs remains unproved. Examples of intergen­ alleles were hypomethylated in the oocyte, suggesting
erational transmission of small RNAs in mammals that it is the repressive histone PTM H3K27me3 rather
with clear implications to offspring phenotypes involve than DNA methylation that constrains maternal-​allele
miRNAs153,157,163 and tRFs154–156,161. Intergenerationally accessibility at these imprinted loci66. Moreover, highly
inheri­ted miRNAs have a role in embryonic devel­ sensitive chromatin immunoprecipitation followed by
opment 157,163, transmission of stress effects20,21 and sequencing (ChIP–seq) profiling of 14 stages of mouse
paramutation of the KIT gene 153, whereas tRFs have embryogenesis and gametogenesis revealed the existence
mostly been linked to the transmission of metabolic of inheritance of maternally provided H3K27me3, which
phenotypes154–156. Altogether, these findings highlight a persists on thousands of distal regulatory regions until
role for small RNAs in heritable silencing of genomic the blastocyst stage122. This finding is in line with the
elements in different species, regulation of embryonic proposed role of maternally contributed H3K27me3
development and transmission of acquired phenotypes. in enhancer regulation at ZGA in D. melanogaster137.
Although many of these examples await independent However, unlike in D. melanogaster, in mice, H3K27me3
experimental validation, in the future, it will be of great is erased from Hox genes upon fertilization and the
interest to unravel the contribution of RNA chemical repressive state of the Hox cluster is maintained through
modifications164 to their MEI potential161 and to explore alternative mechanisms; H3K27me3 is re-established
the evolutionary conservation of such mechanisms and in post-​implantation embryos122 (Fig. 3b). Other Poly­
their potential relevance to humans. comb targets in mouse embryos displayed similar
H3K27me3 dynamics.
MEI and embryogenesis In mouse oocytes, non-​canonically broad H3K4me3
A number of studies published over the past few years domains occupy almost one-​quarter of the genome
have suggested that intergenerational transmission of and coincide with DNA hypomethylation120,121. The
histone PTMs and small RNAs might have important great majority of genes activated in the major ZGA —
implications for the regulation of early embryogenesis the major burst of zygotic transcription by RNA poly­
of the F1 offspring. These observations have been made merase II — as well as distal regulatory elements involved
in both vertebrate and invertebrate organisms and have in ZGA and in the establishment of totipotency, reside
revealed novel modes of embryonic gene regulation. within these broad H3K4me3 domains120,121. However,
ZGA is associated with the decomposition of these broad
Regulation of embryogenesis by inherited histone H3K4me3 domains into sharper, transcription start site-​
PTMs. In D. melanogaster, H3K27me3 can be transmit­ associated H3K4me3 ‘peaks’ in late two-​cell embryos,
ted from oocytes to the embryo together with maternally which is accompanied by the deposition of H3K27ac120
Paramutation supplied E(z), which is the methyltransferase component (Fig. 3c). Establishment of broad H3K4me3 domains
Heritable change in gene of the PRC2 complex137. Ablation of maternally inherited in oocytes, which is mediated by the histone-​lysine
expression of a (paramutable) E(z) in D. melanogaster as well as its homologue enhancer N-​methyltransferase 2B (KMT2B), and their resolution
allele, which is mediated by
of zeste homologue 2 in mice compromises embryonic upon ZGA, which is mediated by the histone demethyl­
trans-​interaction with the
homologous (paramutagenic) development by causing homeotic transformation137,165. ases KDM5A and KDM5B, are crucial for H3K4me3-
allele. Therefore, by establishing appropriate chromatin associated genome silencing during oogenesis and
states in the maternal germ line, Polycomb proteins reactivation in late two-​cell embryos, respectively120,121.
Homeotic transformation and H3K27me3 ensure intergenerational control of Depletion of KMT2B, KDM5A or KDM5B compromised
The formation of a body
structure or an organ in place
early embryonic development before ZGA 137,165,166. early embryonic development120,121,167,168. Together with
of another in an abnormal Specifically, oocyte-​specific depletion of E(z) results in the recent observations of intergenerational H3K27me3
location. the aberrant accumulation of the active-​genes histone inheritance66,137, these findings provide intriguing clues

784 | december 2018 | volume 19 www.nature.com/nrm

 Reviews

a Drosophila melanogaster b Mus musculus Polycomb-targeted

Hox gene promoters Hox gene H3K27me3-
cluster Enhancers Gene-distal regions cluster imprinted regions

Oocyte ZGA Gametes

WT
M
Nucleus
Totipotent nuclei P

ZGA Zygote
E(z) KD M
Pluripotent nuclei
P

Blastocyst
ZGA ICM M

P
Trophoectoderm
c Mus musculus ZGA gene ZGA gene Distal regulatory
promoters promoters elements E6.5
M M

Epiblast P P
Gametes
M E6.5
Extraembryonic ectoderm M
P
P
Visceral endoderm
Zygote
M E9.5 Placenta M
P
P
Early two-cell
embryo M

P
Late two-cell KDM5A and KDM5B
embryo H3K27me3
(ZGA) M
H3K27ac
P
H3K4me3

Closed chromatin
Eight-cell Open chromatin (DHS)
embryo M

P Gene repression

Gene expression

Fig. 3 | Intergenerational inheritance of histone post-​translational and H3K27me3 at the maternal allele, are inherited by the zygote and
modifications contributes to gene regulation during embryogenesis. propagated to the blastocyst stage66. Their nearest genes (n = 76)
a | In Drosophila melanogaster, histone H3 lysine 27 trimethylation display paternal-​a llele-specific gene expression that is largely
(H3K27me3), which is a repressive post-​translational modification, is maintained in the blastocyst (more so in the trophectoderm than in the
maternally transmitted to the embryo and maintained until zygotic gene ICM). In post-​implantation embryos, H3K27me3-dependent imprinting
activation (ZGA), including at Hox loci137. Depletion of the maternally is erased in the epiblast but is partially maintained in extra-​embryonic
inherited Polycomb protein enhancer of zeste (E(z)) results in loss of ectoderm and visceral endoderm at E6.5 and in the placenta until E9.5.
H3K27me3 and aberrant H3K27 acetylation (H3K27ac) at enhancers, c | Non-​canonically broad H3K4me3 domains in mouse oocytes are
leading to erroneous gene activation during ZGA. b | In mouse gametes, transmitted to the zygote upon fertilization and are transformed into
Hox loci and the majority of promoters targeted by Polycomb proteins canonical H3K4me3 peaks in late two-​c ell embryos, where they
are marked by H3K27me3; however, in the zygote and two-​cell embryos, characterize genes and regulatory elements activated during ZGA.
H3K27me3 is reduced at the maternal allele and is completely lost at the Decomposition of non-​canonical H3K4me3 domains coincides with the
paternal allele. H3K27me3 at these regions emerges in the inner cell mass appearance of H3K27ac at ZGA genes in late two-​cell and eight-​cell
(ICM) of the blastocyst and is re-​established in the epiblast at embryonic stages. Compared with the sperm, the paternal pronucleus has
day 6.5 (E6.5). H3K27me3 at gene-​distal regions is inherited from oocytes considerably less H3K4me3, which is re-​established by lysine-​specific
and propagated on the maternal allele in the zygote, two-​cell embryos and histone demethylase 5A (KDM5A) and KDM5B in late two-​c ell
ICM122. DNA methylation-​independent imprinted loci, characterized by embryos and onwards120,121. KD, knockdown; M, maternal; P, paternal;
the presence of DNase I-​hypersensitive sites (DHSs) at the paternal allele WT, wild type.

NATure RevIeWs | MOLeCuLAR CeLL BIOLOGy volume 19 | december 2018 | 785

Reviews

as to how parentally contributed histone PTMs (Table 1) genes regulated by the long terminal repeats of MERVL
provide crucial signals required for the proper (mouse endogenous retrovirus with leucine tRNA
establishment of transcriptional programmes at ZGA. primer) in mouse embryos and embryonic stem cells
(Fig. 4a). These tRFs were trafficked from the epididymis,
Regulation of embryogenesis by inherited small RNAs. thereby providing remarkable evidence of soma-​to-
Small non-​coding RNAs inherited from gametes also germline transmission of paternal dietary effects156,172.
participate in the regulation of early embryonic devel­ Similarly, the injection into zygotes of tRFs from fathers
opment. In mice, inhibition of a sperm-​derived miRNA, that were kept on a HFD resulted in downregulation of
miR-34c, reduced DNA synthesis and suppressed metabolic pathways in both eight-​cell embryos and blasto­
the first zygotic division157. Similarly, maternal loss cysts154. In another study, the transmission of a latent
of the endonuclease Dicer, which is essential for miRNA metabolic phenotype from obese, prediabetic F0 males
formation, was not compatible with the first cell divi­ to F1 and to F2 was linked to the inheritance of several
sion in mouse embryos163, while paternal loss of Dicer RNA species. Those included tRFs and miRNAs such
resulted in severely impaired preimplantation develop­ as miR-10, which constituted ~25% of all miRNAs in
ment169. However, in zebrafish, this is not the case, and the paternally obese F1 mice155 and the targets of which
embryos derived from maternal Dicer mutants only start were enriched in functions related to transcriptional
displaying developmental defects during gastrulation, regulation (Fig. 4b). Notably, the transmission of meta­
somitogenesis, brain and heart development170. These bolic phenotypes discussed above appears to be highly
studies are further supported by recent observations linked to the methylation status of small RNA species by
that small RNAs found in mature sperm are essential DNMT2, which is an area that merits further investiga­
for embryogenesis. Specifically, zygotes generated via tion161. DNA methylation was also recently implicated
intracytoplasmic sperm injection (ICSI) of immature in the transmission of metabolic phenotypes associated
sperm (derived from caput epididymis) largely failed to with in utero malnutrition in mammals116,176. However,
support embryo­genesis and exhibited multiple defects the precise function of DNA methylation in these pro­
during and after implantation171. However, these defects cesses and its underlying causality remain unclear. In
were rescued by the injection of small RNAs (18–40 summary, these results provide evidence for widespread
nucleotides) derived from extracellular vesicles (epididy­ MEI through diverse molecular mechanisms in insects
mosomes) that deliver their small RNA repertoire to and mammals and demonstrate how environmental cues
mature sperm. This finding suggests that small RNAs might be sensed and transmitted to the offspring.
in mature sperm171,172, which consist mostly of miRNAs
but also of tRFs, likely have major roles during early Conclusions and future perspective
mouse embryogenesis. Numerous aspects of MEI in plants and in short-​lived
organisms, such as C. elegans and D. melanogaster,
Inheritance of metabolic phenotypes have been thoroughly described and excellently rev­
A major area of MEI research is related to the study iewed24–28,134,135. By contrast, the functions, mechanisms
of potential adaptive advantages of intergenerational and dimensions of MEI in most animal phyla still
inheritance of histone PTMs, small RNAs and DNA remain poorly understood. The recent advent of genom­
methylation. Of particular interest is the possibility that ics technologies has enabled the thorough exploration
environmental influences might affect the epigenome of MEI phenomena across a broad range of species and
of parental gametes, which can then be transferred to in different biological contexts. For historical reasons177,
the zygote to alter gene regulation. The importance of and owing to its precise, replication-​coupled enzymatic
meiotic inheritance of the histone PTMs H3K9me3 machinery89,90, DNA methylation has garnered much
and H3K27me3 in response to environmental stimuli attention as a potential mechanism of MEI in animals.
was recently demonstrated in D. melanogaster. Paternal Nevertheless, the links between DNA methylation and
high sugar consumption was shown to alter the chroma­ MEI remain a topic of debate70,72,116,178. Recent insights
tin state of embryonic, larval and metabolic genes in F1 from genomics studies suggest that other molecular
offspring, accompanied by increased triglyceride levels mechanisms, such as histone PTMs66,120,121,136,137 and small
and body weight173. This alteration was due to the tran­ RNAs, including tRFs154–156,161, likely have important roles
scriptional activation of genes residing in ‘black’ (lamin/ in the intergenerational transmission of gene-​regulatory
histone-​H1-associated) and ‘blue’ (Polycomb-​associated) information in the animal kingdom.
heterochromatin174, in line with previous findings of A major criticism related to the study of MEI in
heterochromatin MEI in flies175. Notably, the H3K9me3 mammals is the lack of associated causality88. In other
mark was reduced in the fat body cells of F1 offspring words, it is very challenging to unambiguously demon­
males whose fathers were fed a high-​sugar diet173. strate that the regulatory changes acquired in the paren­
Small RNAs have also been recently described as tal germ line can be efficiently relayed to the offspring to
key determinants of intergenerational transmission of alter gene expression. An even more challenging aspect
metabolic phenotypes in mammals. Recent work shed of MEI studies is the lack of proof that such epigenetic
light on how small (28–34 nucleotide) tRFs derived from changes, acquired in one generation and passed on to the
tRNA 5ʹ ends and inherited through sperm can poten­ next, have any adaptive roles3. The recent developments
tially regulate embryogenesis in mice154–156,171 (Table 1). in CRISPR–Cas9-driven targeted epigenome remodel­
tRFs derived from tRNA–Gly–GCC of fathers fed a ling applied to easily tractable model organisms such
low-​protein diet were shown to mediate repression of as zebrafish could tackle these challenges. For example,

786 | december 2018 | volume 19 www.nature.com/nrm

 Reviews

a Normal diet HFD or LPD b

Lean
× a/a avy/a Obese and
×
prediabetic
F0
Altered
F0 abundance
Oocyte Sperm of tRFs
(HFD, LPD)
and let-7 Latent insulin
miRNAs (LPD) resistance and Normal Normal
impaired glucose diet diet
DNMT2-mediated tolerance* Lean
tRF methylation
a/a × a/a
Two-cell Repression of MERVL-
embryo regulated genes (LPD)
F1
Eight-cell Altered Sperm Oocyte
embryo abundance
of tRFs and
Downregulation of miRNAs
metabolic pathway genes (HFD)

F1 Blastocyst
Latent insulin
resistance and
impaired glucose Normal
Normal diet tolerance* diet
F2

Glucose intolerance and a/a

insulin resistance (HFD)

Pancreatic Downregulation of metabolic

islet pathway genes (HFD)

Fig. 4 | Inheritance of metabolic phenotypes through small RNAs in mammals. a | The abundance of tRNA fragments (tRFs)
and let-7 family microRNAs (miRNAs) is altered in the sperm of mice exposed to a high-​fat diet (HFD) or low-​protein diet (LPD).
Specifically , increased abundance of tRFs (such as tRF–Gly–GCC) and decreased abundance of the let-7 miRNA was observed
in the sperm of male mice fed a LPD156. The sperm of mice fed a HFD showed upregulation of tRFs and downregulation
of miRNAs154. Downregulation of genes controlled by the long terminal repeats of MERVL (mouse endogenous retrovirus
with leucine tRNA primer) was observed in two-​cell embryos fathered by LPD mice, and antisense-mediated inhibition of
tRF–Gly–GCC resulted in derepression of the MERVL-​regulated genes156. Downregulation of metabolic pathway genes was
observed in eight-​cell embryos and blastocyst-​stage embryos of HFD fathers154. Adult F1 offspring of HFD fathers exhibit
impaired glucose tolerance, insulin resistance and downregulation of metabolic genes in pancreatic islets154. F1 offspring of
LPD fathers exhibit upregulation of a cholesterol biosynthesis enzyme in the liver156. DNA (cytosine-5)-methyltransferase 2
(DNMT2) is a tRNA methyltransferase that participates in the transmission of metabolic phenotypes by altering the secondary
structure and stability of tRNAs161. b | F1 progeny (a/a) of obese agouti viable yellow (avy) heterozygous males exhibit impaired
glucose tolerance and insulin resistance following HFD challenge (asterisk)155. The transmission of these phenotypes might be
caused by altered levels of tRFs and miRNAs such as miR-10, which were present in the F1 sperm of mice fed a normal diet but
sired by obese fathers. The impaired glucose tolerance and insulin resistance phenotype was also observed in F2 mice during
HFD challenge (asterisk). No metabolic phenotypes were detected in F3 (ref.155).

the fusion of the Cas9 protein to catalytic domains of diverse species. The ever-​decreasing cost of sequencing
DNMTs179–181 or of histone-​modifying enzymes182 could technologies is also expected to result in more whole-​
enable targeted epigenetic modification of genomes that genome sequences to complement MEI studies, which
could then be monitored through generations. This tar­ will help in differentiating whether the observed inheri­
geted modification will facilitate precise functional stud­ tance phenomena are epigenetic, genetic or both. Until
ies that do not depend on often-​lethal gene knockouts, then, the evidence for MEI that spans multiple gener­
which can hamper the interrogation of heritable effects, ations in mammals remains limited to sporadic examples.
and will also allow the study of the potential depend­ To date, the strongest example of such a phenomenon
ence of MEI phenomena on the DNA sequence. It is remains the inheritance of the DNA methylation state
worth noting that, in many cases, cis-​acting factors are of the agouti locus68.
required for the initiation of epigenetic inheritance85–87, In this Review, we discussed various molecular mech­
but in some cases the deletion of the initiating sequence anisms of parental reach into the embryonic develop­
after a chromatin state has been established does not pre­ ment of the offspring66,120,121,136,137,154–156. Whether such
clude epigenetic inheritance183. Thus CRISPR–Cas9 epi­ phenomena are conserved in humans and to what extent
genome and genome editing systems will have the ability to awaits further clarification. It is plausible that some of
assess such factors in a high-​throughput manner and in these mechanisms would function in humans owing

NATure RevIeWs | MOLeCuLAR CeLL BIOLOGy volume 19 | december 2018 | 787

Reviews

to the high conservation of the underlying enzymatic genome-​wide assays and the increased frequency of
machinery184. On the other hand, it is worth noting that studies carried out on donated human embryonic mat­
considerable differences in sperm nucleosome retention erial, it is reasonable to expect such proof-​of-principle
and early embryonic chromatin establishment have been studies sooner rather than later. Overall, the recent revo­
observed between humans and mice118,145. Studying epi­ lution in genomics technologies, including targeted
genetic inheritance in humans will remain a problem genome and epigenome manipulation, is expected to
owing to long generation times, which restrict the col­ provide long-​sought answers to fundamental questions
lection of biological specimens, and owing to the genetic about the functions and extent of MEI in animals.
diversity of the human population. Nevertheless, with
the reduced amounts of input material required for Published online 13 November 2018

1. Richards, E. J. Inherited epigenetic variation silenced chromatin in Saccharomyces cerevisiae. 46. Shirayama, M. et al. piRNAs initiate an epigenetic
— revisiting soft inheritance. Nat. Rev. Genet. 7, Annu. Rev. Biochem. 72, 481–516 (2003). memory of nonself RNA in the C. elegans germline.
395–401 (2006). 23. Casadesus, J. & Low, D. Epigenetic gene regulation in Cell 150, 65–77 (2012).
2. Daxinger, L. & Whitelaw, E. Understanding the bacterial world. Microbiol. Mol. Biol. Rev. 70, References 44–46 describe how piRNAs
transgenerational epigenetic inheritance via the 830–856 (2006). contribute to transgenerational epigenetic memory
gametes in mammals. Nat. Rev. Genet. 13, 153–162 24. Quadrana, L. & Colot, V. Plant transgenerational in Caenorhabditis elegans.
(2012). epigenetics. Annu. Rev. Genet. 50, 467–491 (2016). 47. Grishok, A., Tabara, H. & Mello, C. C. Genetic
3. Miska, E. A. & Ferguson-​Smith, A. C. Transgenerational 25. Hollick, J. B. Paramutation and related phenomena in requirements for inheritance of RNAi in C. elegans.
inheritance: models and mechanisms of non-​DNA diverse species. Nat. Rev. Genet. 18, 5–23 (2017). Science 287, 2494–2497 (2000).
sequence-​based inheritance. Science 354, 59–63 26. Houri-​Zeevi, L. & Rechavi, O. A. Matter of time: small 48. Vastenhouw, N. L. et al. Gene expression: long-​term
(2016). RNAs regulate the duration of epigenetic inheritance. gene silencing by RNAi. Nature 442, 882 (2006).
4. Greally, J. M. A user’s guide to the ambiguous word Trends Genet. 33, 46–57 (2017). 49. Kalinava, N., Ni, J. Z., Peterman, K., Chen, E. & Gu, S. G.
‘epigenetics’. Nat. Rev. Mol. Cell Biol. 19, 207–208 27. Minkina, O. & Hunter, C. P. Intergenerational Decoupling the downstream effects of germline
(2018). transmission of gene regulatory information in nuclear RNAi reveals that H3K9me3 is dispensable
5. Adams, F. The Genuine Works of Hippocrates Caenorhabditis elegans. Trends Genet. 34, 54–64 for heritable RNAi and the maintenance of
(William Wood, NY, USA, 1891). (2017). endogenous siRNA-​mediated transcriptional silencing
6. Lamarck, J. B. Philosophie Zoologique (JB Baillière, 28. Serobyan, V. & Sommer, R. J. Developmental systems in Caenorhabditis elegans. Epigenetics Chromatin 10,
Paris, 1809). of plasticity and trans-​generational epigenetic 6 (2017).
7. Waddington, C. H. Genetic assimilation of an acquired inheritance in nematodes. Curr. Opin. Genet. Dev. 45, 50. Lev, I. et al. MET-2-dependent H3K9 methylation
character. Int. J. Org. Evol. 7, 118–126 (1953). 51–57 (2017). suppresses transgenerational small RNA inheritance.
8. Waddington, C. H. Genetic assimilation of the bithorax 29. Springer, N. M. & Schmitz, R. J. Exploiting induced Curr. Biol. 27, 1138–1147 (2017).
phenotype. Evolution 10, 1 (1956). and natural epigenetic variation for crop improvement. This work demonstrates how induction of RNAi in
9. Rutherford, S. L. & Lindquist, S. Hsp90 as a capacitor Nat. Rev. Genet. 18, 563–575 (2017). nematodes deficient in the H3K9
for morphological evolution. Nature 396, 336–342 30. Fire, A. et al. Potent and specific genetic interference methyltransferase MET-2 results in permanent
(1998). by double-​stranded RNA in Caenorhabditis elegans. RNA responses that persist even after 30
This work describes the key role of Hsp90 in the Nature 391, 806–811 (1998). generations.
buffering of naturally occurring phenotypic 31. Bernstein, E., Caudy, A. A., Hammond, S. M. & 51. Greer, E. L. et al. Transgenerational epigenetic
variation in fruit flies. Hannon, G. J. Role for a bidentate ribonuclease in inheritance of longevity in Caenorhabditis elegans.
10. Zhao, C. Q., Young, M. R., Diwan, B. A., Coogan, T. P. the initiation step of RNA interference. Nature 409, Nature 479, 365–371 (2011).
& Waalkes, M. P. Association of arsenic-​induced 363–366 (2001). 52. Katz, D. J., Edwards, T. M., Reinke, V. & Kelly, W. G. A.
malignant transformation with DNA hypomethylation 32. Ketting, R. F. et al. Dicer functions in RNA interference C. elegans LSD1 demethylase contributes to germline
and aberrant gene expression. Proc. Natl Acad. and in synthesis of small RNA involved in developmental immortality by reprogramming epigenetic memory.
Sci. USA 94, 10907–10912 (1997). timing in C. elegans. Genes Dev. 15, 2654–2659 Cell 137, 308–320 (2009).
11. James, S. J. et al. Mechanisms of DNA damage, DNA (2001). 53. Gaydos, L. J., Wang, W. & Strome, S. Gene repression.
hypomethylation, and tumor progression in the folate/ 33. Yigit, E. et al. Analysis of the C. elegans Argonaute H3K27me and PRC2 transmit a memory of repression
methyl-​deficient rat model of hepatocarcinogenesis. family reveals that distinct Argonautes act sequentially across generations and during development. Science
J. Nutr. 133, 3740S–3747S (2003). during RNAi. Cell 127, 747–757 (2006). 345, 1515–1518 (2014).
12. Baccarelli, A. & Bollati, V. Epigenetics and 34. Sijen, T. et al. On the role of RNA amplification in 54. Klosin, A., Casas, E., Hidalgo-​Carcedo, C., Vavouri, T.
environmental chemicals. Curr. Opin. Pediatr. 21, dsRNA-​triggered gene silencing. Cell 107, 465–476 & Lehner, B. Transgenerational transmission of
243–251 (2009). (2001). environmental information in C. elegans. Science 356,
13. Blaschke, K. et al. Vitamin C induces Tet-​dependent 35. Pak, J. & Fire, A. Distinct populations of primary and 320–323 (2017).
DNA demethylation and a blastocyst-​like state in ES secondary effectors during RNAi in C. elegans. Science This report demonstrates H3K9me3-associated,
cells. Nature 500, 222–226 (2013). 315, 241–244 (2007). temperature-​induced changes in gene expression,
14. Vickers, M. H. Early life nutrition, epigenetics and 36. Swarts, D. C. et al. The evolutionary journey of which are heritable for at least 14 generations in
programming of later life disease. Nutrients 6, Argonaute proteins. Nat. Struct. Mol. Biol. 21, the nematode Caenorhabditis elegans.
2165–2178 (2014). 743–753 (2014). 55. Lev, I., Gingold, H. & Rechavi, O. H3K9me3 is
15. Klengel, T. & Binder, E. B. Epigenetics of stress-​related 37. Guang, S. et al. An Argonaute transports siRNAs from required for transgenerational inheritance of small
psychiatric disorders and gene x environment the cytoplasm to the nucleus. Science 321, 537–541 RNAs that target a unique subset of newly evolved
interactions. Neuron 86, 1343–1357 (2015). (2008). genes. bioRxiv https://doi.org/10.1101/338582
16. Cropley, J. E., Suter, C. M., Beckman, K. B. & 38. Rechavi, O., Minevich, G. & Hobert, O. Transgenerational (2018).
Martin, D. I. Germ-​line epigenetic modification of the inheritance of an acquired small RNA-​based antiviral 56. Lee, H. J., Hore, T. A. & Reik, W. Reprogramming the
murine A vy allele by nutritional supplementation. response in C. elegans. Cell 147, 1248–1256 (2011). methylome: erasing memory and creating diversity.
Proc. Natl Acad. Sci. USA 103, 17308–17312 39. Burton, N. O., Burkhart, K. B. & Kennedy, S. Cell Stem Cell 14, 710–719 (2014).
(2006). Nuclear RNAi maintains heritable gene silencing in 57. Eckersley-​Maslin, M. A., Alda-​Catalinas, C. & Reik, W.
17. Dunn, G. A. & Bale, T. L. Maternal high-​fat diet Caenorhabditis elegans. Proc. Natl Acad. Sci. USA Dynamics of the epigenetic landscape during the
effects on third-​generation female body size via the 108, 19683–19688 (2011). maternal-​to-zygotic transition. Nat. Rev. Mol. Cell Biol.
paternal lineage. Endocrinology 152, 2228–2236 40. Ishidate, T. et al. ZNFX-1 functions within perinuclear 19, 436–450 (2018).
(2011). nuage to balance epigenetic signals. Mol. Cell 70, 58. Ferguson-​Smith, A. C. Genomic imprinting: the
18. Crews, D. et al. Epigenetic transgenerational 639–649 (2018). emergence of an epigenetic paradigm. Nat. Rev.
inheritance of altered stress responses. Proc. Natl 41. Wan, G. et al. Spatiotemporal regulation of liquid-​like Genet. 12, 565–575 (2011).
Acad. Sci. USA 109, 9143–9148 (2012). condensates in epigenetic inheritance. Nature 557, 59. Reik, W., Collick, A., Norris, M. L., Barton, S. C.
19. Padmanabhan, N. et al. Mutation in folate metabolism 679–683 (2018). & Surani, M. A. Genomic imprinting determines
causes epigenetic instability and transgenerational 42. Bagijn, M. P. et al. Function, targets, and evolution methylation of parental alleles in transgenic mice.
effects on development. Cell 155, 81–93 (2013). of Caenorhabditis elegans piRNAs. Science 337, Nature 328, 248–251 (1987).
20. Gapp, K. et al. Implication of sperm RNAs in 574–578 (2012). 60. Sapienza, C., Peterson, A. C., Rossant, J. & Balling, R.
transgenerational inheritance of the effects of early 43. Lee, H. C. et al. C. elegans piRNAs mediate the Degree of methylation of transgenes is dependent on
trauma in mice. Nat. Neurosci. 17, 667–669 (2014). genome-​wide surveillance of germline transcripts. Cell gamete of origin. Nature 328, 251–254 (1987).
21. Rodgers, A. B., Morgan, C. P., Leu, N. A. & Bale, T. L. 150, 78–87 (2012). 61. Bartolomei, M. S., Zemel, S. & Tilghman, S. M.
Transgenerational epigenetic programming via sperm 44. Ashe, A. et al. piRNAs can trigger a multigenerational Parental imprinting of the mouse H19 gene. Nature
microRNA recapitulates effects of paternal stress. epigenetic memory in the germline of C. elegans. Cell 351, 153–155 (1991).
Proc. Natl Acad. Sci. USA 112, 13699–13704 150, 88–99 (2012). 62. Bourc’his, D., Xu, G. L., Lin, C. S., Bollman, B.
(2015). 45. Luteijn, M. J. et al. Extremely stable Piwi-​induced & Bestor, T. H. Dnmt3L and the establishment of
22. Rusche, L. N., Kirchmaier, A. L. & Rine, J. gene silencing in Caenorhabditis elegans. EMBO J. maternal genomic imprints. Science 294, 2536–2539
The establishment, inheritance, and function of 31, 3422–3430 (2012). (2001).

788 | december 2018 | volume 19 www.nature.com/nrm

 Reviews

63. Coan, P. M., Burton, G. J. & Ferguson-​Smith, A. C. experiments in mammals. Nat. Methods 14, 118. Brykczynska, U. et al. Repressive and active histone
Imprinted genes in the placenta — a review. Placenta 243–249 (2017). methylation mark distinct promoters in human and
26, S10–S20 (2005). 89. Bestor, T. H. The DNA methyltransferases of mouse spermatozoa. Nat. Struct. Mol. Biol. 17,
64. Frost, J. M. & Moore, G. E. The importance of mammals. Hum. Mol. Genet. 9, 2395–2402 (2000). 679–687 (2010).
imprinting in the human placenta. PLOS Genet. 6, 90. Jurkowska, R. Z., Jurkowski, T. P. & Jeltsch, A. 119. Erkek, S. et al. Molecular determinants of nucleosome
e1001015 (2010). Structure and function of mammalian DNA retention at CpG-​rich sequences in mouse spermatozoa.
65. Nicholls, R. D. & Knepper, J. L. Genome organization, methyltransferases. Chembiochem 12, 206–222 Nat. Struct. Mol. Biol. 20, 868–875 (2013).
function, and imprinting in Prader–Willi and Angelman (2011). References 117–119 describe potentially heritable
syndromes. Annu. Rev. Genom. Hum. Genet. 2, 91. Suzuki, M. M. & Bird, A. DNA methylation landscapes: DNA and histone modifications in the mammalian
153–175 (2001). provocative insights from epigenomics. Nat. Rev. sperm.
66. Inoue, A., Jiang, L., Lu, F., Suzuki, T. & Zhang, Y. Genet. 9, 465–476 (2008). 120. Dahl, J. A. et al. Broad histone H3K4me3 domains in
Maternal H3K27me3 controls DNA methylation-​ 92. Schubeler, D. Function and information content of mouse oocytes modulate maternal-​to-zygotic
independent imprinting. Nature 547, 419–424 (2017). DNA methylation. Nature 517, 321–326 (2015). transition. Nature 537, 548–552 (2016).
This work demonstrates the existence of 93. Simpson, V. J., Johnson, T. E. & Hammen, R. F. 121. Zhang, B. et al. Allelic reprogramming of the histone
H3K27me3-dependent imprinting independently Caenorhabditis elegans DNA does not contain modification H3K4me3 in early mammalian
of DNA methylation during mammalian 5-methylcytosine at any time during development or development. Nature 537, 553–557 (2016).
embryogenesis. aging. Nucleic Acids Res. 14, 6711–6719 (1986). References 120 and 121 describe the
67. Lane, N. et al. Resistance of IAPs to methylation 94. Lyko, F., Ramsahoye, B. H. & Jaenisch, R. DNA intergenerational inheritance of the histone
reprogramming may provide a mechanism for methylation in Drosophila melanogaster. Nature 408, modification H3K4me3 in mammals and its roles
epigenetic inheritance in the mouse. Genesis 35, 538–540 (2000). during early mammalian embryogenesis.
88–93 (2003). 95. Mayer, W., Niveleau, A., Walter, J., Fundele, R. 122. Zheng, H. et al. Resetting epigenetic memory by
68. Morgan, H. D., Sutherland, H. G., Martin, D. I. & Haaf, T. Demethylation of the zygotic paternal reprogramming of histone modifications in mammals.
& Whitelaw, E. Epigenetic inheritance at the agouti genome. Nature 403, 501–502 (2000). Mol. Cell 63, 1066–1079 (2016).
locus in the mouse. Nat. Genet. 23, 314–318 (1999). 96. Oswald, J. et al. Active demethylation of the paternal 123. Macleod, D., Clark, V. H. & Bird, A. Absence of
69. Waterland, R. A. & Jirtle, R. L. Transposable elements: genome in the mouse zygote. Curr. Biol. 10, 475–478 genome-​wide changes in DNA methylation during
targets for early nutritional effects on epigenetic gene (2000). development of the zebrafish. Nat. Genet. 23,
regulation. Mol. Cell. Biol. 23, 5293–5300 (2003). 97. Santos, F., Hendrich, B., Reik, W. & Dean, W. 139–140 (1999).
70. Blewitt, M. E., Vickaryous, N. K., Paldi, A., Koseki, H. Dynamic reprogramming of DNA methylation in the 124. Veenstra, G. J. & Wolffe, A. P. Constitutive genomic
& Whitelaw, E. Dynamic reprogramming of DNA early mouse embryo. Dev. Biol. 241, 172–182 methylation during embryonic development of
methylation at an epigenetically sensitive allele in (2002). Xenopus. Biochim. Biophys. Acta 1521, 39–44
mice. PLOS Genet. 2, e49 (2006). 98. Smith, Z. D. et al. A unique regulatory phase of DNA (2001).
71. Xu, Q. & Xie, W. Epigenome in early mammalian methylation in the early mammalian embryo. Nature 125. Bogdanovic, O. et al. Temporal uncoupling of the DNA
development: inheritance, reprogramming and 484, 339–344 (2012). methylome and transcriptional repression during
establishment. Trends Cell Biol. 28, 237–253 99. Wossidlo, M. et al. 5-Hydroxymethylcytosine in the embryogenesis. Genome Res. 21, 1313–1327 (2011).
(2018). mammalian zygote is linked with epigenetic 126. Jiang, L. et al. Sperm, but not oocyte, DNA methylome
72. Hitchins, M. P. Constitutional epimutation as a reprogramming. Nat. Commun. 2, 241 (2011). is inherited by zebrafish early embryos. Cell 153,
mechanism for cancer causality and heritability? 100. Wang, L. et al. Programming and inheritance of 773–784 (2013).
Nat. Rev. Cancer 15, 625–634 (2015). parental DNA methylomes in mammals. Cell 157, 127. Potok, M. E., Nix, D. A., Parnell, T. J. & Cairns, B. R.
73. Suter, C. M., Martin, D. I. & Ward, R. L. Germline 979–991 (2014). Reprogramming the maternal zebrafish genome after
epimutation of MLH1 in individuals with multiple 101. Amouroux, R. et al. De novo DNA methylation drives fertilization to match the paternal methylation pattern.
cancers. Nat. Genet. 36, 497–501 (2004). 5hmC accumulation in mouse zygotes. Nat. Cell Biol. Cell 153, 759–772 (2013).
74. Hitchins, M. P. et al. Inheritance of a cancer-​associated 18, 225–233 (2016). 128. Murphy, P. J., Wu, S. F., James, C. R., Wike, C. L.
MLH1 germ-​line epimutation. N. Engl. J. Med. 356, 102. Morgan, H. D., Santos, F., Green, K., Dean, W. & Cairns, B. R. Placeholder nucleosomes underlie
697–705 (2007). & Reik, W. Epigenetic reprogramming in mammals. germline-​to-embryo DNA methylation reprogramming.
75. Lynch, H. T. et al. Review of the Lynch syndrome: Hum. Mol. Genet. 1, R47–R58 (2005). Cell 172, 993–1006 (2018).
history, molecular genetics, screening, differential 103. Borgel, J. et al. Targets and dynamics of promoter This work describes placeholder nucleosomes,
diagnosis, and medicolegal ramifications. Clin. Genet. DNA methylation during early mouse development. marked by H3K4me1 and H2AZ, which have a role
76, 1–18 (2009). Nat. Genet. 42, 1093–1100 (2010). in intergenerational inheritance of paternal
76. Morak, M. et al. Further evidence for heritability of an 104. Guo, H. et al. The DNA methylation landscape of epigenetic modifications in zebrafish.
epimutation in one of 12 cases with MLH1 promoter human early embryos. Nature 511, 606–610 (2014). 129. Almeida, R. D. et al. 5-Hydroxymethyl-​cytosine
methylation in blood cells clinically displaying HNPCC. 105. Smith, Z. D. et al. DNA methylation dynamics of the enrichment of non-​committed cells is not a universal
Eur. J. Hum. Genet. 16, 804–811 (2008). human preimplantation embryo. Nature 511, feature of vertebrate development. Epigenetics 7,
77. Crepin, M. et al. Evidence of constitutional MLH1 611–615 (2014). 383–389 (2012).
epimutation associated to transgenerational 106. Birnbaum, R. Y. et al. Coding exons function as tissue-​ 130. Bogdanovic, O. et al. Active DNA demethylation at
inheritance of cancer susceptibility. Hum. Mutat. 33, specific enhancers of nearby genes. Genome Res. 22, enhancers during the vertebrate phylotypic period.
180–188 (2012). 1059–1068 (2012). Nat. Genet. 48, 417–426 (2016).
78. Kaati, G., Bygren, L. O. & Edvinsson, S. Cardiovascular 107. Blattler, A. et al. Global loss of DNA methylation 131. Kamstra, J. H., Sales, L. B., Alestrom, P. & Legler, J.
and diabetes mortality determined by nutrition during uncovers intronic enhancers in genes showing Differential DNA methylation at conserved non-​genic
parents’ and grandparents’ slow growth period. Eur. J. expression changes. Genome Biol. 15, 469 (2014). elements and evidence for transgenerational
Hum. Genet. 10, 682–688 (2002). 108. Popp, C. et al. Genome-​wide erasure of DNA inheritance following developmental exposure to
79. Roseboom, T., de Rooij, S. & Painter, R. The Dutch methylation in mouse primordial germ cells is mono(2-ethylhexyl) phthalate and 5-azacytidine in
famine and its long-​term consequences for adult affected by AID deficiency. Nature 463, 1101–1105 zebrafish. Epigenetics 10, 20 (2017).
health. Early Hum. Dev. 82, 485–491 (2006). (2010). 132. Stadler, M. B. et al. DNA-​binding factors shape the
80. Bygren, L. O. et al. Change in paternal grandmothers’ 109. Guibert, S., Forne, T. & Weber, M. Global profiling of mouse methylome at distal regulatory regions. Nature
early food supply influenced cardiovascular mortality DNA methylation erasure in mouse primordial germ 480, 490–495 (2011).
of the female grandchildren. BMC Genet. 15, 12 cells. Genome Res. 22, 633–641 (2012). 133. Thurman, R. E. et al. The accessible chromatin
(2014). 110. Hackett, J. A. et al. Germline DNA demethylation landscape of the human genome. Nature 489, 75–82
81. Yehuda, R. et al. Holocaust exposure induced dynamics and imprint erasure through (2012).
intergenerational effects on FKBP5 methylation. 5-hydroxymethylcytosine. Science 339, 448–452 134. Cavalli, G. & Paro, R. The Drosophila Fab-7
Biol. Psychiatry 80, 372–380 (2016). (2013). chromosomal element conveys epigenetic inheritance
82. Heijmans, B. T. et al. Persistent epigenetic differences 111. Hill, P. W. S. et al. Epigenetic reprogramming enables during mitosis and meiosis. Cell 93, 505–518
associated with prenatal exposure to famine in humans. the transition from primordial germ cell to gonocyte. (1998).
Proc. Natl Acad. Sci. USA 105, 17046–17049 (2008). Nature 555, 392–396 (2018). 135. Cavalli, G. & Paro, R. Epigenetic inheritance of active
83. Painter, R. C. et al. Transgenerational effects of 112. Seisenberger, S. et al. The dynamics of genome-​wide chromatin after removal of the main transactivator.
prenatal exposure to the Dutch famine on neonatal DNA methylation reprogramming in mouse primordial Science 286, 955–958 (1999).
adiposity and health in later life. BJOG 115, germ cells. Mol. Cell 48, 849–862 (2012). 136. Ciabrelli, F. et al. Stable polycomb-​dependent
1243–1249 (2008). 113. Gkountela, S. et al. DNA demethylation dynamics in transgenerational inheritance of chromatin states in
84. Lappalainen, T. & Greally, J. M. Associating cellular the human prenatal germline. Cell 161, 1425–1436 Drosophila. Nat. Genet. 49, 876–886 (2017).
epigenetic models with human phenotypes. Nat. Rev. (2015). This work demonstrates how nuclear organization
Genet. 18, 441–451 (2017). 114. Guo, F. et al. The transcriptome and DNA methylome and Polycomb group proteins contribute to the
85. Coleman, R. T. & Struhl, G. Causal role for inheritance landscapes of human primordial germ cells. Cell 161, formation of stable and transgenerationally
of H3K27me3 in maintaining the OFF state of a 1437–1452 (2015). heritable epialleles in fruit flies.
Drosophila HOX gene. Science 356, eaai8236 (2017). 115. Tang, W. W. et al. A unique gene regulatory network 137. Zenk, F. et al. Germ line-​inherited H3K27me3 restricts
86. Laprell, F., Finkl, K. & Muller, J. Propagation of resets the human germline epigenome for development. enhancer function during maternal-​to-zygotic
Polycomb-​repressed chromatin requires sequence-​ Cell 161, 1453–1467 (2015). transition. Science 357, 212–216 (2017).
specific recruitment to DNA. Science 356, 85–88 116. Radford, E. J. et al. In utero effects. In utero This study reports regulation of enhancer activity at
(2017). undernourishment perturbs the adult sperm the onset of zygotic genome activation through
87. Wang, X. & Moazed, D. DNA sequence-​dependent methylome and intergenerational metabolism. Science maternally-​inherited H3K27me3 in fruit fly embryos.
epigenetic inheritance of gene silencing and histone 345, 1255903 (2014). 138. Wu, S. F., Zhang, H. & Cairns, B. R. Genes for embryo
H3K9 methylation. Science 356, 88–91 (2017). 117. Hammoud, S. S. et al. Distinctive chromatin in human development are packaged in blocks of multivalent
88. Bohacek, J. & Mansuy, I. M. A guide to designing sperm packages genes for embryo development. chromatin in zebrafish sperm. Genome Res. 21,
germline-​dependent epigenetic inheritance Nature 460, 473–478 (2009). 578–589 (2011).

NATure RevIeWs | MOLeCuLAR CeLL BIOLOGy volume 19 | december 2018 | 789

Reviews

139. Teperek, M. et al. Sperm is epigenetically programmed 165. Erhardt, S. et al. Consequences of the depletion of 192. Lachner, M., O’Carroll, D., Rea, S., Mechtler, K.
to regulate gene transcription in embryos. Genome zygotic and embryonic enhancer of zeste 2 during & Jenuwein, T. Methylation of histone H3 lysine 9
Res. 26, 1034–1046 (2016). preimplantation mouse development. Development creates a binding site for HP1 proteins. Nature 410,
140. Lindeman, L. C. et al. Prepatterning of developmental 130, 4235–4248 (2003). 116–120 (2001).
gene expression by modified histones before zygotic 166. Posfai, E. et al. Polycomb function during oogenesis is 193. Bernstein, E. et al. Mouse polycomb proteins bind
genome activation. Dev. Cell 21, 993–1004 (2011). required for mouse embryonic development. Genes differentially to methylated histone H3 and RNA and
141. Akkers, R. C. et al. A hierarchy of H3K4me3 and Dev. 26, 920–932 (2012). are enriched in facultative heterochromatin. Mol. Cell.
H3K27me3 acquisition in spatial gene regulation in 167. Andreu-​Vieyra, C. V. et al. MLL2 is required in oocytes Biol. 26, 2560–2569 (2006).
Xenopus embryos. Dev. Cell 17, 425–434 (2009). for bulk histone 3 lysine 4 trimethylation and 194. Czermin, B. et al. Drosophila enhancer of Zeste/ESC
142. Vastenhouw, N. L. et al. Chromatin signature of transcriptional silencing. PLOS Biol. 8, e1000453 complexes have a histone H3 methyltransferase
embryonic pluripotency is established during genome (2010). activity that marks chromosomal Polycomb sites. Cell
activation. Nature 464, 922–926 (2010). 168. Liu, X. et al. Distinct features of H3K4me3 and 111, 185–196 (2002).
143. Hontelez, S. et al. Embryonic transcription is controlled H3K27me3 chromatin domains in pre-​implantation 195. Kuzmichev, A., Nishioka, K., Erdjument-​Bromage, H.,
by maternally defined chromatin state. Nat. Commun. embryos. Nature 537, 558–562 (2016). Tempst, P. & Reinberg, D. Histone methyltransferase
6, 10148 (2015). 169. Yuan, S. et al. Sperm-​borne mi-​RNAs and endo-​siRNAs activity associated with a human multiprotein complex
144. Balhorn, R. The protamine family of sperm nuclear are important for fertilization and preimplantation containing the Enhancer of Zeste protein. Genes Dev.
proteins. Genome Biol. 8, 227 (2007). embryonic development. Development 143, 635–647 16, 2893–2905 (2002).
145. van de Werken, C. et al. Paternal heterochromatin (2016). 196. Birve, A. et al. Su(z)12, a novel Drosophila Polycomb
formation in human embryos is H3K9/HP1 directed 170. Giraldez, A. J. et al. MicroRNAs regulate brain group gene that is conserved in vertebrates and
and primed by sperm-​derived histone modifications. morphogenesis in zebrafish. Science 308, 833–838 plants. Development 128, 3371–3379 (2001).
Nat. Commun. 5, 5868 (2014). (2005). 197. Pasini, D., Bracken, A. P., Jensen, M. R., Lazzerini
146. Puschendorf, M. et al. PRC1 and Suv39h specify 171. Conine, C. C., Sun, F., Song, L., Rivera-​Perez, J. A. Denchi, E. & Helin, K. Suz12 is essential for mouse
parental asymmetry at constitutive heterochromatin in & Rando, O. J. Small RNAs gained during epididymal development and for EZH2 histone methyltransferase
early mouse embryos. Nat. Genet. 40, 411–420 transit of sperm are essential for embryonic activity. EMBO J. 23, 4061–4071 (2004).
(2008). development in mice. Dev. Cell 46, 470–480 (2018). 198. Margueron, R. et al. Role of the polycomb protein EED
147. Brind’Amour, J. et al. An ultra-​low-input native 172. Sharma, U. et al. Small RNAs are trafficked from the in the propagation of repressive histone marks. Nature
ChIP-seq protocol for genome-​wide profiling of rare epididymis to developing mammalian sperm. Dev. Cell 461, 762–767 (2009).
cell populations. Nat. Commun. 6, 6033 (2015). 46, 481–494 (2018). 199. Simon, J., Chiang, A., Bender, W., Shimell, M. J.
148. Rotem, A. et al. Single-​cell ChIP-​seq reveals cell 173. Ost, A. et al. Paternal diet defines offspring chromatin & O’Connor, M. Elements of the Drosophila bithorax
subpopulations defined by chromatin state. state and intergenerational obesity. Cell 159, complex that mediate repression by Polycomb group
Nat. Biotechnol. 33, 1165–1172 (2015). 1352–1364 (2014). products. Dev. Biol. 158, 131–144 (1993).
149. Clark, S. J., Lee, H. J., Smallwood, S. A., Kelsey, G. 174. Filion, G. J. et al. Systematic protein location mapping 200. Chan, C. S., Rastelli, L. & Pirrotta, V. A. Polycomb
& Reik, W. Single-​cell epigenomics: powerful new reveals five principal chromatin types in Drosophila response element in the Ubx gene that determines an
methods for understanding gene regulation and cell cells. Cell 143, 212–224 (2010). epigenetically inherited state of repression. EMBO J.
identity. Genome Biol. 17, 72 (2016). 175. Seong, K. H., Li, D., Shimizu, H., Nakamura, R. & 13, 2553–2564 (1994).
150. Siklenka, K. et al. Disruption of histone methylation Ishii, S. Inheritance of stress-​induced, ATF-2-dependent 201. Perino, M. et al. MTF2 recruits Polycomb Repressive
in developing sperm impairs offspring health epigenetic change. Cell 145, 1049–1061 (2011). Complex 2 by helical-​shape-selective DNA binding.
transgenerationally. Science 350, aab2006 (2015). 176. Martinez, D. et al. In utero undernutrition in male Nat. Genet. 50, 1002–1010 (2018).
This work demonstrates how overexpression of the mice programs liver lipid metabolism in the second-​ 202. Santos-​Rosa, H. et al. Active genes are
H3K4 methylase LSD1 during mouse generation offspring involving altered Lxra DNA tri-methylated at K4 of histone H3. Nature 419,
spermatogenesis results in transgenerationally methylation. Cell Metab. 19, 941–951 (2014). 407–411 (2002).
heritable developmental phenotypes. 177. Cubas, P., Vincent, C. & Coen, E. An epigenetic 203. Bernstein, B. E. et al. Genomic maps and comparative
151. Brennecke, J. et al. An epigenetic role for maternally mutation responsible for natural variation in floral analysis of histone modifications in human and mouse.
inherited piRNAs in transposon silencing. Science symmetry. Nature 401, 157–161 (1999). Cell 120, 169–181 (2005).
322, 1387–1392 (2008). 178. Shea, J. M. et al. Genetic and epigenetic variation, but 204. Koch, C. M. et al. The landscape of histone
152. de Vanssay, A. et al. Paramutation in Drosophila linked not diet, shape the sperm methylome. Dev. Cell 35, modifications across 1% of the human genome in five
to emergence of a piRNA-​producing locus. Nature 750–758 (2015). human cell lines. Genome Res. 17, 691–707 (2007).
490, 112–115 (2012). 179. Liu, X. S. et al. Editing DNA methylation in the 205. Shilatifard, A. The COMPASS family of histone H3K4
153. Rassoulzadegan, M. et al. RNA-​mediated non-​ mammalian genome. Cell 167, 233–247 (2016). methylases: mechanisms of regulation in development
mendelian inheritance of an epigenetic change in 180. Vojta, A. et al. Repurposing the CRISPR-​Cas9 system and disease pathogenesis. Annu. Rev. Biochem. 81,
the mouse. Nature 441, 469–474 (2006). for targeted DNA methylation. Nucleic Acids Res. 44, 65–95 (2012).
154. Chen, Q. et al. Sperm tsRNAs contribute to 5615–5628 (2016). 206. Kim, V. N., Han, J. & Siomi, M. C. Biogenesis of small
intergenerational inheritance of an acquired metabolic 181. Pflueger, C. et al. A modular dCas9-SunTag DNMT3A RNAs in animals. Nat. Rev. Mol. Cell Biol. 10,
disorder. Science 351, 397–400 (2016). epigenome editing system overcomes pervasive 126–139 (2009).
155. Cropley, J. E. et al. Male-​lineage transmission of off-target activity of direct fusion dCas9-DNMT3A 207. Goll, M. G. et al. Methylation of tRNAAsp by the DNA
an acquired metabolic phenotype induced by constructs. Genome Res. 28, 1193–1206 (2018). methyltransferase homolog Dnmt2. Science 311,
grand-​paternal obesity. Mol. Metab. 5, 699–708 182. Kwon, D. Y., Zhao, Y. T., Lamonica, J. M. & Zhou, Z. 395–398 (2006).
(2016). Locus-​specific histone deacetylation using a synthetic 208. Hock, J. & Meister, G. The Argonaute protein family.
156. Sharma, U. et al. Biogenesis and function of tRNA CRISPR-​Cas9-based HDAC. Nat. Commun. 8, 15315 Genome Biol. 9, 210 (2008).
fragments during sperm maturation and fertilization (2017).
in mammals. Science 351, 391–396 (2016). 183. Ragunathan, K., Jih, G. & Moazed, D. Epigenetics. Acknowledgements
References 154–156 demonstrate the importance Epigenetic inheritance uncoupled from sequence-​ The authors thank J. Cropley, F. Zenk, A. Ashe and R. Lister
of small RNAs in the intergenerational specific recruitment. Science 348, 1258699 for comments on the manuscript. The authors apologize to
transmission of metabolic phenotypes in mice. (2015). colleagues whose work could not be cited owing to space
157. Liu, W. M. et al. Sperm-​borne microRNA-34c is 184. Bannister, A. J. & Kouzarides, T. Regulation of limitations.
required for the first cleavage division in mouse. chromatin by histone modifications. Cell Res. 21,
Proc. Natl Acad. Sci. USA 109, 490–494 (2012). 381–395 (2011).
Author contributions
158. Yang, Q. et al. Highly sensitive sequencing reveals 185. Meaney, M. J. Maternal care, gene expression, and
All authors contributed to researching data for the article,
dynamic modifications and activities of small RNAs the transmission of individual differences in stress
discussion of content and the writing and editing of the
in mouse oocytes and early embryos. Sci. Adv. 2, reactivity across generations. Annu. Rev. Neurosci. 24,
manuscript.
e1501482 (2016). 1161–1192 (2001).
159. Rechavi, O. et al. Starvation-​induced transgenerational 186. Goll, M. G. & Bestor, T. H. Eukaryotic cytosine
inheritance of small RNAs in C. elegans. Cell 158, methyltransferases. Annu. Rev. Biochem. 74, 481–514 Competing interests
277–287 (2014). (2005). The authors declare no competing interests.
160. Schott, D., Yanai, I. & Hunter, C. P. Natural RNA 187. Bird, A. DNA methylation patterns and epigenetic
interference directs a heritable response to the memory. Genes Dev. 16, 6–21 (2002). Publisher’s note
environment. Sci. Rep. 4, 7387 (2014). 188. Peters, A. H. et al. Loss of the Suv39h histone Springer Nature remains neutral with regard to jurisdictional
161. Zhang, Y. et al. Dnmt2 mediates intergenerational methyltransferases impairs mammalian claims in published maps and institutional affiliations.
transmission of paternally acquired metabolic heterochromatin and genome stability. Cell 107,
disorders through sperm small non-​coding RNAs. 323–337 (2001). Reviewer information
Nat. Cell Biol. 20, 535–540 (2018). 189. Rea, S. et al. Regulation of chromatin structure by Nature Reviews Molecular Cell Biology thanks O. Rechavi
This study describes the key role of the tRNA site-​specific histone H3 methyltransferases. Nature and the other anonymous reviewer(s) for their contribution to
methyltransferase DNMT2 in the transmission of 406, 593–599 (2000). the peer review of this work.
metabolic phenotypes in mice. 190. Schultz, D. C., Ayyanathan, K., Negorev, D., Maul, G. G.
162. Houwing, S. et al. A role for Piwi and piRNAs in germ & Rauscher, F. J. 3rd. SETDB1: a novel KAP-1-associated
Supplementary information
cell maintenance and transposon silencing in histone H3, lysine 9-specific methyltransferase that
Supplementary information is available for this paper at
zebrafish. Cell 129, 69–82 (2007). contributes to HP1-mediated silencing of euchromatic
https://doi.org/10.1038/s41580-018-0074-2.
163. Tang, F. et al. Maternal microRNAs are essential for genes by KRAB zinc-​finger proteins. Genes Dev. 16,
mouse zygotic development. Genes Dev. 21, 644–648 919–932 (2002).
(2007). 191. Bannister, A. J. et al. Selective recognition of Related links
164. Jonkhout, N. et al. The RNA modification landscape methylated lysine 9 on histone H3 by the HP1 chromo Geneimprint: http://www.geneimprint.com
in human disease. RNA 23, 1754–1769 (2017). domain. Nature 410, 120–124 (2001).

790 | december 2018 | volume 19 www.nature.com/nrm