Research Article

Disruption of histone methylation in developing sperm impairs offspring health transgenerationally

See allHide authors and affiliations

Science  06 Nov 2015:
Vol. 350, Issue 6261, aab2006
DOI: 10.1126/science.aab2006

Generations affected by histone changes

Parent and even grandparent environmental exposure can transmit adverse health effects to offspring. The mechanism of transmission is unclear, but some studies have implicated variations in DNA methylation. In a mouse model, Siklenka et al. found that alterations in histone methylation during sperm formation in one generation leads to reduced survival and developmental abnormalities in three subsequent generations (see the Perspective by McCarrey). Although changes in DNA methylation were not observed, altered sperm RNA content and abnormal gene expression in offspring were measured. Thus, chromatin may act as a mediator of molecular memory in transgenerational inheritance.

Science, this issue p. 10.1126/science.aab2006; see also p. 634

Structured Abstract


Despite the father transmitting half of the heritable information to the embryo, the focus of preconception health has been the mother. Paternal effects have been linked to complex diseases such as cancer, diabetes, and obesity. These diseases are increasing in prevalence at rates that cannot be explained by genetics alone and highlight the potential for disease transmission via nongenetic inheritance, through epigenetic mechanisms. Epigenetic mechanisms include DNA methylation, posttranslational modifications of histones, and noncoding RNA. Studies in humans and animals suggest that epigenetic mechanisms may serve in the transmission of environmentally induced phenotypic traits from the father to the offspring. Such traits have been associated with altered gene expression and tissue function in first and second offspring generations, a phenomenon known as intergenerational or transgenerational inheritance, respectively. The mechanisms underlying such paternal epigenetic transmission are unclear.


Sperm formation involves rapid cell division and distinctive transcription programs, resulting in a motile cell with highly condensed chromatin. Within the highly compacted sperm nucleus, few histones are retained in a manner that suggests an influential role in development. Despite being the major focus of studies in epigenetic inheritance, the role of DNA methylation in paternal epigenetic inheritance is unresolved, as only minor changes in DNA methylation in sperm at CpG-enriched regions have been associated with transmission of environmentally induced traits. Instead, there may be a combination of molecular mechanisms underlying paternal transgenerational epigenetic inheritance involving changes in histone states and/or RNA in sperm. The function of sperm histones and their modifications in embryonic development, offspring health, and epigenetic inheritance is unknown. By overexpressing the human KDM1A histone lysine 4 demethylase during mouse spermatogenesis, we generated a mouse model producing spermatozoa with reduced H3K4me2 within the CpG islands of genes implicated in development, and we studied the development and fitness of the offspring.


Male transgenic offspring were bred with C57BL/6 females, generating the experimental heterozygous transgenic (TG) and nontransgenic (nonTG) brothers. Each generation from TGand nonTG animals (F1 to F3 in our transgenerational studies) was bred with C57BL/6 females, and the offspring (pups from generations F1 to F4) were analyzed for intergenerational and transgenerational effects. We found that KDM1A overexpression in one generation severely impaired development and survivability of offspring. These defects lasted for two subsequent generations in the absence of KDM1A germline expression. We characterized histone and DNA methylation states in the sperm of TG and nonTG sires. Overexpression of KDM1A was associated with a specific loss of H3K4me2 at more than 2300 genes, including many developmental regulatory genes. Unlike in other examples of paternal transgenerational inheritance, we observed no changes in sperm DNA methylation associated with primarily CpG-enriched regions. Instead, we measured robust and analogous changes in sperm RNA content of TG and nonTG descendants, as well as in their offspring, at the two-cell stage. These changes in expression and the phenotypic abnormalities observed in offspring correlated with altered histone methylation levels at genes in sperm. This study demonstrates that KDM1A activity during sperm development has major developmental consequences for offspring and implicates histone methylation and sperm RNA as potential mediators of transgenerational inheritance. Our data emphasize the complexity of transgenerational epigenetic inheritance likely involving multiple molecular factors, including the establishment of chromatin states in spermatogenesis and sperm-borne RNA.


Correct histone methylation during spermatogenesis is critical for offspring development and survival over multiple generations. These findings demonstrate the potential of histone methylation as a molecular mechanism underlying paternal epigenetic inheritance. Its modification by environmental influences may alter embryo development and complex disease transmission across generations. An essential next step is to establish functional links between environmental exposures, the composition of the sperm epigenome, and consequent altered gene expression and metabolic processes in offspring. Considering the mounting evidence, it may soon be reasonable to suggest that future fathers protect their sperm epigenome.

Disruption of histone methylation in developing sperm by exposure to the KDM1A transgene in one generation severely impaired development and survivability of offspring.

These defects were transgenerational and occurred in nonTG descendants in the absence of KDM1A germline expression. Developmental defects in offspring and embryos were associated with altered RNA expression in sperm and embryos.


A father’s lifetime experiences can be transmitted to his offspring to affect health and development. However, the mechanisms underlying paternal epigenetic transmission are unclear. Unlike in somatic cells, there are few nucleosomes in sperm, and their function in epigenetic inheritance is unknown. We generated transgenic mice in which overexpression of the histone H3 lysine 4 (H3K4) demethylase KDM1A (also known as LSD1) during spermatogenesis reduced H3K4 dimethylation in sperm. KDM1A overexpression in one generation severely impaired development and survivability of offspring. These defects persisted transgenerationally in the absence of KDM1A germline expression and were associated with altered RNA profiles in sperm and offspring. We show that epigenetic inheritance of aberrant development can be initiated by histone demethylase activity in developing sperm, without changes to DNA methylation at CpG-rich regions.

Birth defects occur in 3% of human babies and can be caused by genetic factors or environmental exposures, although 50% of birth defects are idiopathic (1, 2). The focus on prevention has largely been geared toward the mother, despite the father contributing half of the genetic information and possibly some epigenetic information to the embryo. Studies in humans and animals suggest that epigenetic mechanisms may serve in the transmission of environmentally induced phenotypic traits from parents to offspring (38). Such traits have been associated with altered gene expression and tissue function in first, second, and/or third offspring generations (38). Depending on the number of generations and parental origin, this phenomenon is referred to as intergenerational or transgenerational epigenetic inheritance (Fig. 1). Maternal and paternal transmission of such effects has been related to alterations of DNA methylation in germ cells (7, 911). DNA methylation occurs at the 5 position of cytosine residues and is associated with heritable gene silencing when promoter sequences containing multiple CpG dinucleotides [CpG islands (CGIs)] are methylated. Likewise, DNA methylation suppresses transcription of endogenous retrotransposable elements (12). In sperm, DNA methylation at gene promoters is uncommon, whereas repetitive elements are frequently methylated (1315). Despite being the major focus of studies in epigenetic inheritance, the importance of DNA methylation in paternal epigenetic inheritance is unresolved. Most studies reported only minor changes in DNA methylation in sperm at CpG-enriched regions that have been associated with the transmission of environmentally induced traits (5, 7, 8, 10, 11). Paternal transgenerational epigenetic inheritance may therefore involve changes in histone states (1618), RNA (19), and/or other sperm components.

Fig. 1 Transgenerational and intergenerational definitions in maternal and paternal epigenetic inheritance.

(A) In the pregnant female mouse (F0), exposure to environmental factors (toxicants, nutrients, stress) can alter the soma and the germ line of the F1 generation. These are considered intergenerational effects. When the F1 offspring are bred, any phenotypic effects in F2 animals are also considered to be intergenerational, as the F2 generation originates from the F1 germ cell that was exposed in utero. In maternal epigenetic inheritance, the F3 animals are the first generation with no direct connection to the exposure and are considered to have a transgenerational phenotype, should abnormalities be detected. (B) In males, exposure of the F0 mouse includes his sperm that fertilizes the oocyte to produce the F1 generation. Thus, phenotypic effects in the F1 animals are considered to be intergenerational. Breeding F1 mice to generate F2 animals results in the first unexposed generation. In contrast to females, F2 males and males of subsequent generations can be subject to transgenerational phenotypes [based on (64)].

During the final stages of sperm formation, chromatin undergoes extensive remodeling in which most histones are replaced by sperm-specific protamine proteins enabling extensive compaction of DNA in sperm nuclei (20). Nonetheless, a small percentage of histones (about 1% in mice and 15% in humans) are retained in mature sperm (16, 18). Several recent studies characterized the genome-wide distribution of nucleosomes in mature sperm of mice and humans, with partially discordant results (17, 18, 21), as reviewed by Saitou and Kurimoto (22). Using a protocol that enables efficient opening of the extremely compact mouse sperm nucleus, followed by micrococcal nuclease digestion and deep sequencing, we observed an approximately 10-fold overrepresentation of sequences localizing to promoters enriched in CpG dinucleotides in nucleosome-associated DNA (17). Histone retention in mouse sperm is predictable, occurring mainly at regions of high CpG density and low DNA methylation—for instance, at promoters of housekeeping and development-regulating genes (17). Retention at nonmethylated CGIs is conserved between mice and humans (1618, 23). In sperm, promoter regions of housekeeping genes contain the histone variant H3.3 and are marked by histone H3 lysine 4 (H3K4) di- and trimethylation. In contrast, developmental regulatory genes in sperm contain both H3.3 and canonical H3.1/H3.2 histones and are marked by trimethylation of H3K27 (H3K27me3) at their promoters (16, 17). The function of sperm histones and their modifications in embryonic development, offspring health, and epigenetic inheritance is largely unknown.

In Caenorhabditis elegans, loss of the H3K4me2 demethylase spr-5 is associated with a transgenerational phenotype appearing as progressive sterility due to stable accumulation of H3K4me2 in primordial germ cells (24). The mouse homolog, lysine demethylase 1A (Kdm1a), controls gene expression and development by catalyzing mainly the removal of mono- and dimethylation of H3K4 (25). To study the role of retained histones in sperm for embryo development and transgenerational epigenetic inheritance, we targeted H3K4 methylation, an epigenetic mark associated with developmental genes in sperm (17). We aimed to generate a mouse model for producing spermatozoa with reduced H3K4me2 within the CGIs of genes implicated in development. Toward this goal, we overexpressed the human KDM1A histone lysine 4 demethylase during spermatogenesis in mice and studied the development and fitness of offspring.

Offspring sired by heterozygous transgenic (TG) fathers and nontransgenic (nonTG) descendants had increased rates of birth defects, neonatal mortality, and altered gene expression. We characterized histone and DNA methylation states in the sperm of TG and nonTG sires. Overexpression of KDM1A was associated with specific losses of H3K4me2 in more than 2300 genes, including many developmental regulatory genes. Unlike in other examples of paternal transgenerational inheritance (11, 26, 27), we observed no changes in sperm DNA methylation at CpG-enriched regions. Instead, we measured robust and analogous changes in sperm RNA content of TG and nonTG descendants, as well as in their offspring at the two-cell stage. These changes in expression and the phenotypic abnormalities observed in offspring correlate with altered histone methylation levels for genes in sperm. This study demonstrates that KDM1A activity and related reduced histone methylation during sperm development have catastrophic consequences for offspring. Additionally, our study implicates histone methylation and sperm RNA as potential mediators of transgenerational inheritance.


The KDM1A transgene is specifically overexpressed in the mouse male germ line

We generated two lines of transgenic mice that express human KDM1A histone lysine 4 demethylase under the control of the gonad-specific, truncated form of the human polypeptide chain elongation factor 1 α (EF-1α) promoter (28) (Fig. 2A and fig. S1, A and B). This promoter is active in testicular germ cells but not somatic tissue (Fig. 2B and fig. S1, C and D). Transgenic KDM1A mRNA localized to spermatogonia and spermatocytes and weakly in round spermatids (Fig. 2C). A similar distribution was observed for the transgene marker green fluorescent protein (GFP) (fig. S1C). Histopathological evaluation of testes revealed normal spermatogenesis at all ages examined, from the first wave of spermatogenesis and throughout adulthood, in the offspring of both lines. Likewise, testis and epididymal weights, as well as sperm counts, were largely unaffected (fig. S2). To assess DNA integrity, we monitored phosphorylation of γH2AX and used terminal deoxynucleotidyl transferase–mediated deoxyuridine triphosphate nick end labeling (TUNEL) to detect DNA damage. The cellular distribution of H2AX phosphorylation in spermatogonia, spermatocytes, and spermatids was invariant between transgenic and control animals (fig. S3A). Likewise, we did not observe an increased number of TUNEL-positive cells in the testes of transgenic males (fig. S3, C and D). Finally, expression of the retrotransposon LINE-1 did not differ between TG, nonTG, and control testes (fig. S3B). Together, these data indicate that KDM1A expression does not induce DNA damage or genome instability (29).

Fig. 2 Transgenic KDM1A expression in developing male germ cells impairs the development and survival of offspring.

(A) Human (hs) KDM1A is expressed under the control of the human elongation factor 1 alpha promoter (EF1α), which has a truncated regulatory region that drives transcription in germ cells. Kozak (K) sequence, internal ribosomal entry site (IRES), EGFP, and bovine growth hormone (BGH) are present in the polyadenylation signal. (B) Reverse transcriptase PCR analysis shows that KDM1A transgene expression is restricted to the testis. (C) In situ hybridization reveals KDM1A mRNA (blue) in spermatogonia (SG), spermatocytes (SC), and round spermatids (RS). Elongated spermatids (ES) were negative. Scale bars, 10 μm. (D and E) Pedigrees. (D) In line 1, the female heterozygous transgenic founder (TG0) was bred with a C57BL/6 male to generate TG1 male pups and their nontransgenic brothers (nonTG1). We subsequently bred heterozygous transgenic males (TG1-4) with C57BL/6 females. We assayed the effects of paternal transgenic exposure on nonTG descendants over multiple generations (nonTG3-5). (E) In line 2, a female heterozygous transgenic founder (TG0) was bred with a WT C57BL/6 male to generate TG1 offspring. The first breeding of the founder resulted in only viable female transgenic offspring. Therefore, from the TG2 generation onward, the transgene was passed through the father. For both lines, we bred TG2-4 and nonTG3-5 males with WT C57BL/6 females. After E18.5 embryonic analysis from line 1 TG3 sires and their nonTG3 brothers, we sacrificed the males and used their sperm for epigenomic analysis. (F and G) Reduced survival of offspring sired by TG2-3 and nonTG3 males in line 1 (G) and TG2-4 males in line 2 (H), in comparison with C57BL/6-sired offspring. Survivor curves with different letters are significantly different (log rank statistical test). n indicates the number of offspring; at P < 0.05, as marked by asterisks (H and I) PND 6 TG-sired pups with abnormal skin are shown in comparison to a normal pup sired by a control mating (asterisk).

Transgenerational paternal effects in TG- and nonTG-sired descendants

We conducted breedings from TG sires and nonTG sires to determine possible inter- and transgenerational paternal effects of KDM1A overexpression on offspring fitness. Pedigrees and generations are shown in Fig. 2. Male transgenic offspring were bred with C57BL/6 females, generating the experimental heterozygous TG and nonTG brothers. Males from TG2-to-TG4 (TG2-4) and nonTG3-to-nonTG5 (nonTG3-5) generations (representing F0 to F3 in our transgenerational studies) were bred with C57BL/6 females, and offspring (pups from generations F1 to F4) were analyzed for intergenerational and transgenerational effects (Fig. 2, D and E). First, we analyzed the pups of TG and nonTG sires over multiple generations for development and survival (Fig. 2, F to I). We determined the litter size at birth and weighed, sexed, and examined the pups 36 and 48 hours after birth and on postnatal days (PND) 6 and 21 (table S1). Neonatal survival rates of offspring sired by TG2-4 and nonTG3 males were reduced compared with survival rates of C57BL/6 controls (P < 0.05) (Fig. 2, F and G). We observed cumulative effects of transgene expression on survivability: With each successive generation of exposure of developing sperm to the transgene (TG2-4), survivability of transgenic offspring (encompassing generations F1 to F3) decreased in both lines (P < 0.05) (Fig. 2, F and G). We also observed pups that were abnormal and runted (<75% of average body weight of the litter), with apparent abnormalities in the limbs, skeleton (table S1), and skin (flaky as pups, gray-and-white flecked as adults) (Fig. 2, H and I). Analysis of offspring sired by individual males showed that the described phenomenon is not driven by transmission of abnormalities by a few males to many offspring but by many males transmitting epigenetic alterations to a variable number of offspring (fig. S4). Moreover, frequencies of pup survival and abnormalities were not related to whether or not the offspring inherited the transgene from a transgenic father (P < 0.05). The presence of possible alterations in the epigenome of haploid spermatozoa that do not carry the actual transgene is probably due to the fact that in TG heterozygous sires, KDM1A is active in germ cells at various stages of spermatogenesis, such as spermatogonia and spermatocytes, which are diploid cells. Moreover, even after the two meiotic divisions, haploid spermatids share mRNAs via cellular bridges resulting from incomplete mitotic and meiotic divisions. Consequently, although the transgene is transcribed in only 50% of spermatids, all haploid spermatids within the cellular syncytium will be affected (30, 31). Because transgene expression is restricted to the developing germ line (Fig. 2, B and C, and fig. S1, C and D) and nonTG sires do not express the transgene in their developing sperm, these data indicate that the aberrant phenotypes observed in nonTG pups (F1 to F3) are related to transgenerational epigenetic phenomena.

C57BL/6 mothers likely culled pups that were perceived as abnormal. As a consequence, we may have underestimated the phenotypic abnormalities assessed in the pups (table S1). To address this possibility and to further characterize the developmental defects across generations, we performed detailed phenotypic analysis at embryonic day 18.5 (E18.5). Each TG2-4 (F0 to F2) and nonTG3-5 (F1 to F3) sire was mated with at least two females, and pregnancies were assessed for litter size, pre- and postimplantation loss (combined to give total pregnancy loss), and abnormal fetuses (Figs. 3 and 4). In TG-sired offspring, abnormalities were varied and affected multiple systems. Examples of developmental errors included skeletal anomalies observed as malformed digits, spinal defects, abnormal craniofacial structures, and failures in the development of limbs and body segments (Figs. 3 and 4). Likewise, in pups sired by nonTG3 (F1) and nonTG4 (F2) animals, we observed F2 and F3 offspring with visible abnormalities and skeletal defects occurring with significantly greater frequency than in control offspring (P < 0.05) (Fig. 3 and table S2). Abnormalities in F3 offspring (sired by a nonTG4 father) represent transgenerational effects (Figs. 3 and 4). Rates of pregnancy loss increased in several TG and nonTG generations (Fig. 3). No gross visible abnormalities were observed in E18.5 F4 offspring sired by nonTG5 males (Fig. 3). To further identify the specific nature of the skeletal defects and to confirm observations in nonTG offspring, a subset of pups was analyzed specifically for skeletal abnormalities. These analyses revealed abnormalities such as errors in ossification, spinal defects, and missing or abnormal craniofacial structures in both TG and nonTG offspring (Fig. 4 and table S2). The dilution of the gross abnormalities in F4 offspring sired by nonTG5 males was confirmed by our skeletal analysis (table S2). Thus, by three generations after transgene exposure, we observed a normalization of offspring phenotypes in terms of birth defects at the levels detectable in our analysis (gross morphological and skeletal). Nonetheless, we have probably underestimated offspring abnormalities, as in-depth pathology analysis was not performed, and abnormalities were limited to those detected by observation of pups or skeletal analysis.

Fig. 3 E18.5 fetuses sired by TG and nonTG descendants show a range of abnormalities.

(A) Phenotypic frequencies observed in embryonic day E18.5. Total pregnancy loss per group [%] = (no. of CL – no. of total embryos) per group/no. of CL per group × 100; no. of abnormal fetuses per group [%] = (sum of abnormal fetuses per group/no. of embryos per group) × 100. Major abnormalities included craniofacial and skeletal defects, edema, hemorrhagic gut, extra digits, and missing eyes. Minor abnormalities included pale skin, color disfigurement, and growth retardation. (B) Normal E18.5 fetus sired by a C57BL/6 control father. (C) Runt with limb abnormality (arrow) and hemorrhagic foci. (D) Extra digit on left hindpaw. (E) Blunted snout, underdeveloped forearms and digits (thin arrow), and multiple hemorrhagic foci (wide arrow). (F) Craniofacial abnormalities and edema. (G) Hemorrhagic gut and underdeveloped limbs with failure in ossification. (H to J) Severely malformed fetuses. (K) Umbilical hernia. (L) Line 1 and (M) line 2 total pregnancy loss per group. (N) Line 1 and (O) line 2 total abnormalities per group (line 1: C57BL/6, n = 298 abnormalities; TG2, n = 134; TG3, n = 197; nonTG3, n = 142; nonTG4, n = 249; nonTG5, n = 186; line 2: C57BL/6, n = 298; TG3, n = 30; TG4, n = 144; nonTG4, n = 139). Total pregnancy loss per group [%] = (no. of CL – no. of total embryos) per group/no. of CL per group × 100; abnormal fetuses per group [%] = (sum of abnormal fetuses per group/no. of embryos per group) × 100. Statistical test for litter size: Student’s t test; statistical test for total pregnancy loss and abnormalities per group: Fisher’s exact test. *P < 0.05; **P < 0.01; ***P < 0.001.

Fig. 4 Common skeletal abnormalities in pups sired by TG and nonTG3-4 descendants.

(A) Control-sired C56BL/6 pup at E18.5 and the corresponding skeletal stain. Red, bone; blue, cartilage. (B) nonTG4-sired fetus at E18.5 and the corresponding skeletal stain. The nonTG4-sired fetus had an abnormal craniofacial structure with reduced parietal ossification (arrow) and an extended cartilage deposit in the snout. Front limb digits were also underdeveloped (arrow). (C) Control C57BL/6-sired fetus. (D) TG3-sired fetus with hypo-ossified parietal, frontal, and supraoccipital bones and abnormal eye socket. (E) Lack of temporal bones (arrow), hypo-ossified skull, and abnormal snout. (F) nonTG4-sired fetus micrognathia. (Refer to Fig. 2D for origin or generation and nomenclature and fig. S5 for a reference for normal skeletal anatomy and staining.)

Transgenic KDM1A expression alters H3K4 dimethylation in sperm

To relate the developmental defects in offspring to possible changes in chromatin states in parental spermatozoa, we profiled the KDM1A target H3K4me2 in a genome-wide manner in sperm from TG3 males (n = 11 animals), their nontransgenic littermates (nonTG3, n = 9), and age-matched C57BL/6 controls (n = 11), using the previously mentioned method that was specifically developed for analyzing chromatin states in highly condensed sperm (32). We prepared nucleosomes by micrococcal nuclease digestion and performed chromatin immunoprecipitation (ChIP) with H3K4me2-specific antibodies, followed by next-generation sequencing (ChIP-seq) (fig. S6). In read-count–normalized libraries, we observed a reduction in H3K4me2 enrichments at CpG-rich sequences within transcription start site (TSS) regions of different genes in TG3 sperm compared with that of controls (Fig. 5A). To identify TSS regions with altered H3K4me2 profiles in a genome-wide manner, we calculated the enrichment of H3K4me2 for genomic regions encompassing 250 base pairs (bp) upstream and downstream (+/−) of TSS for the different sperm samples. Because we previously showed that nucleosomes are retained in mouse spermatozoa at CpG-rich sequences [e.g., close to TSS (17)], the selected windows thus correspond to regions with the highest nucleosome occupancy in sperm and are most likely to transmit any histone-encoded information to the offspring. Reduced H3K4me2 levels occurred at 28.7% of the 8171 TSS regions marked by H3K4me2 in TG3 samples, whereas 3.4% of the TSS regions displayed modest elevated levels of H3K4me2 (Fig. 5, B and C). The predominant down-regulation of H3K4me2 levels is consistent with the reported H3K4me2 demethylase activity of KDM1A.

Fig. 5 H3K4me2 occupancy is severely reduced at nucleosome-containing TSS regions of CGI genes in TG3 sperm only.

(A) Snapshots of H3K4me2 occupancy at different loci in sperm of TG3 and WT sires (this study). Occupancies of H3K4me3, H3K27me3, and H3.3 in sperm and of Kdm1a in spermatocytes of WT animals were taken from (17, 33). Pdpk1 and E2F6 represent loci marked by high and intermediate H3K4me3 occupancy, respectively, whereas Gsc represents H3K27me3-marked genes, as observed in WT sperm. All three genes show reduced H3K4me2 levels at TSS sequences strongly enriched for H3.3-containing nucleosomes in WT sperm. We interpret the high histone modification occupancies at flanking sequences with low H3.3 nucleosome occupancy as the result of high ChIP efficiencies even at sites with low histone retention levels in the majority of sperm. (B and C) Heatmaps showing reduced (B) or unchanged (C) H3K4me2 occupancy at TSSs in TG3 sperm for three groups of genes classified according to the H3K4me3 and H3K27me3 states in WT sperm. The plots show CpG density; nucleosomes; and H3.3, H3.1/H3.2, H3K4me3, and H3K27me3 coverages in WT mice (17) and H3K4me2 coverages around TSSs (±3 kilobases) in sperm of C57BL/6 WT, KDM1A TG3, and nonTG3 littermates and Kdm1a coverage in a WT spermatocyte (33). (D to F) Boxplots for the distributions of CpG percentage (D), Kdm1a occupancy around TSSs (±500 bp) (E), and H3.3 enrichment around TSSs (±250 bp) (F) for different groups of genes. “Down” and “unchanged” refer to genes with reduced or unchanged levels, respectively, of H3K4me2 in sperm of KDM1A TG3 males. Number of genes per group is as follows (from left to right): 507, 1121, 1437, 3840, 397, and 593.

Direct sequencing of nucleosome-associated DNA did not reveal differences in nucleosome occupancies between TSS regions with differential H3K4me2 levels in TG3 sperm, nor were differences observed between TG3 mice, nonTG3 littermates, and control samples (fig. S7). These results exclude impairment of nucleosome retention as a putative cause underlying the reduction of H3K4me2 at certain loci in TG3 sperm. To assess the reasoning underlying reduced H3K4me2 occupancy levels at certain TSSs, we categorized the regions according to CpG density and various chromatin attributes, such as the occupancy of the histone variant H3.3 and levels of H3K4me3 and H3K27me3, as measured in the sperm of control animals (17); these attributes are associated with specific gene functions (Fig. 5, B and C) (17). We found that regions with reduced H3K4me2 levels are CpG-rich (Fig. 5D), which suggests that targeting of human KDM1A is related to CpG density. Endogenous murine Kdm1a is particularly localized at CpG-dense TSS regions in mouse spermatocytes that are characterized by H3K4 methylation in sperm (17, 33) (Fig. 5, A to C). Moreover, regions with reduced H3K4me2 levels in TG3 sperm have higher occupancy levels of murine Kdm1a as compared with regions that have unchanged H3K4me2 levels (Fig. 5E). This finding points to specificity in human KDM1a activity at target sites of endogenous murine Kdm1a. Together, these data suggest that transgenic human KDM1A is targeted to selected regions that are extensively bound by endogenous murine Kdm1a during spermatogenesis.

As measured in control sperm, TSS regions with reduced H3K4me2 levels are strongly enriched for H3.3 (Fig. 5F) (17). Because high levels of H3.3 at strong CGIs in sperm correspond to high nucleosome turnover at CGIs in spermatids (17), locally reduced H3K4me2 levels at strong CGI promoters in TG3 sperm may reflect the inability of H3K4 histone methyltransferases to counteract local TG-conferred KDM1A activity and may thereby fail to reestablish the H3K4me2 state during nucleosome turnover in spermatids of TG3 males.

The results of our gene ontology analysis suggest that transgenic KDM1A regulates specific classes of genes (tables S3 and S4). Genes of the H3K4me3 high- and intermediate-chromatin classes with reduced H3K4me2 levels in TG3 males serve functions in various cellular metabolic protein processes. For example, in TG3 sperm, reduced levels of H3K4me2 were observed for Pdpk1 (3-phosphoinositide dependent protein kinase 1) and E2F6 (Fig. 5A). Conditional ablation of Pdpk1 in the pancreas leads to diabetes (34), and knockout mice have impaired embryogenesis, growth, and nervous system development (35, 36). The loss of E2F6 results in skeletal defects (37). H3K27me3-marked genes with reduced levels of H3K4me2 serve various functions during embryonic development. For instance, Gsc (Goosecoid homeobox) has a reduced amount of H3K4me2 in TG3 sperm at a region overlapping H3K27me3, H3K4me3, and Kdm1a binding (Fig. 5, B and C). A targeted mutation of Gsc gave rise to mice with craniofacial abnormalities and skeletal defects (38). These phenotypes reported in gene ablation models are highly reminiscent of those observed in TG1-4 and nonTG3-4 offspring, yet the connection to altered histone methylation in developing sperm remains unclear. Our gene ontology analysis of embryonically expressed genes indicates that genes with a loss of H3K4me2 are involved in morphogenesis, patterning, vasculature development, cartilage development, and catabolic and metabolic processes (table S4). In contrast, H3K4me3 high- and intermediate-chromatin genes with unchanged H3K4me2 levels in TG3 males execute functions in spermatogenesis and various nuclear processes, respectively. These results emphasize the targeting of hKDM1A to specific subsets of genes that may preferentially alter offspring development. Moreover, the gene subsets are directly in line with our phenotypes, where defects were not observed on spermatogenesis in transgenic sires but were seen in the TG-sired offspring.

Changes to histone H3K4me2 are independent of DNA methylation state

In contrast to TG3 spermatozoa, our ChIP-seq analysis of sperm from nonTG3 males (littermates of TG3 animals) did not reveal any differences in H3K4me2 occupancy in comparison to sperm from TG3 or control males (Fig. 5, B and C). These data suggest that the phenotypic aberrations seen in the offspring of nonTG3 sires are not directly related to reduced H3K4me2 levels, as observed in mature sperm of TG3 sires. Previously, KDM1A was reported to mediate the removal of H3K9 mono- and dimethylation in some cases (39, 40). In various somatic cells, H3K9me2 is an abundant heterochromatic modification that is localized in large domains throughout the genome (41). H3K9me2 generally does not localize to and is mutually exclusive with regions marked by H3K4me2 and/or H3K27me3, which are prevalent at CGI regions containing nucleosomes in sperm (41). Such anticorrelation minimizes a possible contribution of H3K9me2 to paternal transmission of the embryonic impairment traits observed in TG and nonTG offspring.

Altered DNA methylation has been associated with phenotypes induced by paternal environmental exposures and has been suggested as a mechanism underlying epigenetic inheritance (42). To address whether reduced H3K4me2 may cause heritable changes in DNA methylation levels, we first used a targeted approach to assess DNA methylation in sperm from TG3 (F1) and nonTG3 (F1) littermates. Target selection was based on enrichment for CpGs and identification of regions as being differentially methylated in TG3 sperm for H3K4me2 in comparison to C57BL/6 controls (three overlapping windows of 250 bp each). Selected targets were analyzed for altered DNA methylation by quantitative Sequenom MassARRAY, a technique based on bisulfite conversion followed by mass spectrometry (MS) analysis with a resolution at the CpG level. In all 24 targeted genes, we failed to observe any significant differences in DNA methylation in TG3 and nonTG3 sperm in comparison to controls (fig. S8).

Next, we used reduced representation bisulfite sequencing (RRBS) for a genome-wide comparison of DNA methylation levels, predominantly at CpG islands in control, TG3, and nonTG3 sperm. Analysis based on methylation percentages of single CpG sites found across the genome showed a high degree of similarity between samples of all three genotypes (R = 0.98 to 0.99) with no apparent group clustering (fig. S9A). When testing individual CpG sites for significant association with genotype, we found only a few more than expected by chance (fig. S9, B and C). In summary, using targeted Sequenom analysis and RRBS, we did not observe an overrepresentation of changes in CpG methylation levels in spermatozoa of TG3 and nonTG3 transgenic mice versus control samples. These findings indicate that DNA methylation at CpG islands is not implicated in the molecular processes leading to the observed transgenerational phenotypes.

Overexpression of KDM1A is associated with altered sperm RNA content

RNA analysis of sperm from TG3 and nonTG3 mice revealed a common nongenetic molecular change (Fig. 6). Sperm is a rich source of diverse RNAs that may function as potential mediators of paternally transmitted effects (43). We used the Affymetrix GeneChip ST2.0 Array to compare the RNA content of sperm from TG3 and nonTG3 mice to that of controls. This method enables the detection of 28,000 coding transcripts and more than 7000 noncoding RNAs, including ~2000 long intervening/intergenic noncoding RNAs (lincRNAs). TG3 and nonTG3 spermatozoa displayed a comparable molecular change in RNA content relative to controls, with 564 RNAs commonly misregulated among 650 RNAs altered in TG3 sperm and 619 in nonTG3 sperm [log2 fold change > 1.0; false discovery rate (FDR) < 0.05]. Among the shared misregulated RNAs in nonTG and TG sperm, 67 were noncoding RNAs and 471 were protein-coding transcripts (table S5). As would be predicted with reduction of a gene-activating histone modification, 99% of differential RNAs were reduced in TG3 and nonTG3 sperm compared to controls (643 of 650 and 613 of 619, respectively). More than 60% of RNA-associated promoters were marked by H3K4me3 in wild-type (WT) sperm, potentially reflecting transcriptional activity at preceding stages of male germ cell development (17) (Fig. 6B). Only ~3% of RNAs (15 of 564) were marked by H3K27me3 in WT sperm. Finally, 41 commonly deregulated transcripts were associated with genes with reduced H3K4me2 levels at their TSS in TG3 sperm (Fig. 6B). These findings present the possibility that transcripts and regulatory RNAs are transmitted via TG3 and nonTG3 sperm to the embryo and may be implicated in transmission of the phenotype to the offspring.

Fig. 6 Differential gene expression in two-cell embryos as related to sperm chromatin content.

(A) RNA content in TG3 and nonTG3 sperm is depicted by a heatmap of selected genes, as related to identified regions of H3K4me2 loss in TG3 sperm or regions enriched for H3K4me3 or H3K27me3 in C57BL/6 sperm. (B) Venn diagrams comparing the number of RNA transcripts with significantly different levels of abundance in the sperm of TG3 and nonTG3 males with respect to controls. (C) Heatmap depicting selected RNAs that differed from control-sired embryos and those sired by TG8 or nonTG8 males (F2 generation embryos), as related to identified regions of H3K4me2 loss in TG3 sperm or sperm-enriched regions for H3K4me3 or H3K27me3. (D) Venn diagrams depicting differential gene expression in two-cell embryos sired by TG8, nonTG8, or control animals, as related to identified regions of H3K4me2 loss in TG3 sperm or sperm-enriched regions for H3K4me3 or H3K27me3. (E) The phenotypes observed in TG and nonTG offspring resemble those of mouse mutants for differentially expressed genes.

TG and nonTG sperm alter gene expression in early embryos

To understand whether there was a link between genes bearing altered histone methylation in the TG3 sperm to embryo gene expression and offspring phenotypes, we performed array expression analysis from TG8, nonTG8, and control-sired two-cell embryos. We chose to examine two-cell embryos because, in terms of developmental timing, they are close to sperm in which we observed H3K4me2 changes. We hypothesized that potential changes in paternal chromatin states would be the least diluted in two-cell embryos compared with later stages of development. We identified 874 RNAs in embryos sired by TG8 and 123 RNAs in nonTG8 embryos (F2 generation) that were differentially expressed (table S6) (FDR < 0.3). Most deregulated genes in TG8-sired embryos were up-regulated (80%; 703 of 874 RNAs), whereas about half of the genes were up-regulated in nonTG8-sired embryos (46%; 56 of 123 RNAs). Notably, 71 genes were commonly misexpressed in two-cell embryos sired by TG8 and nonTG8 littermates (Fig. 6, D and E, and table S6) (FDR < 0.3). Moreover, 110 genes deregulated in two-cell embryos had reduced H3K4me2 at TSSs in TG3 sperm. A subset of these genes is also marked by H3K27me3 and H3K4me3 (Fig. 6, C to E). Some of the differentially expressed genes in the embryos that overlap with reduced H3K4me2 in TG3 sperm, or regions in sperm enriched in H3K27me or H3K4me3, can be linked to the developmental abnormalities observed in TG- and nonTG-sired pups (Fig. 6E). These data establish that the triggering event leading to altered H3K4me2 in developing sperm has consequences for gene expression in embryos.


Here we show that increased expression of the chromatin modifier KDM1A during spermatogenesis induces major developmental defects, which were transmitted paternally for three generations, even in the absence of transgene expression in the germ line of nontransgenic offspring. Our data suggest that the critical initiating event in our model is the alteration in histone methylation in developing male germ cells, leading to offspring abnormalities. We show that changes to histone methylation in the hKDM1A model occurred in the absence of changes to DNA methylation at CGIs in sperm. We also observed altered RNA profiles in sperm of TG and nonTG males (F1 generation) and their two-cell embryonic offspring (F2 generation) (Fig. 6). Our data emphasize that the mechanisms of transgenerational epigenetic inheritance are complex and probably involve several molecular factors, such as the establishment of chromatin states in spermatogenesis, DNA methylation, and sperm-borne RNA.

We observed a large diversity in aberrant phenotypes, including abnormalities in bone and skin, reduced survivability, and retarded growth. These aberrations were present in many offspring sired by many fathers, rather than many offspring from a few fathers. Together with the high penetrance of phenotypes that was resolved by the fourth generation of nontransgenic offspring, these observations point to an epigenetic mechanism driving impaired development. It is unlikely that hKDM1A expression during spermatogenesis behaves comparably to a chemical mutagen or induces a mutator phenotype in germ cells or offspring, given the very high frequency of abnormalities observed (4446). Consistent with that notion, our assessments using multiple approaches revealed no DNA damage or chromatin instability. Additionally, the normal sperm-head morphology lends further support to the stability of sperm chromatin. Human and mouse mutants with altered sperm histone content have misshapen sperm heads that lack compaction (47, 48). Unlike prior studies of transgenerational inheritance, however, we used several methods—including genome-wide DNA methylation analysis by RRBS and targeted analysis with Epityper—to confirm that DNA methylation occurring mainly at CGIs was not altered in TG or nonTG sperm.

Transgenic KDM1A expression induces a major local reduction in H3K4me2 occupancy at many CGI-containing promoters in sperm from TG sires. Reductions in sperm H3K4me2 levels are observed at sites of high Kdm1a occupancy in spermatocytes, supporting the notion that transgenic KDM1A functions preferentially at sites of endogenous Kdm1a proteins. Because nonTG3 animals sire offspring with developmental defects, as do TG3 males, reduction in H3K4me2 as observed in TG3 sperm probably does not directly mediate epigenetic inheritance of developmental defects across generations. Instead, our data point to the idea that disruptions in H3K4me2 have a cascading effect, altering transcription at genes in spermatogenesis and the RNA content of sperm. Correspondingly, H3K4me2 changes in sperm were correlated with changes in expression from genes bearing H3K4me3, and those bearing H3K27me3, that are normally already present in sperm at CGIs of subsets of genes known to control embryonic development (17). Correspondingly, genes with altered expression in two-cell embryos sired by either TG or nonTG males were associated with genes in sperm enriched in histone methylation (H3K4me2, H3K4me3, and H3K27me3). This suggests that altered histone methylation in sperm can influence early embryo gene expression. Similarly, mutation of the Smarca5MommeD4 allele of the Snf2h chromatin remodeler during spermatogenesis has been associated with changes in expression of the epigenetic-sensitive agouti viable yellow (Avy) gene. As in our study, WT offspring of heterozygous Smarca5MommeD4 sires had variable phenotypes (49), which indicates that proper chromatin regulation during spermatogenesis plays an important role in offspring fitness and gene expression.

Our data also offer the possibility of sperm RNA content as a contributing factor to the offspring phenotypes. We observed substantial overlap in the RNA content of TG and nonTG sperm, with ~85% similarity in RNAs that were different from control sperm RNAs. An example of paternal effects via RNA was demonstrated in C. elegans, where induced viral expression led to paternal transmission of small viral RNA (viRNA) across multiple generations. As in our model, the WT offspring worm descendants inherited the viRNA molecules and the phenotype of viral protection. This transgenerational effect was extinguished by the third generation beyond exposure to the transgene, thus indicating a gradual loss of the phenotype, consistent with our observations (50). However, the likelihood that a diffusible agent such as RNA could drive the observed transgenerational phenotype is low, because germ cell fate in mammals is specified during gastrulation, which occurs many cell divisions after fertilization. In the absence of a means to amplify a diffusible RNA signal, this signal will be lost over the many cell divisions until germline specification. Therefore, we propose a model for mammals in which altered chromatin in developing sperm is associated with abnormal sperm-borne RNAs. These RNAs could then signal to chromatin to regulate gene expression during development (Fig. 7). Regulatory RNAs made up a significant portion of the differential RNA content in sperm from TG and nonTG males compared with controls. Functions of regulatory RNA include control of pluripotency, differentiation, and guidance in the setting of the epigenome (5153). Sperm is known to transmit long noncoding RNAs (lncRNAs) to the embryo; these RNAs have been postulated to affect the postfertilization genome (19). If lncRNAs are not correctly transcribed in developing germ cells, altered lncRNA levels could then be transported to the embryo, where gene expression could be modified. LncRNAs interact with chromatin regulatory machinery and serve in guiding chromatin changes. Moreover, lincRNA knockouts (Fendrr, Peril, and Mdgt) display peri- and postnatal lethality before PND 20 (54), a common observation in our transgenic descendants.

Fig. 7 Disruption of histone methylation in developing sperm by KDM1A transgene overexpression from one generation severely impaired embryo development and survivability of offspring.

These defects were transgenerational and occurred in nonTG descendants in the absence of KDM1A germline expression, which suggests that regions in the nonTG germline escape epigenetic reprogramming. Developmental defects in offspring were associated with altered RNA content in sperm and gene expression in embryos.


Our data show that, independent of DNA methylation at CGIs, abnormalities in histone methylation during spermatogenesis are associated with altered embryo gene expression and development. Our findings of transgenerational epigenetic inheritance indicate genetic lesions in chromatin modifiers and environmentally induced alterations in histone methylation during spermatogenesis as underlying causes of birth defects and disease that may be traceable to the father. Individual responses to environment and inheritance of sensitivities and/or predisposition to disease may be shaped by factors such as gene copy number of chromatin modifiers.

Materials and methods

Generation of KDM1 transgenic mice

For the generation of KDM1 TG mice, full-length human KDM1 cDNA [National Center for Biotechnology Information (NCBI) Nucleotide Database and Consensus CDS Database accession numbers NM_015013 and CCDS30627, respectively] was subcloned into a modified IRES-EGFP (internal ribosomal entry site–enhanced green fluorescent protein) vector (Clontech). We replaced the cytomegalovirus promoter of the expression vector by an isoform of the human EF-1α promoter driving germ cell–specific expression of the KDM1 transgene (28). We microinjected the transgenic construct into C57BL/6 zygotes and obtained two female transgenic founders. We crossed founders with C57BL/6 males to create the two transgenic lines. We subsequently transmitted the transgene by heterozygous males bred to C57BL/6 females. The two transgenic lines showed similar developmental phenotypes in offspring and served as a control for transgene integration site effects. We extracted genomic DNA from tail biopsies and performed Southern blot analysis to determine the transgene copy number. All C57BL/6 control mice were obtained from Charles River Laboratories International (Wilmington, MA). Mice were provided with water and standard mouse chow ad libitum and housed under conditions of controlled light (12-hour light, 12-hour dark cycle) at 21°C and 50% humidity. All animal procedures were approved by the Animal Care and Use Committee of McGill University, Montreal, Canada.

In situ hybridization

In situ hybridizations were performed as previously described (55), using digoxigenin-labeled RNA probes (Roche).

Genotyping by PCR on genomic DNA

Genotyping was performed on tail-tip DNA as previously described (56). Oligonucleotide primers used for genotyping by polymerase chain reaction (PCR) were: hsEF1alpha-fw1 (TTC TCA AGC CTC AGA CAG TGG), Flag-re1 (TCG TCA TCG TCC TTG TAG TCC), hsLSD1-ex19-fw1 (TAC GAT CCG TAA CTA CCC AGC), and pIRES-EGFP-re2 (TCT TAG CGC AGA AGT CAT GCC).

RNA isolation and reverse transcriptase PCR

Total RNA was extracted using Purezol (BioRad, Hercules, CA) according to the manufacturer’s instructions and was treated with DnaseI (RNeasy Mini Kit, Qiagen). Isolated RNA was reverse transcribed with random primers using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Foster City, CA). cDNAs were subjected to PCR using the following oligonucleotide primers: beta-Act1-fw [5′ ACC TTC AAC ACC CCM GCC ATG TAC G 3′ (M = A/C)], beta-Act2-re [5′ CTR ATC CAC ATC TGC TGG AAG GTG G 3′ (R = A/G)], hKDM1A-ex2-fw1 (5′ TGG ATG AAA GCT TGG CCA ACC 3′), hsLSD1-ex3-re1 (5′ AGA AGT CAT CCG GTC ATG AGG 3′), and hsKDM1A-ex11-fw1 (5′ GCC ACC CAG AGA TAT TAC 3′). For embryo transgene expression analysis, major organs from three transgenic embryos at E18.5 were homogenized via mortar and pestle and then extruded through a 25-gauge needle. RNA was extracted using the Qiagen RNeasy RNA purification kit (product no. 1050349, lot no. 145015691). RNA (1 μg) from each sample was converted to cDNA using the High-Capacity cDNA Reverse Transcriptase Kit (Applied Biosystems; product no. 4368814, lot no. 1108153). To determine expression of hsKDM1A, hsKDM1A-ex2-fw1 and hsLSD1-ex3-re1 primers were used as above. Primers for Gapdh were used as a loading control (forward: 5′ ACT TTG GCA TTG AAG GGC TG 3′; reverse: 5′ TGG AAG AGT GGG AGT TGC TGT TG 3′).

The PCR protocol was as follows: predenaturation of one cycle at 94°C for 4 min, followed by PCR amplification repeated for 40 cycles of denaturation at 94°C for 40 s, annealing at 60°C for 25 s, and elongation at 72°C for 60 s. The final elongation was performed at 72°C for 10 min. PCR reactions were visualized on the Qiaxcel Advanced System using the QiAxcel DNA Screening Kit (Product 74104; Lot 145019088)

Sperm isolation, count, and morphology

For sperm counts, spermatozoa were recovered from paired cauda epididymides of sexually mature mice as follows: Cauda epididymides were placed into phosphate-buffered saline (PBS), cut, and gently agitated at 37°C to allow sperm to swim out. After 5 min of incubation, a 10-μl aliquot of the sperm solution was removed and subjected to hemacytometric counts following standard procedures. For isolation of sperm for ChIP-seq, a swim-up procedure was used to ensure pure cell population (32).

Analysis of pups

We bred C57BL/6 (n = 18 animals), line 1 TG2-3 (n = 10 and 14) and nonTG3-5 (n = 8, 7, and 4), and line 2 TG2-4 (n = 4, 6, and 5) males with C57BL/6 females. Shortly after copulation, as determined by the presence of a vaginal plug, males were removed from females. Litter size was determined at birth, and pups were sexed, weighed, and examined at 36 and 48 hours after birth, as well as on PND 6 and 21. TG and nonTG pups with and without external malformations and C57BL/6 controls were subjected to skeletal staining. Abnormal pups were identified by the appearance of gross morphological defects, such as the presence of a skin abnormality, underdeveloped limb, or runting. Frequency of morphologically abnormal pups was calculated based on the total number of offspring per generation.

Analysis of E18.5 fetuses and assessment of pregnancy loss

We bred C57BL/6 (n = 21 animals), line 1 (L1) TG2-3 (n = 8 and 12) and nonTG3-5 (n = 10, 16, and 14), and line 2 (L2) TG3-4 (n = 2 and 9) and nonTG4 (n = 9) males with C57BL/6 females. Control matings were age-matched so that statistical comparisons were always between matings in which the sires were the same age. Copulation was determined by the presence of a vaginal plug. The day of the plug was established as E0.5. Fetuses were collected from the uteri of the female mice at E18.5 after mating. Each fetus was carefully examined, and sex, weight, and crown-to-rump length were determined. Moreover, placental gross morphology, weight, and diameter were ascertained. Fetuses with external malformations and the corresponding controls were subjected to skeletal staining or were fixed in Bouin’s fixative solution and subsequently subjected to histopathological analysis. We determined the number of ovulations by counting the number of corpora lutea (CL) in E18.5 pregnant females from the above matings (C57BL/6, n = 32; L1 TG2-3, n = 15 and 24; L1 nonTG3-5, n = 18, 30, and 20; L2 TG3-4, n = 4 and 16; L2 nonTG4, n = 17; n = number of females per group). Ovaries were placed into PBS in separate drops, and CL were counted blindly with respect to the number of embryos actually recovered using a dissecting microscope with top lighting. The sum of pre- and postimplantation losses was used as a measure of total pregnancy loss. This was done by comparing the number of CL produced with the number of total embryos per group: total pregnancy loss per group [%] = (no. of CL – no. of total embryos) per group/no. of CL per group × 100.

Skeletal staining and histopathology

Alcian blue and Alizarin red staining of cleared skeletal preparations was performed according to Hogan et al. (57). Skin and viscera were removed, and carcasses were fixed overnight in 95% ethanol (EtOH). Afterward, carcasses were stained overnight in Alcian blue 8GS (80 ml of 95% EtOH/20 ml of acetic acid/15 mg of Alcian blue 8GS), fixed in 95% EtOH for 2 to 5 hours, and transferred to 2% KOH for 24 hours. Skin and muscles were then removed, and carcasses were stained overnight in 1% KOH/0.015% Alizarin red S. Next, skeletons were cleared in 1% KOH/20% glycerol for ~48 hours. Skeletons were stored in 20% glycerol for analysis.

ChIP sequencing

As described previously (17, 32), ChIP sequencing experiments were performed using an antibody against H3K4me2 (Millipore 07-030). ChIP-seq libraries were prepared using the Illumina ChIP-seq DNA Sample Prep Kit (catalog no. IP-102-1001) and sequenced on Illumina GA II. Processing and alignment of the ChIP-seq data were performed as described previously (17) and were based on mouse mm9 assembly (July 2007 Build 37 assembly by NCBI and Mouse Genome Sequencing Consortium).

Classification of H3K4me2 signal around TSS in KDM1A transgenic mice

Enrichment values for regions surrounding TSSs (±250 bp) were calculated in a similar way as described in (17). Enrichments refer to the ratio of H3K4me2 signal in sperm from KDM1A TG3 males over the H3K4me2 signal from sperm of C57BL/6 WT mice. Genes with ≥5 reads (log2 scale) in the regions analyzed in both replicates of WT H3K4me2 were referred to as H3K4me2-positive (fig. S6A). We used the averages of two H3K4me2 data sets as the WT H3K4me2 measurements. To identify genes with altered H3K4me2 levels in KDM1A TG3 sperm, we determined the ratio of H3K4me2 occupancy around TSSs (±250 bp) in TG3 over WT sperm (fig. S6B). Because the H3K4me2 ratios were not normally distributed for down-regulated TSS regions, we performed k-means clustering analysis coupled to heatmap visualization to set the cut-off for genes with down-regulated H3K4me2 levels (see Fig. 4, B and C, and fig. S6). To increase sensitivity, k-means clustering accounted for the occupancies of nucleosomes, H3K4me3, H3K27me3, H3.3, and H3.1/H3.2 histones in WT sperm. For clusters with more than 1000 genes, we randomly chose 1000 genes for heatmap visualization purposes. For setting the right cut-off, we arbitrarily identified the upper 10% of TSS regions with elevated H3K4me2 levels in TG3 over WT sperm. Chromatin snapshots (Fig. 4A) were generated as described by Erkek et al. (17). Chromatin images represent the read counts per base averaged over a moving 300-bp interval. Averaged read counts were normalized for total read counts across samples.

Sequenom MassARRAY methylation analysis

DNA (1 μg) from TG (n = 5 animals), nonTG (n = 5), and control C57/BL6 (n = 5) sperm was bisulfite-treated with the EZ DNA Methylation Gold Kit (Zymoresearch, D5007). We used the Sequenom EpiDesigner application to design primers for the amplification of different amplicons of selected targets. Targeted regions encompassed a 250-bp window showing a large reduction of H3K4me2 (ratio of TG versus WT control <–2.5) and high CpG density and H3.3 occupancy (17). Sequenom MassARRAY methylation analysis was then performed using the MassARRAY Compact System (Sequenom, San Diego, CA). This system is based on MS analysis for qualitative and quantitative detection of DNA methylation using homogeneous MassCLEAVE base-specific cleavage and matrix-assisted laser desorption/ionization–time-of-flight MS. Spectra were elaborated by the Epityper software v1.2.0 (Sequenom), which provides methylation values of each CpG unit, expressed as percentages. Such values result from the calculation of the ratio of mass signals between methylated and nonmethylated DNA.

RRBS analysis

Reduced representation bisulfite sequencing libraries were generated according to previously published protocols using the gel-free technique (58). Briefly, 500 ng of DNA from each of the 15 samples (TG3, n = 5; nonTG3, n = 5; and control, n = 5) was used in the RRBS experiments. Multiplexed samples were used in paired-end sequencing (HiSeq sequencer, Illumina). BSMAP version 2.6 was used to trim reads (phred quality >30 and Illumina adapters), align reads to mm10, and obtain CpG methylation calls (59).

RNA analysis from sperm

Sperm (~7 × 106) were isolated by swim-out from each TG3, nonTG3, or C57BL/6 male. To increase sperm RNA yield for analysis, sperm were pooled from four or five males to give a total of ~20 × 106 sperm per replicate (C57BL/6, one replicate, n = 5 per replicate; TG, three replicates, n = 4 per replicate; nonTG, three replicates, n = 4 per replicate). RNA was extracted from pooled sperm. Somatic cell contamination was avoided by washing with somatic cell lysis buffer (0.1% SDS and 0.5% Triton-X) (60). After two washes with PBS, spermatozoa were homogenized by vortexing in buffer RLT (Qiagen RNeasy), supplemented with β-mercaptoethanol (50 mM) and steel beads (100 mg, 0.2 mm). RNA was extracted through Qiazol (Qiagen), followed by chloroform. The aqueous layer containing RNA was processed with the RNeasy RNA extraction kit (Qiagen).

RNA analysis from two-cell embryos

Two-cell embryos were collected from multiple C57BL/6 pregnancies sired by a C57BL/6 (n = 8 animals), TG (n = 4), or nonTG (n = 5) male. The number of embryos collected per pregnancy ranged from 1 to 23, and therefore, pooling of embryos from multiple sires within each genotype was necessary. On average, 26 embryos were pooled for total RNA extraction. Plug-positive females were sacrificed and oviducts flushed with PBS + 0.01% bovine serum albumin to collect embryos. Cells were stored at –80°C until embryos from two to six breedings were collected. For each genotype analyzed, the following technical replicates were used: C57BL/6 (three replicates), TG (three), and nonTG (two). Experimental details are summarized in table S7.

Microarray data analysis

RNA from sperm and embryos was converted to double-stranded cDNA using the SensationPlus WT kit and hybridized to GeneChip Mouse Exon 2.0 ST arrays (Affymetrix). CEL files for sperm and embryo data were separately read into R (version 3.1.0), where the Bioconductor package oligo was used to calculate transcript-cluster-level expression values with the RMA (robust multi-array average) algorithm. Differential expression analysis between the experimental groups was conducted using the limma package (61). The expected FDR was estimated using the Benjamini-and-Hochberg method, and statistical significance for differential expression of sperm RNA was set to FDR < 5%, coupled with a minimal difference of 1 on the log2 scale (|fold change| > 2). Statistical significance for differential expression of embryo RNA was set to FDR < 30%, due to n = 2 replicates in nonTG-sired embryos. Genes reaching statistical significance were submitted to DAVID (62) and Mouse Genome Informatics ( for functional analysis.

Functional analysis of ChIP-seq data

Functional analysis was performed using the R package topGO (63). Genes depleted in H3K4me2, which were also enriched in high levels of H3K4me3 or in H3K27me3, were selected (see table S3). Enrichment analysis was performed using Fisher’s exact test, and significance was called at P < 0.01.

Statistical analyses

The level of significance for all statistical tests used was set at P < 0.05, and all tests were two-tailed. The assessment of abnormalities in TG- and nonTG-sired pups and E18.5 embryos was conducted on a per-generation basis. Litter size was depicted as mean ± SEM and analyzed with a Student’s t test. The survivability of pups was estimated by applying a Kaplan-Meier analysis, followed by a log-rank test using GraphPad Prism 5 software (GraphPad Software, La Jolla, CA). Total abnormalities and frequency of live pups, total pregnancy loss, and total abnormal E18.5 embryos were tested for a significant difference from control offspring, using Fisher’s exact test, uncorrected for multiple comparisons. DNA methylation data were analyzed using an unpaired Student’s t test, and all data were analyzed with the aid of Systat 13 and the R statistical computing environment.

Supplementary Materials

Figs. S1 to S9

Tables S1 to S7

Reference (65)

References and Notes

  1. Acknowledgments: We thank M. Stadler, H. Royo, and D. Gaidatzis (FMI) for advice on computational data analysis; B. K. Hall (Dalhousie University) and J. Tanny and D. Bernard (McGill University) for advice; X. Giner for technical expertise; the Laboratoire de Transgénèse Center de Recherché, Centre Hospitalier de l’Université de Montréal, Canada, for microinjection and generation of the transgenic hKDM1A founders; Y. Shi (Harvard University) for providing full-length human KDM1 cDNA; T. Pastinen, G. Bourque, M. Caron, and platform personnel of the McGill Epigenomics Mapping and Data Coordinating Centers in the McGill University, as well as Génome Québec Innovation Centre, for setting up the RRBS sequencing data analysis pipeline; and F. Lefebvre for support with microarray analysis. This research was funded by the Canadian Institute of Health Research (S.K. and J.T.), Genome Quebec (S.K. and J.T.), the Reseau de Reproduction Quebecois, Fonds de Recherche du Québec – Nature et Technologies (FRQNT) (S.K. and J.T), Boehringer Ingelheim Fond (S.E.), Swiss National Science Foundation (grant 31003A_125386) (A.H.F.M.P.), and the Novartis Research Foundation (A.H.F.M.P.). The data reported in this paper are available at Gene Expression Omnibus: Microarray data, including both the CEL files and the transcript-level expression values, are under accession number GSE66052; ChIP-seq and nucleosome data are under accession number GSE55471. K.S., S.E., M.G., R.L., S.M., J.T., A.H.F.M.P., and S.K. conceived and designed the experiments. K.S., S.E., M.G., R.L., C.L., T.C., S.M., J.X., J.T., M.S., A.H.F.M.P., and S.K. performed the experiments and analyzed the data. M.H. provided advice on the data analysis and the manuscript. K.S., J.T., A.H.F.M.P., and S.K. wrote the manuscript. We declare no conflicts of interest that would prejudice the impartiality of this work.
View Abstract

Navigate This Article