Establishment of paternal methylation imprint at the H19/Igf2 imprinting control region

The insulator model explains the workings of the H19 and Igf2 imprinted domain in the soma, where insulation of the Igf2 promoter from its enhancers occurs by CTCF in the maternally inherited unmethylated chromosome but not the paternally inherited methylated allele. The molecular mechanism that targets paternal methylation imprint establishment to the imprinting control region (ICR) in the male germline is unknown. We tested the function of prospermatogonia-specific broad low-level transcription in this process using mouse genetics. Paternal imprint establishment was abnormal when transcription was stopped at the entry point to the ICR. The germline epimutation persisted into the paternal allele of the soma, resulting in reduced Igf2 in fetal organs and reduced fetal growth, consistent with the insulator model and insulin-like growth factor 2 (IGF2)’s role as fetal growth factor. These results collectively support the role of broad low-level transcription through the H19/Igf2 ICR in the establishment of its paternal methylation imprint in the male germ line, with implications for Silver-Russell syndrome.


INTRODUCTION
Imprinted genes often cluster in domains and have a special expression pattern of being transcribed from one parental chromosome only, the one inherited from the father, or the one inherited from the mother.The two chromosomes carry different epigenetic markings in the form of DNA methylation at the imprinting control region (ICR) that regulates parental allele-specific expression of genes along the domain.Those imprinted germline differentially methylated regions (gDMRs) originate in the male or female germ lines and are carried into the zygote from the sperm or the oocyte.DMRs are later maintained in the paternally (PAT) or maternally (MAT) inherited chromosome during development of the soma.One outstanding question is why PAT and MAT gDMR establishment differ in the male and female germ lines.
Mouse genetic studies found that the reciprocal maternal and paternal allele-specific expression of the H19 and Igf2 genes depends on an ICR, which behaves as a maternal allele-specific CTCF-mediated insulator in somatic cells (Fig. 1A).DNA methylation inherited from the sperm inhibits insulation in the paternal allele.A major mystery in the imprinting field is what specifies DNA methylation at this PAT DMR in the male germ line.This question is relevant to Silver-Russell syndrome (SRS; OMIM 180860) (1), an imprinting disorder, manifesting in fetal and postnatal growth retardation and related symptoms.Loss of methylation is detected in the ICR1 that controls the H19/IGF2 domain in about 65% of SRS cases (2,3).The cause of this epigenetic deficiency, however, is unknown.It may involve the process of imprint establishment-de novo methylation of ICR1 in the male germ line.
Our genome-wide deep total RNA-sequencing (RNA-seq) experiments (4) revealed that broad low-level transcription runs through the H19/Igf2 ICR specifically in 15.5 days postcoitum (dpc) mouse prospermatogonia (fig.S1), at the time when the establishment of PAT gDMRs, such as the H19/Igf2 ICR, is initiated (4)(5)(6).We hypothesized that this transcription through the ICR is required for targeting de novo methylation to the ICR in prospermatogonia (Fig. 1, A and B) and tested its role using mouse genetics.

Insertion of an RNA terminator cassette by gene targeting truncates the prospermatogonia-specific transcript
We carried out a gene targeting experiment in mouse embryonic stem (ES) cells.We inserted a transcription terminator cassette (Ter) (7) to truncate the fetal male germ cell-specific broad lowlevel transcript (4) before entering the H19/Igf2 ICR sequences (fig.S2) (8,9).
To test whether the insertion results in truncation of the broad low-level transcription, we purified prospermatogonia from 15.5 dpc H19 Ter /H19 Ter and H19 Ter+neo /H19 Ter+neo homozygous fetuses.All these fetuses also carried an Oct4-enhanced green fluorescent protein (EGFP) transgene, crossed in from the TgOG2 transgenic mouse (10) to facilitate fluorescence-activated cell sorting (FACS) of EGFP + germ cells out of fetal testes.We carried out a deep total RNA-seq experiment as earlier (4).We found that the broad low-level transcript stopped inside the terminator cassette before entering the ICR in H19 Ter /H19 Ter prospermatogonia or before entering the Pgkneo cassette in H19 Ter+neo /H19 Ter+neo prospermatogonia (fig.S3).We also noted that the H19 Ter+neo / H19 Ter+neo prospermatogonia exhibited no transcription across the Pgkneo cassette in the forward direction (fig.S3).No transcript exited into neighboring sequences out of the cassettes in either forward or reverse direction (fig.S3).In response to the terminator insertion, transcription was diminished across the ICR in the prospermatogonia of H19 Ter+neo /H19 Ter+neo and H19 Ter /H19 Ter fetuses (Fig. 1C).

Broad low-level transcription across the ICR is required for paternal methylation imprint establishment
We performed bisulfite DNA sequencing in purified prospermatogonia from individual 17.5 dpc heterozygous H19 + /H19 Ter+neo and H19 + /H19 Ter fetuses (Fig. 2A).Note that the allele inherited from the mother is written first in each genotype.In these samples, DNA methylation of the wild-type (WT) allele has reached a high level of DNA methylation by 17.5 dpc as expected (11).However, CpG methylation of the mutant allele was lower than the WT allele in prospermatogonia of each H19 + /H19 Ter+neo fetus (Fig. 2, B and  C).CpG methylation was variable in the H19 Ter+neo allele between fetuses and even more so in the H19 Ter allele, which was hypomethylated in two of seven fetuses (Fig. 2, D and E).In H19 Ter /H19 Ter prospermatogonia, CpG methylation of the ICR was 64 and 13% (Fig. 2F), again a reduction compared to the  expected levels (11).In the RNA-seq data, the H19 Ter+neo / H19 Ter+neo mice showed some low-level transcription through the ICR (Fig. 1C), which appears to be a reinitiation after a complete stop has occurred in the terminator cassette (fig.S3B).According to the model (Fig. 1B), one would predict that the H19 Ter+neo allele should have more DNA methylation compared to that of H19 Ter allele, where a complete lack of transcription was observed (Fig. 1C).The opposite result was found (Fig. 2, C and E), perhaps because the Pgkneo cassette has an effect on ICR methylation in the H19 Ter+neo allele.Nevertheless, these results collectively demonstrate that DNA methylation establishment at the H19/Igf2 ICR does not occur properly in the absence of the prospermatogoniaspecific broad, low-level transcription across the ICR sequences.

Aberrant methylation imprint establishment results in reduced growth in the offspring
On the basis of the insulator model (Fig. 1A) (12)(13)(14)(15), we hypothesized that inheriting a reduced ICR methylation from the male germline in the paternal allele results in reduced growth in the mutant fetuses.In this scenario, the Igf2 gene, encoding a fetal growth factor, is expected to be insulated in both parental alleles.The weight of H19 + /H19 Ter+neo and H19 + /H19 Ter fetuses and their placentas was variably reduced at 18.5 dpc (Fig. 3, A to F).The weight reduction of H19 + /H19 Ter+neo fetuses and placentas, and H19 + /H19 Ter fetuses was statistically significant.To find out whether the reduced growth lasts beyond fetal life, we measured the weight of weanling pups and adults (fig.S4).The weight of the mutant H19 + /H19 Ter+neo pups, males and females, was significantly lower compared with WT siblings.The number of H19 + / H19 Ter+neo mutant pups was much less than expected at weaning.This bias likely resulted from early postnatal lethality, because of the 17 dead pups found during the first 3 days of birth, 12 were H19 + /H19 Ter+neo and 5 were WT.The H19 + /H19 Ter+neo but not H19 + /H19 Ter weanling pups and adults showed a broader distribution of weights than their WT siblings at 3 and 8 weeks.The weight of H19 + /H19 Ter+neo adults was also reduced compared to WT littermates, and, in males but not females, the difference was significant.We noticed that the WT siblings of H19 + /H19 Ter mice, both females (P = 8.2959 × 10 −7 ) and males (P = 2.1692 × 10 −5 ), were significantly smaller at weaning and at 8 weeks of age (P = 0.0010 and P = 0.0576, respectively) than the WT siblings of H19 + /H19 Ter+neo mice.The reason for this difference is not known.It did not initiate at fetal stage, as those WT fetuses and their placentas were not significantly different between crosses (P = 0.8040 and P = 0.2896).The H19 + /H19 Ter mice and H19 + /H19 Ter+neo mutant mice were not different from each other.These results collectively suggest that defective methylation establishment in prospermatogonia has a lasting phenotypic effect in the next generation, as paternal inheritance of the Ter insertion results in reduced fetal and placental growth in both mutant lines and reduced postnatal weight in the H19 + /H19 Ter+neo and perhaps also the H19 + /H19 Ter mutant mice.

The variable methylation persists from epimutant prospermatogonia to the offspring
To test whether the reduced DNA methylation establishment in prospermatogonia persists into the PAT inherited allele of the soma in the next generation, we carried out multiplex indexed bisulfite sequencing in the kidney of H19 + /H19 Ter+neo and H19 + / H19 Ter fetuses at 18.5 dpc.Upon paternal transmission, both the H19 Ter+neo and H19 Ter alleles were variably hypomethylated compared to the WT allele in control fetuses (Fig. 4, A to C). Lighter fetuses displayed lower level of DNA methylation at the ICR (Fig. 4, B and C), suggesting that the variable ICR hypomethylation inherited from the mutant male germline affects growth in the mutant fetuses, and the severity of growth defect corresponds to the severity of methylation defect.These results demonstrate that the variably hypomethylated state of the prospermatogonia is passed on to the offspring in the paternal allele resulting in variable growth phenotype.It is also interesting that prospermatogonia with the lower ICR DNA methylation was obtained from smaller fetuses (Fig. 2, B, D, and F), suggesting heritability of the variable methylation between generations.

Aberrant imprinted gene expression is detected in the offspring out of epimutant prospermatogonia
The insulator model (Fig. 1A) predicts that hypomethylation of the ICR in the mutant paternal allele results in biallelic insulation of the Igf2 promoters from the enhancers ( 16) that lie downstream of H19, as observed in genetic SRS model mice (17)(18)(19).Igf2 levels were greatly and significantly reduced in the kidney, liver, yolk sac, and placenta organs of 18.5 dpc H19 + /H19 Ter+neo fetuses and in the liver and yolk sac of H19 + /H19 Ter fetuses compared with WT fetuses (Fig. 4, D and E).Because spreading of the ICR hypermethylation is required for silencing the H19 promoter during embryo development (11,20), we expected an up-regulation of H19 RNA in the mutant fetuses that inherit hypomethylated ICR from their father (Fig. 1A).H19 levels were variably increased in the liver and yolk sac and placenta of 18.5 dpc H19 + /H19 Ter+neo and H19 + /H19 Ter fetuses, and these differences were statistically significant in the liver of H19 + /H19 Ter+neo fetuses (Fig. 4, F and G).Derepression of H19 in the paternal allele should result in a shift from maternal allele-specific to biallelic expression of H19.To quantify parental alleles, we performed Sequenom allelotyping experiments as earlier (17,18).The ratio of maternal/(maternal + paternal) allele of H19 was reduced in the liver, yolk sac, and placenta and slightly in the kidney of 18.5 dpc H19 + /H19 Ter+neo fetuses compared to control H19 + /H19 + fetuses (Fig. 4H).Relaxation of H19 imprinting was also detected in the liver of H19 + /H19 Ter fetuses compared with control H19 + /H19 + fetuses (Fig. 4I).Parental allele-specific expression of imprinted genes was not affected beyond the H19/Igf2 domain along distal chr7 (figs.S5 and S6).These results demonstrate that compromised methylation imprint establishment in prospermatogonia has severe consequences to the expression of both H19 and Igf2 imprinted genes upon paternal inheritance.

DISCUSSION
This work supports a role of transcription through the ICR in the establishment of de novo DNA methylation at the H19-Igf2 gDMR in prospermatogonia: (i) The RNA-terminator cassette prevents the broad low-level transcript from entering the ICR; (ii) DNA methylation establishment is reduced at the ICR in fetal male germ cells; (iii) DNA hypomethylation is inherited to the offspring in the paternal allele; (iv) ICR hypomethylation in the soma results in biallelic insulation and causes reduced Igf2 and biallelic H19 transcription; and (v) reduced Igf2 levels result in fetal growth retardation.
The stochastic epigenetic effects in this mouse model resemble SRS cases with epigenetic molecular features.We found that DNA methylation of the ICR was partially reduced in both mutant lines.Although the terminator cassette stopped the broad low-level transcript before entering the ICR or the Pgkneo cassette (fig.S3), it did not prevent the restarting of the transcript in the ICR (albeit at a much lower level), which, if it occurs in one germ cell but not in another, can underlie the stochastic nature of the effect on DNA methylation establishment.A partial ICR hypomethylation in the soma allows partial gain of insulation and partial loss of Igf2 transcription with partially reduced fetus and placenta weight.DNA methylation was established at a variable degree in the prospermatogonia of different mutant fetuses and resulted in variable size offspring out of mutant germ cells.Smaller fetuses had more profound hypomethylation at the ICR in their perinatal prospermatogonia.This model provides a useful experimental system for studying the heritability of epigenetic states between generations.
It is interesting to compare the mechanisms of imprinted gDMR establishment between the female and male germ lines.The de novo methyltransferase DNMT3A, its cofactor DNMT3L (21-26), and  the H3K4 demethylase KDM1B (27,28) are required for establishing maternal gDMRs in growing oocytes.In addition, transcription initiated from oocyte-specific alternative promoters is required along the MAT methylated gDMRs for inducing DNA methylation in growing oocytes (25,(29)(30)(31)(32)(33)(34), in an analogous mechanism to gene body methylation (35), where the SETD2-dependent H3K36me3 is required (36) to attract the PWWP domain of DNMT3B (37,38).In fetal male germ cells, transcription that runs across paternal gDMRs is different from that found in growing oocytes.It is very broad, spans hundreds of kilobases, and occurs at such a low level in prospermatogonia that can only be detected by very deep sequencing of total RNA.It is a genome-wide feature of prospermatogonia (4).In those cells, the role of SETD2 is dispensable, but H3K36me2 methyltransferase NSD1 is required for establishing DNA methylation at three paternal DMRs, the H19-Igf2, Dlk1-Gtl2, and part of the Rasgrf1 DMR (39).NSD1-mediated H3K36me2 has been shown in mouse mesenchymal stem cells and ES cells to target the PWWP domain of DNMT3A and DNA methylation to intergenic regions in the genome (40,41).When we carried out chromatin immunoprecipitation sequencing (ChIP-seq) experiments in the 15.5 dpc H19 Ter /H19 Ter mutant prospermatogonia (fig.S7), we detected no difference in H3K36me2 (or H3K36me3 and H3K27me3) levels along the H19/Igf2 gDMR or surrounding sequences.This finding suggests that H3K36me2 affects de novo DNA methylation at those sequences independent of broad low-level transcription.Future experiments will test whether the prospermatogonia-specific transcripts recruit DNMT3A and perhaps NSD1 to other paternal gDMRs.Our RNA-seq experiment detected specific transcription through each PAT methylated imprinted gDMR (4).Therefore, similar to the H19/Igf2 gDMR, paternal imprint establishment of the Dlk1-Gtl2 and perhaps also at one-half of the Rasgrf1 DMR that depends on NSD1 (39) may require transcription-through.The other part of the Rasgrf1 gDMR has a special Piwi-interacting RNA-mediated mechanism for methylation establishment (42) and is methylated by DNMT3C (43).Similarly, we predict that PAT imprinted gDMRs will attain imprinting in the male germ line only when introduced into genomic regions that exhibit broad lowlevel transcription in prospermatogonia.When the H19/Igf2 ICR was inserted upstream of the Afp promoter (44) that has no transcription through (fig.S8), it failed DNA methylation establishment in the germline, although it still became methylated in the embryo.If the transgenic constructs also carry the elements necessary for methylation maintenance, such as the ZFP57 recognition sites (45)(46)(47), then those transgenes will behave imprinted in the offspring.When the human H19/IGF2 ICR1 was knocked into the mouse locus, it showed incomplete (~50%) methylation in the sperm and a further reduced DNA methylation by the two-cell stage and became unmethylated in the paternal allele of the offspring (48).Therefore, the human ICR1 sequences are not capable of imprint maintenance in the mouse.It will be interesting to examine whether transcription runs through the human ICR1 and whether incomplete DNA methylation establishment occurs at those sequences in mouse prospermatogonia.
The limitation of this work is that we cannot fully address the heritability of the stochastic methylation aberrations.We cannot trace germ cell methylation pattern directly into the soma in a specific chromosome because if we examine one physical chromosome copy in the germ line, then it will not give rise to offspring.Another limitation is that, although using the rabbit β-globin RNA truncation cassette (7) has been a well-accepted genetic test for the function of transcription in DNA methylation establishment at MAT imprinted gDMRs (29,31,32,34), we cannot completely exclude the possibility that altered methylation at the H19/Igf2 paternal gDMR is a consequence of the insertional mutation by an unknown mechanism and not a consequence of the reduced read through transcription by specifically terminating the transcript.
In summary, we have uncovered the mechanism that instructs methylation imprint establishment at the specific sequences of the H19/Igf2 ICR at the endogenous location in fetal male germ cells, and by manipulating this mechanism, we generated an epigenetic SRS model resembling the majority of human SRS cases (2), in that it has partial stochastic ICR hypomethylation.By providing fundamental knowledge on imprint establishment at the gDMR and developing a mouse epigenetic model, this work supports future studies aimed at ameliorating microsomia in the epigenetic subset of SRS.This SRS model has no genetic mutation inside the ICR and displays milder phenotypes than our previous genetic SRS models, where the ICR was replaced by the chicken β-globin insulator (17,18).Errors in establishment rather than methylation maintenance may be more relevant in the etiology of ICR1 hypomethylation in patients with SRS, as no evidence was found for ZFP57 mutations in SRS epimutations (49).The methylation deficit at the ICR is consistent with a partial reduction of the Igf2 transcript and insulin-like growth factor 2 (IGF2) protein levels in SRS fetuses, as well as the fetal growth retardation and microsomia symptoms.This SRS model will allow epigenetic, genetic, or pharmacological corrections of growth retardation in SRS of epigenetic origin.In these cases, therapeutically correcting IGF2 signaling at the fetal stage could be beneficial (50).

MATERIALS AND METHODS
All animal experiments were performed according to the National Institutes of Health Guide for the Care and Use of Laboratory animals, with Institutional Care and Use Committee-approved protocols at Van Andel Institute (VAI).Female and male uterus-mate fetuses from multiple litters were used as biological replicates in the experiments that compare RNA levels, allelic expression, DNA methylation, and weight between the genotypes.Fetus, newborn, and weanling numbers are provided in the figures.

Mouse lines
The H19 Ter+neo and H19 Ter mouse models were generated using gene targeting in 129SI ES cells as we did earlier (11,18).After blastocyst injection of two positive ES cell clones, we obtained 13 chimeras (four females and nine males) from one ES cell line.To obtain the H19 Ter+neo founders and rescue the anticipated PAT inherited growth phenotype similar to the genetic rescue of another SRS model (50), three male chimeras were crossed to females in the 129S1 background that carried point mutations in the four CTCF binding sites in the H19/Igf2 ICR (11).The mutation was subsequently maintained through the female germ line to avoid the disadvantage of small paternal mutants in the breeding litters; the breeding pairs were set up by crossing mutant females with WT 129S1 males.This breeding practice provided the experimental heterozygous fathers.To obtain the H19 Ter line, the Pgkneo selection cassette was removed by crossing the H19 Ter+neo mutant mice with the Hprt-Cre expressing 129S1/Sv-Hprt tm1(CAGcre)Mann/J transgenic mouse line (51).The H19 Ter line mutation was also maintained through the female germ line.The JF1/MsJ inbred strain from The Jackson Laboratory (JAX 003720) was used for distinguishing parental alleles due to the presence of single-nucleotide polymorphisms (SNPs) in its genome compared with our mouse model.For germ cell collection, the H19 Ter+neo and H19 Ter lines were crossed with the TgOG2 transgenic mouse B6;CBA-Tg (Pou5f1-EGFP)2Mnn that allows purification of prospermatogonia by FACS (10).

Southern blot hybridization
Ten micrograms of genomic DNA from the polymerase chain reaction (PCR)-positive ES clones were digested with restriction enzyme, then separated by 0.8% agarose gel electrophoresis, and transferred to Amersham Hybond nylon membrane (RPN1210B, GE Healthcare), and then Southern blot hybridization experiments were performed using dioxigenin-labeled probes.Visualization was done by exposing to x-ray film for 30 min.PCR primers for generating these probes were as follows: probe 1 (outside short arm): 5 0 GCA ATC CGT TTT AGG ACT GCG ATG TAC GAG AC 3 0 and 5 0 GCT ACA TTC ACA CGA GCA TCC AGG AGG C 3 0 to detect an 8.7-kb WT and 11.5-kb mutant Bam HI band; probe 2 (neo cassette): 5 0 CGG CAG GAG CAA GGT GAG ATG AC 3 0 and 5 0 CGC TTG GGT GGA GAG GCT ATT CG 3 0 ; probe 3 (inside long arm): 5 0 CAG AGA GCA GCA GAG AAG TGT TAG CTC TTT GGG 3 0 and 5 0 GTA AGT GTC TGT CCC GCT CGT GGT CA 3 0 to detect a 4-kb WT and 7-kb mutant Hind III band.

Purification of fetal male germ cells
To obtain homozygous mutant male fetal germ cells, H19 Ter+neo / H19 Ter+neo or H19 Ter /H19 Ter females that carry the Pou5f1 promoter-driven EGFP transgene from the TgOG2 (10) mouse line were crossed with males of the same genotype.To obtain heterozygous male fetal germ cells, TgOG2 (10) females were crossed with H19 Ter+neo or H19 Ter males.The fetal testicles were dissected, and a single-cell suspension was prepared by trypsin digestion and trituration as described earlier (10).Up to eight testes were digested in 150 μl of 0.25% trypsin (including 0.5% bovine serum albumin) at 36°C for 15 min, tapping gently every 5 min.The trypsin reaction was stopped by adding 450 μl of 20% fetal bovine serum in M2 medium (catalog no.M7167, Sigma-Aldrich).The EGFP + germ cells were purified by flow cytometry using a Beckman Coulter MoFlo Astrios sorter.

Construction and sequencing of directional total RNA-seq libraries
Total RNA was isolated using RNA-Bee reagent (Tel-Test) from FACS sorted 15.5 dpc male germ cells, followed by deoxyribonuclease I treatment.Libraries were prepared by the VAI Genomics Core from 500 ng of total RNA using the KAPA RNA HyperPrep Kit (Kapa Biosystems, Wilmington, MA, USA).Ribosomal RNA material was reduced using the QIAseq FastSelect -rRNA HMR Kit (QIAGEN, Germantown, MD, USA).RNA was sheared to 300 to 400 bp.Before PCR amplification, cDNA fragments were ligated to IDT for Illumina TruSeq UD indexed adapters (Illumina Inc., San Diego, CA) and amplified with 8 cycles of PCR.Quality and quantity of the finished libraries were assessed using a combination of Agilent DNA High Sensitivity chip (Agilent Technologies Inc.) and QuantiFluor dsDNA System (Promega Corp., Madison, WI, USA).Individually indexed libraries were pooled, and 50-bp, paired-end sequencing was performed on an Illumina NovaSeq 6000 sequencer to an average depth of 100 million raw paired reads per transcriptome.Base calling was done by Illumina RTA3, and output of noncrystallographic symmetry was demultiplexed and converted to FastQ format with Illumina Bcl2fastq v1.9.0.

Quantitative reverse transcription PCR
Total RNA was isolated using RNA-Bee reagent (Tel-Test).Contaminating DNA was removed with the DNA-free DNA Removal Kit (Ambion).cDNA was prepared from 400 ng of total RNA using the SuperScript III Random Primer Synthesis kit (Invitrogen) for the quantitative PCR assays as we did earlier (17).The primer sequences are listed in table S1.

Sequenom allelotyping
To measure the portion of each parental allele in the total transcript levels, multiplex Sequenom (now Agena Bioscience) allelotyping assays (59) were used.These assays are based on SNPs that distinguish between the inbred JF1/MsJ (JF1) and 129S1 (129) mouse strains.Each "unextended" primer (UEP) abuts a SNP in a target transcript, and the incorporating nucleotide differs in molecular mass between the parental alleles.The abundance of the extended UEP is quantified by mass spectrometry.Amplified cDNA samples were spotted onto a 384 SpectroCHIP Array (Agena Bioscience).Automated spectra acquisition was performed in a MassArrayCompact mass spectrometer (Sequenom) using the Spectroacquire program (Sequenom) and was analyzed by MassArray Typer v3.4.RNA-mixing standards were routinely run to verify linear response in measured versus input allele-specific transcription: For example, total RNA from JF1 and 129 embryos was mixed indifferent percent ratios (0:100, 10:90, 30:70, 50:50,70:30, 90:10, and 100:0) before cDNA preparation and Sequenom allelotyping.A 50:50 RNA mix was used for RNA skew correction.The percentage of transcription of each allele in the total expression was calculated at each given SNP.DNA samples were used from each embryo to verify the heterozygous genotype at each SNP.The UEP and PCR primer sequences are listed in table S2.

Bisulfite sequencing of germ cells
A total of 2000 FACS-sorted male fetal germ cells were lysed with proteinase K solution and converted with EZ DNA methylation kit (D5001, ZYMO) according to the manufacturer's instructions.PCR was performed using the genotype-specific forward primers and the shared reverse primer (table S3) and Zymo Taq hot start polymerase (catalog no.E2002, ZYMO) to amplify the converted DNA.The purified PCR fragment was ligated into pGEM-T vector (catalog no.A3600, Promega).Sanger sequencing of 15 independent colonies per region was performed by GENEWIZ.The QUMA online software was used for methylation analysis.Colonies that had identical pattern of the occasional incomplete conversion were excluded from the displays.

Multiplex bisulfite sequencing
Fetal kidney DNA was purified by Proteinase K digestion, phenolchloroform extraction, and ethanol precipitation.The EZ DNA methylation kit (D5001, ZYMO) was used for the bisulfite conversion according to the instructions of the kit.Multiplex PCR was performed with indexed primers (table S4) and Zymo Taq hot start polymerase (E2002, ZYMO) to amplify the converted DNA.The amplified DNA fragments were resolved using a 1% agarose gel, and the DNA was isolated from the gel slices using Monarch DNA Gel Extraction Kit (catalog no.T1020S).Fifty nanograms of each fragment was amplified with different index primers as listed below for Amplicon-EZ sequencing by GENEWIZ.

Bioinformatics analysis of multiplex bisulfite amplicon sequencing
Reads were trimmed using TrimGalore v0.6.0 (https://github.com/FelixKrueger/TrimGalore) (53) with default settings to remove lowquality bases and Illumina adapters.Demultiplexing was done using CutAdapt v3.2 (53).The primer sequences, including the barcode sequences, were provided as anchored 5 0 adapters using the "-g" and "-G" parameters, requiring no mismatches and no indels ("-e 0 --no-indels").To account for read pairs where the barcode adapters were on the opposite read, the unmatched reads from the first Cu-tAdapt run were processed with CutAdapt again using the same settings, except R1 and R2 were specified in reverse (R1 as R2 and vice versa).Last, the R1 and R2 reads from the two rounds of demultiplexing were concatenated together for each sample.Reads were aligned to the corresponding target sequence, padded with five N's on each end to avoid issues during methylation context inference by the aligner, using Bismark (60) v0.23.0 with the parameters, "--non_directional" and "--local."Per-CpG methylation percentages were obtained using the "bismark_methylation_extractor" script in Bismark with the parameters, "--bedGraph" and "--paired-end."

Supplementary Materials
This PDF file includes: Figs.S1 to S8 Tables S1 to S4

Fig. 1 .
Fig. 1.Testing the function of broad low-level transcription at the H19/Igf2 ICR.(A) Insulator model.CTCF molecules (yellow ovals) bind in the unmethylated ICR allele in the chromosome inherited from mother and block Igf2 from the shared enhancers (green dot).Methylation of the ICR inherited from the father via the sperm prevents CTCF binding and thus allows activation of Igf2.The panel to the right shows the expected aberrant gene expression pattern when the ICR is not methylated properly in the paternal chromosome, as in the case of the majority of SRS patients.(B) Hypothesis: Paternal methylation imprint is established in prospermatogonia by de novo methyltransferase DNMT3A in response to broad low-level transcription that runs through the ICR.The right panel depicts the expected outcome when transcription is terminated before entering the ICR in prospermatogonia.(C) Transcription across the ICR is terminated by the knock-in terminator cassette.Genome browser view of normalized total RNA sequencing (RNA-seq) bigwig files in the forward (top six lanes) and reverse (bottom six lanes) directions.Biological replicates of 15.5 dpc purified prospermatogonia from wild-type (WT), H19 Ter+neo /H19 Ter+neo , and H19 Ter /H19 Ter homozygous fetuses.The genes and the H19/Igf2 ICR and the insertion site for the termination cassette are marked.

Fig. 2 .
Fig. 2. De novo DNA methylation of the H19/Igf2 ICR requires transcription through the ICR in prospermatogonia.(A) Schematic map of the bisulfite sequencing regions in the WT and mutant alleles.(B to F) Results of bisulfite sequencing in EGFP + FACS-purified prospermatogonia from 17.5 dpc fetuses.Open and closed circles indicate unmethylated and methylated CpGs, respectively.(B) Examples of H19 + /H19 Ter+neo fetuses.Dashed vertical lines separate the ICR sequences from the flanking regions.The weight of each fetus is indicated to the left.(C) Individual H19 + /H19 Ter+neo fetuses at 17.5 dpc obtained from three litters of three mutant males out of three independent founders.The percentage of methylated CpGs of 13 CpGs in the ICR fragment is plotted in the WT and knock-in allele.(D) Examples of H19 + /H19 Ter fetuses .(E) Individual H19 + /H19 Ter fetuses obtained from three litters of two males out of two founder males.(F) Examples of DNA methylation in prospermatogonia of the H19 Ter / H19 Ter fetuses.

Fig. 4 .
Fig. 4. ICR hypomethylation is transmitted from the germline to the offspring and results in aberrant expression of the Igf2 and H19 imprinted genes.(A to C) Multiplex bisulfite sequencing results shown in the PAT inherited allele in the kidney of 18.5 dpc fetuses (A to C). Percentage of methylation is plotted at each CpG in the ICR and flanking regions (left of the dotted line).(A) WT H19 + /H19 + (n = 8), (B) H19 + /H19 Ter+neo (n = 11), and (C) H19 + /H19 Ter (n = 11).(D to G) Quantitative reverse transcription polymerase chain reaction (PCR) in four fetal organs at 18.5 dpc.Experimental crosses were carried out as in Fig. 3.The average value for each organ is marked with a horizontal bar (P value by two-tailed t tests).(D) Igf2 in H19 + /H19 Ter+neo fetuses.(E) Igf2 in H19 + /H19 Ter fetuses.(F) H19 in H19 + /H19 Ter+neo fetuses.(G) H19 in H19 + /H19 Ter fetuses.(H and I) Sequenom allelotyping of allele-specific transcription in four organs of mutant and control 18.5 dpc fetuses.The ratio of maternal/(maternal + paternal) allele for the H19 and Igf2 imprinted genes is displayed by the color scale.Genotypes are shown at the top.(H) H19 + /H19 Ter+neo fetuses.(I) H19 + /H19 Ter fetuses.