Paternally Induced Transgenerational

Environmental Reprogramming of
Metabolic Gene Expression in Mammals
Benjamin R. Carone,1,10 Lucas Fauquier,1,10 Naomi Habib,4,5,10 Jeremy M. Shea,1,10 Caroline E. Hart,1 Ruowang Li,2
Christoph Bock,6,7 Chengjian Li,1 Hongcang Gu,6 Phillip D. Zamore,1,3 Alexander Meissner,6,7 Zhiping Weng,2
Hans A. Hofmann,8 Nir Friedman,4,9 and Oliver J. Rando1,*
1Department of Biochemistry and Molecular Pharmacology
2Program in Bioinformatics and Integrative Biology
3Howard Hughes Medical Institute

University of Massachusetts Medical School, Worcester, MA 01605, USA

4School of Computer Science and Engineering, The Hebrew University, Jerusalem 91904, Israel
5Department of Molecular Genetics and Biotechnology, Faculty of Medicine, The Hebrew University, Jerusalem 91120, Israel
6Broad Institute, Cambridge, MA 02142, USA
7Department of Stem Cell and Regenerative Biology, Harvard University, Cambridge, MA 02138, USA
8Section for Integrative Biology, Institute for Cellular & Molecular Biology, Institute for Neuroscience, University of Texas at Austin, Austin,

TX 78712, USA
9Institute of Life Sciences, The Hebrew University, Jerusalem 91904, Israel
10These authors contributed equally to this work

*Correspondence: [email protected]
DOI 10.1016/j.cell.2010.12.008

SUMMARY 2005; Rando and Verstrepen, 2007). One implication of epige-

netic inheritance systems is that they provide a potential mech-
Epigenetic information can be inherited through the anism by which parents could transfer information to their
mammalian germline and represents a plausible offspring about the environment they experienced. In other
transgenerational carrier of environmental informa- words, mechanisms exist that could allow organisms to ‘‘inform’’
tion. To test whether transgenerational inheritance their progeny about prevailing environmental conditions. Under
of environmental information occurs in mammals, we certain historical circumstances—for example, repeated expo-
sure over evolutionary time to a moderately toxic environment
carried out an expression profiling screen for genes
that persists for tens of generations—such non-Mendelian infor-
in mice that responded to paternal diet. Offspring mation transfer could be adaptive (reviewed in Jablonka and
of males fed a low-protein diet exhibited elevated Lamb, 1995; Rando and Verstrepen, 2007). Whether or not
hepatic expression of many genes involved in lipid organisms can inherit characters induced by ancestral environ-
and cholesterol biosynthesis and decreased levels ments has far-reaching implications, and this type of inheritance
of cholesterol esters, relative to the offspring of males has come to be called ‘‘Lamarckian’’ inheritance after the early
fed a control diet. Epigenomic profiling of offspring evolutionary theorist J.B. Lamarck, although it is worth noting
livers revealed numerous modest (20%) changes that both Darwin and Lamarck believed in the inheritance of
in cytosine methylation depending on paternal diet, acquired characters.
including reproducible changes in methylation over Despite these theoretical considerations, at present there is
a likely enhancer for the key lipid regulator Ppara. scant evidence for transgenerational effects of the environment,
particularly in mammals. The majority of examples of transge-
These results, in conjunction with recent human
nerational environmental effects described have been maternal
epidemiological data, indicate that parental diet can effects (see Harris and Seckl, 2010; Whitelaw and Whitelaw,
affect cholesterol and lipid metabolism in offspring 2008; Youngson and Whitelaw, 2008 for review), including
and define a model system to study environmental in utero passage of photoperiod information in various rodents
reprogramming of the heritable epigenome. (Horton, 2005), cultural inheritance of stress reactivity and
maternal grooming behavior in rats (Meaney et al., 2007; Weaver
INTRODUCTION et al., 2004), and metabolic and psychiatric sequelae of fetal
malnutrition in humans and rodents (Hales and Barker, 2001;
The past few decades have seen an important expansion of our Harris and Seckl, 2010; Symonds et al., 2009). However,
understanding of inheritance, as a wide variety of epigenetically maternal effects are difficult to separate from direct effects of
inherited traits have been described (Jablonka and Lamb, 1995, in utero environmental exposure on offspring.

1084 Cell 143, 1084–1096, December 23, 2010 ª2010 Elsevier Inc.
 A small number of studies have identified heritable epigenetic offspring were reared with their mothers until 3 weeks old, at
effects of environmental perturbations on offspring. Treatment of which point their livers were harvested for RNA isolation. DNA
pregnant rat mothers with the endocrine disruptor vinclozolin microarrays were used to profile global gene expression differ-
results in decreased fertility and behavorial changes in several ences in the livers of the offspring from the two types of crosses
generations of offspring (Anway et al., 2005; Crews et al., (Table S1).
2007). In another study, withholding methyl donors from preg- Testing for differences between 26 matched pairs of mice from
nant female mice resulted in decreased cytosine methylation the two F1 groups, we found a significant overabundance of
across the agouti viable yellow Avy reporter locus (Waterland differentially expressed genes, relative to the null hypothesis
and Jirtle, 2003), and the altered cytosine methylation profile per- that the parental treatment does not affect offspring (1595 genes
sisted well beyond the first generation (Cropley et al., 2006). at a false discovery rate—FDR—of 0.001, Figures S1B and S1C).
Whereas demonstration of multigenerational changes (e.g., We also identified a more robust (t test with null hypothesis of
an F2 effect) is important when using maternal treatment proto- mean change 0.2, FDR of 0.01) group of 445 genes whose
cols to rule out simple plastic responses of offspring to the in expression strongly depended on the diet consumed by their
utero environment, paternal effects avoid this issue as fathers fathers (Figure 1B). In our analysis we focus on this more robust
often contribute little more than sperm to offspring. A handful group of genes; however, all the phenomena described below
of paternal effects have been documented in the literature— are true for the larger group as well. These gene expression
pre-mating fasting of male mice has been reported to affect changes were observed in 13 (7 low-protein, 6 control) litters in
serum glucose levels in offspring (Anderson et al., 2006), and experiments spanning several years, carried out in three different
chronic exposure of male rats to high-fat diet affects pancreatic animal facilities (Figures S2A and S2B). In principle, random
islet biology in offspring (Ng et al., 2010). Furthermore, epidemi- factors should be distributed equally between our two groups
ological data from human populations link experience of famine given the numbers of offspring examined, but we directly
in paternal grandfathers to obesity and cardiovascular disease address a number of potential artifacts nonetheless, including
two generations later (Kaati et al., 2002; Pembrey et al., 2006). changes in cell populations, circadian cycle, litter size, order
These results motivate a deeper exploration of the mechanisms of sacrifice, and cage location (Figure S2, see Experimental
of pre-mating paternal diet on offspring phenotype. Procedures).
It is therefore of great interest to determine what environ- We confirmed our results by q-RT-PCR (Figures 1C, Fig-
mental conditions have transgenerational effects in mammals, ure S1A). Squalene epoxidase (Sqle), which catalyzes the first
and to characterize the mechanisms that mediate these effects. oxygenation step in sterol biosynthesis, exhibited an 3-fold
Here, we describe a genomic screen for transgenerational increase in the low-protein cohort in our microarray data, and
effects of paternal diet on gene expression in offspring in mice. q-RT-PCR showed a similar average expression difference
Expression of hundreds of genes changes in the offspring of across over 25 animals, gathered in crosses carried out several
males fed a low-protein diet, with coherent upregulation of lipid years apart (Figure 1C). The differences we observe occur in
and cholesterol biosynthetic pathways. Epigenomic profiling in both male and female progeny (Figure 1C, Figure S2C), though
offspring livers identified changes in cytosine methylation at these dietary history-dependent differences are superimposed
a putative enhancer for the key lipid transcription factor Ppara, on a baseline of differential expression between the sexes.
and these changes correlated with the downregulation of this
gene in offspring. Interestingly, we did not find effects of paternal Upregulation of Proliferation and Lipid Biosynthesis
diet on methylation of this locus in sperm, and overall sperm Genes in Low-Protein Offspring
cytosine methylation patterns were largely conserved under To help define the physiological differences between our
various dietary regimes. These results establish an inbred, cohorts, we calculated enrichments of various Gene Ontology
genetically tractable model system for the study of transgenera- (GO) processes in the differentially expressed genes. Genes
tional effects of diet and may have implications for the epidemi- upregulated in our treatment group’s offspring were enriched
ology of several major human diseases. for a number of categories of genes involved in fat and choles-
terol biosynthesis, including lipid biosynthesis (p < 9 3 10 26),
RESULTS steroid biosynthesis (p < 3 3 10 19), cholesterol biosynthesis
(p < 2 3 10 12), and oxidation-reduction (p < 4 3 10 10). Another
Experimental Paradigm major group of upregulated genes are annotated to be involved
Male mice were fed control or low-protein diets (11% rather than in S phase, such as DNA replication (p < 2 3 10 9) and related
20% protein, with the remaining mass made up with sucrose) annotations. Downregulated genes were enriched for GO anno-
from weaning until sexual maturity. Note that although the rele- tations such as sequence-specific DNA binding (p < 6 3 10 6)
vant dietary change in this experiment could be protein content, and ligand-dependent nuclear receptor activity (p < 6 3 10 5),
sucrose content, fat/protein ratio, etc., for simplicity we refer to although the number of genes matching these annotations was
the diet as low protein throughout the text. Mice on either diet small (14 and 5, respectively).
were then mated to females reared on control diet (Figure 1A The increase in S phase genes likely indicates a hyperprolifer-
and Figure S1A available online). Fathers were removed after 1 ative state, whereas the metabolic expression differences sug-
or 2 days of mating, limiting their influence on their progeny gest that lipid metabolism is altered in these animals. To explore
to the mating itself. All mothers were maintained on control the mechanisms responsible for these altered gene expression
diet throughout the course of the experiment. After birth, the programs, we asked whether the observed gene expression

Cell 143, 1084–1096, December 23, 2010 ª2010 Elsevier Inc. 1085
Figure 1. A Screen for Genes Regulated by Paternal Diet
(A) Experimental design. Male mice were fed control or low (11%) protein diet from weaning until sexual maturity, then were mated to females that were raised
on control diet. Males were removed after 1 or 2 days of mating. Livers were harvested from offspring at 3 weeks, and RNA was prepared, labeled, and hybridized
to oligonucleotide microarrays.
(B) Overview of microarray data, comparing offspring of sibling males fed different diets—red boxes indicate higher RNA levels in low-protein than control
offspring, green indicates higher expression in controls. Boxes at the top indicate comparisons between two male (purple) or two female (yellow) offspring.
Each column shows results from a comparison of a pair of offspring. Only genes passing the stringent threshold for significant change (Figure S1B) are shown.
Data are clustered by experiment (columns) and by genes (rows).
(C) Validation of microarray data. Quantitative RT-PCR was used to determine levels of Squalene epoxidase (Sqle) relative to the control gene Vitronectin (Vtn),
which showed no change in the microarray dataset. Animals are grouped by paternal diet and by sex, and data are expressed as DCT between Sqle and Vtn,
normalized relative to the average of control females.
Additional validation is shown in Figure S1A. p values were calculated using t test. See also Table S1, Figure S1, and Figure S2.

1086 Cell 143, 1084–1096, December 23, 2010 ª2010 Elsevier Inc.
 Figure 2. Multiple Pathways Are Affected by
Paternal Diet
Comparison of upregulated gene expression
profile with a compendium of public datasets of
hepatic gene expression. A clustering of our upre-
gulated genes according to their notation in the 28
significant (p < 0.00025) overlapping signatures
from an assembled compendium of 120 publicly
available murine liver signatures under various con-
ditions and genetic perturbations (GEO; Horton
et al., 2003; Yang et al., 2009). For each significant
profile, the majority of overlapping genes are shown
as yellow, whereas genes with opposite regulation
(i.e., down rather than up in the dataset in question)
are blue. The genes divide into two distinct clusters,
one enriched in DNA replication and the other in
various categories of fat and cholesterol biosyn-
thesis. See also Table S2 and Figure S3.

and apoptosis-related genes and down-

regulate genes involved in carboxylic
acid metabolism (analysis not shown).

Transgenerational Effects on Lipid

Metabolism
We further focused on cholesterol bio-
synthesis genes. Coherent upregulation
differences might reflect altered regulation of a small number of of genes involved in cholesterol metabolism is observed in the
pathways. We checked for significant overlaps of the gene offspring of low-protein fathers (Figure 3A). Figure 3B shows
expression profile observed in our low-protein offspring with a more detailed comparison between our upregulated dataset
a compendium of 120 publicly available murine liver gene and published data (Horton et al., 2003) for genes activated
expression datasets (Experimental Procedures). Our low-protein by a major transcriptional regulator of cholesterol metabolism,
offspring gene expression profile significantly (p < .05 after Bon- SREBP. Many of the genes upregulated in low-protein offspring
ferroni correction) overlapped gene expression changes from 28 have previously been shown to be upregulated by overexpres-
published profiles (Figure 2, Table S2), including gene expres- sion of SREBP-1a or SREBP-2 or downregulated by loss of the
sion profiles associated with perturbation of transcription fac- SREBP-activating gene Scap.
tors that regulate cholesterol and lipid metabolism (SREBP To explore the correspondence between hepatic gene ex-
[Horton et al., 2003], KLF15 [Gray et al., 2007], PPARa [Rakh- pression and physiology, we measured lipid levels in three pairs
shandehroo et al., 2007], and ZFP90 [Yang et al., 2009]). Our of control and treatment livers to determine whether increased
gene expression dataset also significantly matched hepatic levels of lipid biosynthesis genes were associated with changes
gene expression in a variety of mice with mutations affecting in lipid levels (Figure 3C, Experimental Procedures). Livers in the
growth hormone (GH) and insulin-like growth factor 1 (IGF-1) cohort with low-protein diet fathers were depleted of cholesterol
levels (Boylston et al., 2004; Madsen et al., 2004; Tsuchiya and cholesterol esters (whose levels were reduced more than
et al., 2004). Hierarchical clustering according to the enriched 2-fold). Additional differences were found in specific lipid classes,
public profiles revealed two types of prominent gene functions such as substantial increases in relative levels of saturated cardi-
in our data: DNA replication (p < 6 3 10 14) and lipid or choles- olipins, saturated free fatty acids, and saturated and monounsat-
terol biosynthesis (p < 2 3 10 27) (Figure 2). The partial overlap urated triacylglycerides in low-protein offspring (Table S3).
observed with each of many different transcription factor and Together, these results demonstrate that paternal diet affects
growth factor profiles suggests that the altered gene expression metabolites of key biomedical importance in offspring.
profile observed in low-protein offspring is likely related to
reprogramming of multiple distinct pathways. MicroRNAs in Offspring
To assess whether the reprogrammed state in offspring repro- Small (19–35 bp) RNAs such as microRNAs (miRNAs) have
duces the paternal response to low-protein diet, we measured recently been implicated in epigenetic inheritance in mice (Wag-
global gene expression changes in the livers of pairs of animals ner et al., 2008). To determine whether altered small RNA popu-
weaned to control or low-protein diet as in Figure 1A. Genes that lations might drive our reprogramming effect, we characterized
change in offspring are not the same as the genes induced in the by high-throughput sequencing the small (19–35 bp) RNA popu-
parental generation by these protocols (Figure S3). Instead, lation from control and low-protein offspring livers (Ghildiyal
males fed the low-protein diet upregulate immune response et al., 2008) and mapped reads to known miRNAs (Table S4).

Cell 143, 1084–1096, December 23, 2010 ª2010 Elsevier Inc. 1087
 Figure 3. Altered Cholesterol Metabolism in
the Low-Protein Cohort
(A) Cholesterol biosynthesis. Genes annotated as
cholesterol biosynthesis genes are shown, with
colors indicating average difference in expression
in low-protein versus control comparisons.
(B) Many genes upregulated in the low-protein
cohort are SREBP targets. Upregulated cluster
from Figure 1B is shown, along with data from Hor-
ton et al. (2003). Genes scored as up in both repli-
cates from Horton et al. (2003) are shown as
yellow, genes scored as down are blue. Columns
show data from transgenic mice overexpressing
SREBP-1a or SREBP-2 or from Scap knockout
mice.
(C) Cholesterol levels are decreased in livers of
low-protein offspring. Data from lipidomic profiling
of liver tissue from three control and three low-
protein animals are shown as mean ± standard
deviation. Red line indicates no change. p values
were calculated using a paired t test on log-trans-
formed lipid abundance data. Cholesterol esters,
CE; phosphatidylethanolamine, PE; free choles-
terol, FC; triacylglycerol, TAG; phopshatidylcho-
line, PC; cardiolipin, CL; phosphatidylserine, PS ;
free fatty acid, FA; lysophosphatidylcholine,
LYPC; and diacylglycerol, DAG.
See also Table S3.

miR-98 and downregulated miR-210.

Many of these upregulated miRNAs are
associated with proliferation in liver, with
miR-21 and miR-199 both associated
with hepatocellular carcinoma (Jiang
et al., 2008), whereas let-7 is well-known
as a tumor suppressor (Jerome et al.,
2007). The increase in growth-associated
miRNAs is consistent with the hyperproli-
ferative gene expression profile observed
in the offspring of low-protein diet fathers.
We found no statistically significant
overlap (p > 0.05) between the predicted
targets of the miRNAs here and the
gene expression changes we observe,
though the subtle (50%) changes in
miRNA abundance we observe might be
expected to have little effect on
mRNA—even when specific miRNAs are
artificially introduced in cells, downregu-
lation of target mRNAs is less than
2-fold for the majority of predicted targets
(Hendrickson et al., 2008). Our results
therefore suggest that miRNAs are likely
A number of miRNAs changed expression in the offspring from to be additional targets of the reprogramming pathway yet are
low-protein diet fathers (Figure 4). Changes were often subtle likely not the direct upstream regulators of the entire response
in magnitude (50%), but were reproduced in four control versus (but see Wagner et al., 2008).
low-protein comparisons (paired t test), and given the number of
sequencing reads obtained for these RNAs this magnitude of Cytosine Methylation in Offspring
difference is well outside of counting error (Table S4). Offspring How are offspring reprogrammed by paternal diet? Cytosine
of low-protein cohort upregulated miR-21, let-7, miR-199, and methylation is a widespread DNA modification that is

1088 Cell 143, 1084–1096, December 23, 2010 ª2010 Elsevier Inc.
 it is associated with the enhancer chromatin mark H3K4me1
(Heintzman et al., 2007) in murine liver (F. Yue and B. Ren,
personal communication). Ppara is downregulated in the
majority (but not all) of offspring livers (Table S1, Figure 6B),
and the overall gene expression profile in our offspring livers
significantly matches the gene expression changes observed
in Ppara knockout mice (Figure 2), suggesting that epigenetic
regulation of this single locus could drive a substantial fraction
of the observed gene expression changes in offspring. Indeed,
variance of Ppara mRNA levels alone can be used to explain
13.7% of the variance in the entire gene expression dataset
(although this of course does not determine causality).
We therefore assayed the methylation status of this locus by
bisulfite sequencing in an additional 17 offspring livers (8 control
and 9 low-protein), finding average differences of up to 8%
Figure 4. Proliferation-Related MicroRNAs Respond to Paternal Diet methylation between low-protein and control livers at several
Small (<35 nt) RNAs from the livers of eight offspring (four control, four low- CpGs in this locus (Figure 6C). Importantly, these pooled data
protein) were isolated and subjected to high-throughput sequencing. underestimate the potential role of this locus in reprogramming
MicroRNAs that exhibited consistent changes in all four pairs of animals are as they include animals exhibiting a range of changes in Ppara
shown, with average change shown as a bar and individual comparisons gene expression—individual animal pairs with large differences
shown as points. See also Table S4.
in Ppara mRNA levels exhibit methylation differences of up to
30% at various cytosines across this locus. Figure 6D shows
environmentally responsive and carries at least some heritable individual bisulfite clones for three pairs of animals with varying
information between generations (Bartolomei et al., 1993; Crop- extents of Ppara downregulation (not all animals used for meth-
ley et al., 2006; Holliday, 1987; Rakyan et al., 2003; Waterland ylation analysis were analyzed by microarray). Taken together,
and Jirtle, 2003). As imprinted loci are often involved in growth these results identify a differentially methylated locus that is
control (Moore and Haig, 1991), we first asked whether a subset a strong candidate to be one of the upstream controllers of the
of candidate imprinted loci exhibited altered cytosine methyla- hepatic gene expression response.
tion in low-protein offspring (Figure S4A). As these loci did not
exhibit significant changes in methylation, we therefore turned Cytosine Methylation, RNA, and Chromatin in Sperm
to genome-scale mapping studies to search for differentially The link between paternal diet and offspring methylation patterns
methylated loci between control and low-protein offspring. lead us to consider the hypothesis that paternal diet affects cyto-
We performed reduced representation bisulfite sequencing sine methylation patterns in sperm. We therefore isolated highly
(RRBS) (Meissner et al., 2008) to characterize cytosine methyla- pure (>99%) sperm from the caudal epididymis of males
tion at single-nucleotide resolution across 1% of the mouse consuming control or low-protein diet. We assayed the Ppara
genome (Table S5). RRBS was performed for livers from a pair enhancer for methylation by bisulfite sequencing but found no
of control and low-protein offspring, and fraction of methylated significant changes between males consuming control or low-
CpGs was calculated for a variety of features such as promoters, protein diet (Figure S4B). These results indicate either that cytosine
enhancers, and other nongenic CpG islands. In general, we found methylation in sperm is not the relevant paternally transmitted die-
that cytosine methylation was well-correlated between control tary information at this locus (but changes at some point during
and low-protein offspring (Figures 5A and 5B). However, we did development; Blewitt et al., 2006), or that we captured animals
observe widespread modest (10%–20%) changes in CpG whose offspring would not manifest significant changes in expres-
methylation between the two samples (red and green dots in sion of the associated genes—as seen in Figure 1B or Figure 6B,
Figures 5A and 5B), consistent with many observations indicating Ppara downregulation is variably penetrant in low-protein offspring.
that environmental changes tend to have small quantitative To globally investigate effects of paternal diet on sperm cyto-
effects on cytosine methylation in the next generation (Blewitt sine methylation, we isolated sperm from four males—two
et al., 2006; Heijmans et al., 2008; Ng et al., 2010; Weaver et al., consuming control diet, one consuming low-protein diet, and
2004). Importantly, changes in promoter methylation did not glob- one subjected to a caloric restriction regimen. We then surveyed
ally correlate with changes in gene expression in offspring, indi- cytosine methylation patterns across the entire genome via
cating that the gene expression program in offspring is unlikely MeDIP-Seq (immunoprecipitation using antibodies against
to be epigenetically specified at each individual gene (Figure 5C). 5me-C followed by deep sequencing; Jacinto et al., 2008; Weber
Of course, widespread gene expression differences can be et al., 2005) (Figure 7A, Figure S5A, and Figure S6). Notably,
caused by changes to a small number of upstream regulators, global cytosine methylation profiles were highly correlated
and a number of differentially methylated regions are associated between any pair of samples, indicating that the sperm ‘‘epige-
with cholesterol- or lipid-related genes (Table S5). nome’’ is largely unresponsive to these differences in diet
Most interestingly, we found a substantial (30%) increase in (Figures 7B–7D, Figures S5B–S5E). Indeed, littermates on
methylation at an intergenic CpG island 50 kb upstream of different diets (Figures 7B and 7C) were better-correlated for
Ppara (Figure 6A). This locus is likely an enhancer for Ppara, as promoter methylation than were the pairs of control animals

Cell 143, 1084–1096, December 23, 2010 ª2010 Elsevier Inc. 1089
 from different litters (Figure 7D). Although these results do not rule
out cytosine methylation in sperm as the relevant carrier of epige-
netic information about paternal diet, the high correlation
between samples, coupled with the absence of cytosine methyl-
ation changes at the Ppara enhancer in sperm, leads us to
consider alternative epigenetic information carriers including
RNA (Rassoulzadegan et al., 2006; Wagner et al., 2008) and chro-
matin (Arpanahi et al., 2009; Brykczynska et al., 2010; Hammoud
et al., 2009; Ooi and Henikoff, 2007).
We used Affymetrix microarrays to analyze RNA levels for
three pairs of males and for two matched epididymis samples
(Figure S6, Figure S7A, Table S6). Curiously, low-protein and
caloric restriction samples consistently exhibited more
‘‘sperm-like’’ RNA populations (as opposed to epididymis
RNA) than did control samples (Figures S7B and S7C). Whether
this reflects systematic contamination issues or biological differ-
ences in sperm maturity or quality is presently unknown,
although we note that we confirmed consistently higher levels
of the sperm-specific Dnahc3 by q-RT-PCR in an additional
7/8 low-protein sperm samples (Figure S7E). We note that
control sperm samples were routinely >99.5% sperm as assayed
by microscopy (Figure S6), but nonetheless we cannot
completely rule out systematic contamination issues. With this
possibility in mind, we identified genes that were differentially
packaged in control versus low-protein sperm by correcting for
potential epididymal contamination (Figures S7B–S7F). Interest-
ingly, we observed downregulation of a number of transcription
factors and chromatin regulators such as Smarcd3 and Ppard,
although q-RT-PCR validation was not statistically significant
due to high inter-animal variability (Figure S7F).
Although the downregulation of Smarcd3 was not significantly
confirmed by q-RT-PCR, this could reflect the variable pene-
trance of paternal diet on offspring described above. Given
that heterozygous mutants in chromatin remodelers can affect
offspring phenotype even when the mutant allele segregates
away (Chong et al., 2007), we used an initial genome-wide
mapping (not shown) of overall histone retention (pan-H3 ChIP)
abundance and the key epigenetic histone modification
H3K27me3 in sperm to identify targets for single locus analysis.
We observed a consistent decrease in H3K27me3 in low-protein
sperm at the promoter of Maoa (Monoamine oxidase) in 5/5 pairs
of sperm samples and a decrease in H3K27me3 at Eftud1 in 4/5
paired samples (Figures S7G and S7H). These results demon-
strate proof of principle that the sperm epigenome is regulated
by dietary conditions, although the biological implications of
these observations are not yet clear.

DISCUSSION

Taken together, our results demonstrate that paternal diet

affects lipid- and proliferation-related gene expression in the

and green dots indicate promoters with significant (p < 0.05) methylation
Figure 5. Transgenerational Effects of Paternal Diet on Hepatic changes of over 10%.
Cytosine Methylation (B) As in (A), for nongenic CpG islands.
(A) Genomic DNA from control and low-protein offspring livers was subjected (C) Promoter cytosine methylation changes are uncorrelated with gene
to reduced representation bisulfite sequencing (RRBS). For all annotated expression changes. For each promoter, the average change in cytosine
promoters, average fraction of CpGs that were methylated is shown for the methylation is compared to the change in mRNA abundance from Figure 1B.
control sample (x axis) compared to the low-protein sample (y axis). Red See also Table S5 and Figure S4.

1090 Cell 143, 1084–1096, December 23, 2010 ª2010 Elsevier Inc.
Figure 6. Effects of Paternal Diet on Methylation of a Putative Ppara Enhancer
(A) Differential methylation of a putative Ppara enhancer. Top panel shows a schematic of chromosome 15: 85,360,000–85,640,000. Zoomed in region represents
chr15: 85,514,715-85,514,920. RRBS data for one control and one low-protein offspring pair are shown below, with assayed CpGs represented as boxes colored
to indicate % of clones methylated. Numbers to the left indicate % methylation, with number of sequence reads covering the CpG in parentheses.
(B) Ppara is downregulated in most low-protein offspring livers. Box plot shows mean, quartiles, and highest and lowest values from Table S1.
(C) Putative enhancer methylation correlates with Ppara downregulation. DNA from eight control and nine low-protein pairs of offspring livers was bisulfite treated,
and at least 13 clones were analyzed for each animal. Percent methylation at each of the 12 CpGs in this region plotted on the y axis; data are shown as mean ±
standard error of the mean (SEM).
(D) Individual bisulfite clones are shown for three control and three low-protein offspring. White circles indicate unmethylated CpGs, black circles indicate meth-
ylated CpGs. Microarray data for change in Ppara RNA levels between the paired animals are shown to the left, in log2. Values under each bisulfite grouping
indicate overall % methylation, with number of clones analyzed in parentheses.

Cell 143, 1084–1096, December 23, 2010 ª2010 Elsevier Inc. 1091
Figure 7. Modest Effects of Diet on the Sperm Epigenome
(A) MeDIP sequencing data are shown for two liver samples (top two tracks) and four sperm samples (bottom four) at a maternally methylated region (Gnas, left)
and a paternally methylated region (Rasgrf1, right).
(B) Comparison of control and low-protein methylation. For each promoter, methylation levels were averaged for 8 kb surrounding the TSS, and values are scat-
terplotted for control sperm (x axis) versus low-protein sperm (y axis). x and y axes are plotted on logarithmic scales.
(C) As in (B), but for control versus caloric restriction.
(D) As in (B), but for the pair of control samples.
Similar results for (B)–(D) are found when focusing on the 1 kb surrounding the TSS (not shown). See Figure S7 for analyses of consistent RNA and chromatin
differences between low-protein and control sperm.

offspring of inbred mice, and that epigenetic information carriers Whether the effects we observe on cholesterol metabolism
in sperm respond to environmental conditions. These results prove advantageous in low-protein conditions remains to be
have potential implications for human health and raise numerous tested, but it will be important to investigate ecologically relevant
mechanistic questions, discussed below. diets in order to speculate more firmly about adaptive signifi-
cance of any observed transgenerational effects. For example,
at present we cannot say with certainty what aspect of the
Paternal Diet Affects Metabolism in Offspring low-protein regimen is sensed by males—it is possible that
Our results clearly identify a set of physiological pathways whose offspring metabolism is affected by overall protein consumption,
expression is sensitive to paternal diet. Specifically, we find that or high sucrose, or fat/protein ratio, or even levels of micronu-
hepatic expression of genes involved in proliferation and choles- trients, as our males consumed diets ad libitum and thus might
terol biosynthesis can be regulated by paternal diet, and these have overconsumed the low-protein diet.
changes are reflected in levels of several lipid metabolites.
Combined with data showing that offspring glucose levels are The Reprogrammed State: Liver
affected by paternal fasting in mice (Anderson et al., 2006), these What is the mechanistic basis for the reprogrammed gene
results demonstrate that paternal diet has wide-ranging effects expression state? Genome-scale analyses of cytosine methyla-
on the metabolism of offspring in rodents. Interestingly, a very tion in offspring livers identified several lipid-related genes that
recent study from Ng et al. (Ng et al., 2010) reported that chronic were differentially methylated depending on paternal diet. Most
exposure of male rats to high-fat diet was associated with notably, a putative enhancer for a major lipid regulator, Ppara,
pancreatic beta cell dysfunction in female offspring. It will natu- exhibited generally higher methylation in low-protein offspring
rally be of great interest in the near future to compare the trans- than in control offspring. Methylation at this locus was variable
generational effects of high-fat and low-protein diets, although between animals, consistent with the partial penetrance of Ppara
one clear difference is that in our system a transgenerational downregulation in our dataset. The overall gene expression
effect is observed in both sex offspring. profile observed in low-protein offspring significantly overlaps

1092 Cell 143, 1084–1096, December 23, 2010 ª2010 Elsevier Inc.
gene expression changes observed in Ppara knockout mice unaffected by these diets. Nonetheless, changes in relatively few
(Rakhshandehroo et al., 2007), leading to the hypothesis that loci can have profound effects in the developing animal, and our
epigenetic Ppara downregulation via enhancer methylation is data do not rule out the possibility of inheritance through sperm
an upstream event that affects an entire downstream regulon cytosine methylation, especially given that MeDIP is unlikely to
in reprogrammed animals. Note that although the hepatic down- identify 10%–20% of differences in methylation at a small
regulation of Ppara suggests a liver-autonomous epigenetic number of cytosines. Importantly, the putative enhancer of Ppara
change, we cannot rule out that hepatic gene expression (Figure 6) was not differentially methylated in sperm. It will there-
changes result from global physiological changes resulting fore be of great interest in the future to determine when during
from downregulation of Ppara in some other tissue. development the differential methylation observed in liver is es-
Interestingly, Ppara expression in liver is also regulated by tablished and to identify the upstream events leading to differen-
maternal diet—offspring of female mice consuming a high-fat tial methylation (Blewitt et al., 2006).
diet exhibit altered hepatic Ppara expression, with increased Interestingly, we did identify effects of diet on RNA content
expression at birth but decreased expression at weaning (Yama- and chromatin packaging of sperm. For example, sperm from
guchi et al., 2010). Together with our data, these results suggest control animals were consistently depleted of the highly
that Ppara is a key nexus that integrates ancestral dietary infor- sperm-specific Dnahc3 gene (Figure S7) relative to sperm from
mation to control offspring metabolism. low-protein animals. We cannot presently determine whether
this represents reproducible differences in contamination, differ-
Mechanistic Basis for Transgenerational Paternal ences in sperm maturity, or something else. Finally, based on our
Effects observation that low-protein sperm tended to be depleted of
Paternal diet could potentially affect offspring phenotype via genes encoding a number of chromatin regulators, we have
a number of different mechanisms. Although we focus here on begun to search for dietary effects on sperm chromatin struc-
epigenetic inheritance systems, it is important to note that ture. Interestingly we found that the Maoa promoter was consis-
parental information can also be passed to offspring via social tently depleted of the key Polycomb-related chromatin mark
or cultural inheritance systems (Avital and Jablonka, 2000; H3K27me3 (Figure S7G), demonstrating as a proof of concept
Champagne and Meaney, 2001; Jablonka and Lamb, 1995; that chromatin packaging of the sperm genome is responsive
Meaney et al., 2007; Weaver et al., 2004). Although such mater- to the environment and motivating genome-wide investigation
nally provided social inheritance is unlikely in our paternal effect into dietary effects on sperm chromatin. Given the common
system—males were typically only in females’ cages for one behavorial changes observed in many transgenerational inheri-
day—it is known that in some animals females can judge mate tance paradigms, the possibility that H3K27me3 at Maoa affects
quality and allocate resources accordingly (Pryke and Griffith, offspring behavior (potentially via altered offspring responses to
2009), and that seminal fluid can influence female postcopula- maternal stress; Harris and Seckl, 2010) will be of great future
tory behavior in Drosophila (Fricke et al., 2008; Wolfner, 2002). interest.
These and other plausible transgenerational information carriers
cannot be excluded at present—ongoing artificial insemination Relevance to Human Disease
and in vitro fertilization experiments will determine whether These results are likely to be relevant for human disease because
sperm carry the relevant metabolic information in our system. not only is maternal starvation in humans correlated with obesity
Here we focused on the hypothesis that paternal dietary infor- and diabetes in children (Lumey et al., 2007), but also, remark-
mation does indeed reside in sperm epigenetic information ably, limited food in paternal grandfathers has been associated
carriers. First, a subset of cytosine methylation patterns in sperm with changed risk of diabetes and cardiovascular disease in
are known to be heritable (Chong et al., 2007; Cropley et al., grandchildren (Kaati et al., 2002; Pembrey et al., 2006). Interest-
2006; Rakyan et al., 2003; Waterland and Jirtle, 2003). Second, ingly, in these studies ancestral access to food and disease risk
several reports suggest that RNA molecules packaged in sperm were not associated with disease risk in the next generation but
can affect offspring phenotype (Rassoulzadegan et al., 2006; were only associated with F2 disease risk. However, it is impor-
Wagner et al., 2008). Third, chromatin structure has been tant to note that the transgenerational effects of food availability
proposed to carry epigenetic information, as sperm are largely for paternal grandfathers depend on the exact period during
devoid of histone proteins but retain them at a subset of develop- childhood of exposure to rich or poor diets (Pembrey et al.,
mentally important loci (Arpanahi et al., 2009; Brykczynska et al., 2006), whereas our experimental protocol involved continuous
2010; Chong et al., 2007; Hammoud et al., 2009). Finally, it is low-protein diet from weaning until mating. Thus, future studies
conceivable that additional or novel epigenetic regulators (such are required to more precisely define when and how ancestral
as prions) are packaged into sperm, or that sperm quality is exposure to a low-protein diet affects epigenetic programming
affected by diet, or that genetic changes are directed by the envi- of offspring metabolism.
ronment (although it is important to emphasize that inbred Together, these results suggest rethinking basic practices in
mouse strains were used in this study). epidemiological studies of complex diseases such as diabetes,
Here, we report whole-genome characterization of cytosine heart disease, or alcoholism. We believe that future environ-
methylation patterns and RNA content in sperm obtained from mental exposure histories will need to include parental exposure
mice maintained on control, low-protein, and caloric restriction histories as well as those of the patients to disentangle induced
diets. Globally, cytosine methylation patterns are similar in all epigenetic effects from the currently sought genetic and environ-
three conditions, indicating that the sperm epigenome is largely mental factors underlying complex diseases. Our observations

Cell 143, 1084–1096, December 23, 2010 ª2010 Elsevier Inc. 1093
provide an inbred mammalian model for transgenerational re- fied, gel-purified, and sequenced using a Solexa Genome Analyzer (Illumina)
programming of metabolic phenotype that will enable dissection (Ghildiyal et al., 2008).
of the exposure history necessary for reprogramming and
RRBS
genetic analysis of the machinery involved in reprogramming,
Reduced representation bisulfite sequencing was carried out as previously
and they suggest a number of specific pathways likely to be described (Meissner et al., 2008). Data are available at http://thrifty-
the direct targets of epigenetic reprogramming. epigenome.computational-epigenetics.org.

EXPERIMENTAL PROCEDURES Sperm Isolation

Caudal epididymis was dissected from sacrificed animals, punctured, and
Mouse Husbandry incubated for 30 min in M2 media (Sigma) at 37 C. Supernatant was removed,
All animal care and use procedures were in accordance with guidelines of the pelleted (3000 g for 5 min), washed 23 with PBS and 13 in water, and incu-
Institutional Animal Care and Use Committee. C57/Bl6 mice were obtained bated in Somatic Cell Lysis buffer. Sperm preparations were used only if
from Jackson Labs and from Charles River Laboratories (for different iterations they were >99.5% pure as assessed by microscopy, and q-RT-PCR was
of this experiment). All experiments were performed with mice that had been also used to reject any sperm samples based on the ratio between epidid-
raised for at least two generations on control diet to attempt to minimize any ymis-specific genes Actb or Myh11 and sperm-specific genes Smcp or
transgenerational effects of transitioning to control diet from chow provided Odf1 (Figure S6).
by animal provider. For all comparisons shown, male mice were weaned
from mothers at 21 days of age, and sibling males were put into cages with MeDIP
low-protein or control diet (moistened with water to allow mice to break the Methyl-DNA immunoprecipitation was carried out essentially as described
hard pellets). Females were weaned to control diet. Males were raised on (Weber et al., 2005, 2007). Four micrograms of purified genomic DNA was frag-
diet until 9–12 weeks of age, at which point they were placed with females mented to a mean size of 300 bp using a Covaris machine, denatured, and
for 1 or 2 days. Control and low-protein mating cages were always inter- immunoprecipitated with 5mC antibody (Eurogentec). ChIP material was Sol-
spersed with one another. Note that we always used virgin females to avoid exa sequenced, with 21 million uniquely mappable reads per library.
confounding effects of the female’s litter number, although this results in
many lost litters as first litters were often consumed by their mothers. After 1
ACCESSION NUMBERS
to 2 days, males were removed, and pregnant females were left alone with
control diet and a shepherd shack until their litters were 3 weeks of age. At
All microarray data and deep sequencing data used in this study have been
3 weeks of age offspring were sacrificed by isoflurane and cervical dislocation,
deposited to GEO (http://www.ncbi.nlm.nih.gov/projects/geo/), accession #
and median lobe of liver was rapidly dissected out and flash-frozen in liquid N2.
GSE25899.
Diets
Diets were obtained from Bio-serv, and compositions are listed in Table S7. SUPPLEMENTAL INFORMATION
For most experiments only low-protein diet was sterilized per standard
protocol at Bio-serv. For later experiments, both diets were sterilized. Supplemental Information includes Extended Experimental Procedures, seven
figures, and seven tables and can be found with this article online at doi:10.
RNA Extraction 1016/j.cell.2010.12.008.
Liver samples were ground with a liquid N2-cooled mortar and pestle. Total
RNA for microarray analysis was extracted from liver powder using Trizol. ACKNOWLEDGMENTS

Microarray Hybridization We thank D. Haig for initial discussions motivating this experiment and
Thirty micrograms of total RNA was labeled for 2 hr at 42 C with Superscript II K. Ahmad, P. Kaufman, C. Meiklejohn, Y. Nahmias, and members of the Rando
reverse transcriptase using 4 mg of random hexamer and 4 mg of oligo dT. Cy3- lab for critical reading of the manuscript. O.J.R. is supported in part by a Career
and Cy5-labeled samples were hybridized to home-printed ‘‘MEEBO’’ micro- Award in Biomedical Sciences from the Burroughs Wellcome Fund, by grant
arrays. MEEBO information is available at http://www.ncbi.nlm.nih.gov/geo/ GM088618 from NIGMS, and by the Mathers Foundation and is a member
query/acc.cgi?acc=GPL6352. Microarrays were hybridized at 65 C for 16 hr, of the UMass Diabetes Endocrinology Research Center (DK32520). N.F. is
washed as previously described (Diehn et al., 2002), and scanned using an supported by grants from the NIH and the US-Israel Binational Science Foun-
Axon Genepix 4000B microarray scanner. dation (BSF). H.A.H. is supported by NSF, an Alfred P. Sloan Foundation
Fellowship and a Dwight W. and Blanche Faye Reeder Centennial Fellowship
Comparison to Public Murine Liver Microarray Data in Systematic and Evolutionary Biology. P.D.Z. is supported by grants
We built a compendium of public microarray data consisting of 120 gene GM62862 and GM65236 from the NIGMS. N.H. is supported by the Meidan
expression profiles in the murine liver under various conditions and genetic fellowship. The funders had no role in study design, data collection, and
perturbations. Signatures of differentially expressed genes were determined analysis, decision to publish, or preparation of the manuscript.
using a combination of two one-tailed t tests, with FDR correction of 0.1. O.J.R. and H.A.H. designed the original expression experiment, and a pilot
Profiles significantly enriched with up- or downregulated genes in low-protein was carried out by C.E.H., O.J.R., and H.A.H. O.J.R., B.C, L.F., and J.S. carried
offspring were defined by a hypergeometric p value % 0.05 after correction for out animal husbandry and gene expression experiments, and L.F. and C.L.
multiple hypotheses (p < 0.00025). carried out miRNA profiling. C.B. and A.M. carried out RRBS experiments,
L.F. carried out sperm RNA profiling, and J.S. carried out MeDIP experiments.
Lipid Measurements J.S. and B.C. carried out bisulfite sequencing, and B.C. carried out chromatin
50–100 mg of ground liver tissue from six animals (three paired sets) was sent ChIPs. N.H., O.J.R., and N.F. analyzed the gene expression and microRNA
to Lipomics for ‘‘Truemass’’ mass spectrometry characterization of 450 lipid data, R.L., Z.W., and O.J.R. analyzed the MeDIP data. O.J.R., P.D.Z., and
levels (Table S4). Note that samples 73-1 and 76-1 come from PBS-perfused N.F. wrote the paper.
livers, whereas the other four samples were dissected without perfusion.
Received: October 29, 2010
Small RNA Cloning and Sequencing Revised: December 6, 2010
Total RNA was isolated from ground liver tissue using mirVana (Ambion). 18–35 Accepted: December 8, 2010
nt small RNA was purified from 100 mg of total RNA, ligated to adaptors, ampli- Published: December 23, 2010

1094 Cell 143, 1084–1096, December 23, 2010 ª2010 Elsevier Inc.
REFERENCES associated with prenatal exposure to famine in humans. Proc. Natl. Acad.
Sci. USA 105, 17046–17049.
Anderson, L.M., Riffle, L., Wilson, R., Travlos, G.S., Lubomirski, M.S., and Heintzman, N.D., Stuart, R.K., Hon, G., Fu, Y., Ching, C.W., Hawkins, R.D.,
Alvord, W.G. (2006). Preconceptional fasting of fathers alters serum glucose Barrera, L.O., Van Calcar, S., Qu, C., Ching, K.A., et al. (2007). Distinct and
in offspring of mice. Nutrition 22, 327–331. predictive chromatin signatures of transcriptional promoters and enhancers
Anway, M.D., Cupp, A.S., Uzumcu, M., and Skinner, M.K. (2005). Epigenetic in the human genome. Nat. Genet. 39, 311–318.
transgenerational actions of endocrine disruptors and male fertility. Science Hendrickson, D.G., Hogan, D.J., Herschlag, D., Ferrell, J.E., and Brown, P.O.
308, 1466–1469. (2008). Systematic identification of mRNAs recruited to argonaute 2 by specific
Arpanahi, A., Brinkworth, M., Iles, D., Krawetz, S.A., Paradowska, A., Platts, microRNAs and corresponding changes in transcript abundance. PLoS ONE
A.E., Saida, M., Steger, K., Tedder, P., and Miller, D. (2009). Endonuclease- 3, e2126.
sensitive regions of human spermatozoal chromatin are highly enriched in Holliday, R. (1987). The inheritance of epigenetic defects. Science 238,
promoter and CTCF binding sequences. Genome Res. 19, 1338–1349. 163–170.
Avital, E., and Jablonka, E. (2000). Animal Traditions: Behavioural Inheritance Horton, J.D., Shah, N.A., Warrington, J.A., Anderson, N.N., Park, S.W., Brown,
in Evolution (Cambridge, UK: Cambridge University Press). M.S., and Goldstein, J.L. (2003). Combined analysis of oligonucleotide micro-
Bartolomei, M.S., Webber, A.L., Brunkow, M.E., and Tilghman, S.M. (1993). array data from transgenic and knockout mice identifies direct SREBP target
Epigenetic mechanisms underlying the imprinting of the mouse H19 gene. genes. Proc. Natl. Acad. Sci. USA 100, 12027–12032.
Genes Dev. 7, 1663–1673. Horton, T.H. (2005). Fetal origins of developmental plasticity: animal models of
Blewitt, M.E., Vickaryous, N.K., Paldi, A., Koseki, H., and Whitelaw, E. (2006). induced life history variation. Am. J. Hum. Biol. 17, 34–43.
Dynamic reprogramming of DNA methylation at an epigenetically sensitive
Jablonka, E., and Lamb, M.J. (1995). Epigenetic Inheritance and Evolution: The
allele in mice. PLoS Genet. 2, e49.
Lamarckian Dimension (Oxford, UK: Oxford University Press).
Boylston, W.H., Gerstner, A., DeFord, J.H., Madsen, M., Flurkey, K., Harrison,
Jablonka, E., and Lamb, M.J. (2005). Evolution in Four Dimensions: Genetic,
D.E., and Papaconstantinou, J. (2004). Altered cholesterologenic and lipo-
Epigenetic, Behavioral, and Symbolic Variation in the History of Life (Cam-
genic transcriptional profile in livers of aging Snell dwarf (Pit1dw/dwJ) mice.
bridge, MA: MIT Press).
Aging Cell 3, 283–296.
Jacinto, F.V., Ballestar, E., and Esteller, M. (2008). Methyl-DNA immunoprecip-
Brykczynska, U., Hisano, M., Erkek, S., Ramos, L., Oakeley, E.J., Roloff, T.C.,
itation (MeDIP): hunting down the DNA methylome. Biotechniques 44, 35, 37,
Beisel, C., Schubeler, D., Stadler, M.B., and Peters, A.H. (2010). Repressive
39 passim.
and active histone methylation mark distinct promoters in human and mouse
spermatozoa. Nat. Struct. Mol. Biol. 17, 679–687. Jerome, T., Laurie, P., Louis, B., and Pierre, C. (2007). Enjoy the silence: the
story of let-7 microRNA and cancer. Curr. Genomics 8, 229–233.
Champagne, F., and Meaney, M.J. (2001). Like mother, like daughter:
evidence for non-genomic transmission of parental behavior and stress Jiang, J., Gusev, Y., Aderca, I., Mettler, T.A., Nagorney, D.M., Brackett, D.J.,
responsivity. Prog. Brain Res. 133, 287–302. Roberts, L.R., and Schmittgen, T.D. (2008). Association of microRNA expres-
sion in hepatocellular carcinomas with hepatitis infection, cirrhosis, and
Chong, S., Vickaryous, N., Ashe, A., Zamudio, N., Youngson, N., Hemley, S.,
patient survival. Clin. Cancer Res. 14, 419–427.
Stopka, T., Skoultchi, A., Matthews, J., Scott, H.S., et al. (2007). Modifiers of
epigenetic reprogramming show paternal effects in the mouse. Nat. Genet. Kaati, G., Bygren, L.O., and Edvinsson, S. (2002). Cardiovascular and diabetes
39, 614–622. mortality determined by nutrition during parents’ and grandparents’ slow
growth period. Eur. J. Hum. Genet. 10, 682–688.
Crews, D., Gore, A.C., Hsu, T.S., Dangleben, N.L., Spinetta, M., Schallert, T.,
Anway, M.D., and Skinner, M.K. (2007). Transgenerational epigenetic imprints Lumey, L., Stein, A.D., Kahn, H.S., van der Pal-de Bruin, K.M., Blauw, G.,
on mate preference. Proc. Natl. Acad. Sci. USA 104, 5942–5946. Zybert, P.A., and Susser, E.S. (2007). Cohort profile: the Dutch hunger winter
families study. Int. J. Epidemiol. 36, 1196–1204.
Cropley, J.E., Suter, C.M., Beckman, K.B., and Martin, D.I. (2006). Germ-line
epigenetic modification of the murine A vy allele by nutritional supplementa- Madsen, M.A., Hsieh, C.C., Boylston, W.H., Flurkey, K., Harrison, D., and
tion. Proc. Natl. Acad. Sci. USA 103, 17308–17312. Papaconstantinou, J. (2004). Altered oxidative stress response of the long-
lived Snell dwarf mouse. Biochem. Biophys. Res. Commun. 318, 998–1005.
Diehn, M., Alizadeh, A.A., Rando, O.J., Liu, C.L., Stankunas, K., Botstein, D.,
Crabtree, G.R., and Brown, P.O. (2002). Genomic expression programs and Meaney, M.J., Szyf, M., and Seckl, J.R. (2007). Epigenetic mechanisms of peri-
the integration of the CD28 costimulatory signal in T cell activation. Proc. natal programming of hypothalamic-pituitary-adrenal function and health.
Natl. Acad. Sci. USA 99, 11796–11801. Trends Mol. Med. 13, 269–277.
Fricke, C., Bretman, A., and Chapman, T. (2008). Adult male nutrition and Meissner, A., Mikkelsen, T.S., Gu, H., Wernig, M., Hanna, J., Sivachenko, A.,
reproductive success in Drosophila melanogaster. Evolution 62, 3170–3177. Zhang, X., Bernstein, B.E., Nusbaum, C., Jaffe, D.B., et al. (2008). Genome-
scale DNA methylation maps of pluripotent and differentiated cells. Nature
Ghildiyal, M., Seitz, H., Horwich, M.D., Li, C., Du, T., Lee, S., Xu, J., Kittler, E.L.,
454, 766–770.
Zapp, M.L., Weng, Z., et al. (2008). Endogenous siRNAs derived from transpo-
sons and mRNAs in Drosophila somatic cells. Science 320, 1077–1081. Moore, T., and Haig, D. (1991). Genomic imprinting in mammalian develop-
Gray, S., Wang, B., Orihuela, Y., Hong, E.G., Fisch, S., Haldar, S., Cline, G.W., ment: a parental tug-of-war. Trends Genet. 7, 45–49.
Kim, J.K., Peroni, O.D., Kahn, B.B., et al. (2007). Regulation of gluconeogen- Ng, S.F., Lin, R.C., Laybutt, D.R., Barres, R., Owens, J.A., and Morris, M.J.
esis by Kruppel-like factor 15. Cell Metab. 5, 305–312. (2010). Chronic high-fat diet in fathers programs beta-cell dysfunction in
Hales, C.N., and Barker, D.J. (2001). The thrifty phenotype hypothesis. Br. female rat offspring. Nature 467, 963–966.
Med. Bull. 60, 5–20. Ooi, S.L., and Henikoff, S. (2007). Germline histone dynamics and epigenetics.
Hammoud, S.S., Nix, D.A., Zhang, H., Purwar, J., Carrell, D.T., and Cairns, B.R. Curr. Opin. Cell Biol. 19, 257–265.
(2009). Distinctive chromatin in human sperm packages genes for embryo Pembrey, M.E., Bygren, L.O., Kaati, G., Edvinsson, S., Northstone, K., Sjos-
development. Nature 460, 473–478. trom, M., and Golding, J. (2006). Sex-specific, male-line transgenerational
Harris, A., and Seckl, J. (2010). Glucocorticoids, prenatal stress and the responses in humans. Eur. J. Hum. Genet. 14, 159–166.
programming of disease. Horm. Behav. Published online June 18, 2010. 10. Pryke, S.R., and Griffith, S.C. (2009). Genetic incompatibility drives sex alloca-
1016/j.yhbeh.2010.06.007. tion and maternal investment in a polymorphic finch. Science 323, 1605–1607.
Heijmans, B.T., Tobi, E.W., Stein, A.D., Putter, H., Blauw, G.J., Susser, E.S., Rakhshandehroo, M., Sanderson, L.M., Matilainen, M., Stienstra, R., Carlberg,
Slagboom, P.E., and Lumey, L.H. (2008). Persistent epigenetic differences C., de Groot, P.J., Muller, M., and Kersten, S. (2007). Comprehensive analysis

Cell 143, 1084–1096, December 23, 2010 ª2010 Elsevier Inc. 1095
of PPARalpha-dependent regulation of hepatic lipid metabolism by expression Weber, M., Davies, J.J., Wittig, D., Oakeley, E.J., Haase, M., Lam, W.L., and
profiling. PPAR Res. 2007, 26839. Schubeler, D. (2005). Chromosome-wide and promoter-specific analyses
Rakyan, V.K., Chong, S., Champ, M.E., Cuthbert, P.C., Morgan, H.D., Luu, identify sites of differential DNA methylation in normal and transformed human
K.V., and Whitelaw, E. (2003). Transgenerational inheritance of epigenetic cells. Nat. Genet. 37, 853–862.
states at the murine Axin(Fu) allele occurs after maternal and paternal trans-
Weber, M., Hellmann, I., Stadler, M.B., Ramos, L., Paabo, S., Rebhan, M., and
mission. Proc. Natl. Acad. Sci. USA 100, 2538–2543.
Schubeler, D. (2007). Distribution, silencing potential and evolutionary impact
Rando, O.J., and Verstrepen, K.J. (2007). Timescales of genetic and epige- of promoter DNA methylation in the human genome. Nat. Genet. 39, 457–466.
netic inheritance. Cell 128, 655–668.
Whitelaw, N.C., and Whitelaw, E. (2008). Transgenerational epigenetic inheri-
Rassoulzadegan, M., Grandjean, V., Gounon, P., Vincent, S., Gillot, I., and
tance in health and disease. Curr. Opin. Genet. Dev. 18, 273–279.
Cuzin, F. (2006). RNA-mediated non-mendelian inheritance of an epigenetic
change in the mouse. Nature 441, 469–474. Wolfner, M.F. (2002). The gifts that keep on giving: physiological functions and
Symonds, M.E., Sebert, S.P., Hyatt, M.A., and Budge, H. (2009). Nutritional evolutionary dynamics of male seminal proteins in Drosophila. Heredity 88,
programming of the metabolic syndrome. Nat. Rev. Endocrinol. 5, 604–610. 85–93.
Tsuchiya, T., Dhahbi, J.M., Cui, X., Mote, P.L., Bartke, A., and Spindler, S.R.
Yamaguchi, R., Nakagawa, Y., Liu, Y.J., Fujisawa, Y., Sai, S., Nagata, E., Sano,
(2004). Additive regulation of hepatic gene expression by dwarfism and caloric
S., Satake, E., Matsushita, R., Nakanishi, T., et al. (2010). Effects of maternal
restriction. Physiol. Genomics 17, 307–315.
high-fat diet on serum lipid concentration and expression of peroxisomal pro-
Wagner, K.D., Wagner, N., Ghanbarian, H., Grandjean, V., Gounon, P., Cuzin, liferator-activated receptors in the early life of rat offspring. Horm. Metab. Res.
F., and Rassoulzadegan, M. (2008). RNA induction and inheritance of epige- 42, 821–825.
netic cardiac hypertrophy in the mouse. Dev. Cell 14, 962–969.
Yang, X., Deignan, J.L., Qi, H., Zhu, J., Qian, S., Zhong, J., Torosyan, G., Majid,
Waterland, R.A., and Jirtle, R.L. (2003). Transposable elements: targets for
S., Falkard, B., Kleinhanz, R.R., et al. (2009). Validation of candidate causal
early nutritional effects on epigenetic gene regulation. Mol. Cell. Biol. 23,
genes for obesity that affect shared metabolic pathways and networks. Nat.
5293–5300.
Genet. 41, 415–423.
Weaver, I.C., Cervoni, N., Champagne, F.A., D’Alessio, A.C., Sharma, S.,
Seckl, J.R., Dymov, S., Szyf, M., and Meaney, M.J. (2004). Epigenetic Youngson, N.A., and Whitelaw, E. (2008). Transgenerational epigenetic
programming by maternal behavior. Nat. Neurosci. 7, 847–854. effects. Annu. Rev. Genomics Hum. Genet. 9, 233–257.

1096 Cell 143, 1084–1096, December 23, 2010 ª2010 Elsevier Inc.