Recapitulation of gametic DNA methylation and its post-fertilization maintenance with reassembled DNA elements at the mouse Igf2/H19 locus

Background Paternal allele-specific DNA methylation of the H19 imprinting control region (ICR) regulates imprinted expression of the Igf2/H19 genes. The molecular mechanism by which differential methylation of the H19 ICR is established during gametogenesis and maintained after fertilization, however, is not fully understood. We previously showed that a 2.9-kb H19 ICR fragment in transgenic mice was differentially methylated only after fertilization, demonstrating that two separable events, gametic and post-fertilization methylation, occur at the H19 ICR. We then determined that CTCF/Sox-Oct motifs and the 478-bp sequence of the H19 ICR are essential for maintaining its maternal hypomethylation status and for acquisition of paternal methylation, respectively, during the post-fertilization period. Results Using a series of 5′-truncated H19 ICR transgenes to dissect the 478-bp sequence, we identified a 118-bp region required for post-fertilization methylation activity. Deletion of the sequence from the paternal endogenous H19 ICR caused loss of methylation after fertilization, indicating that methylation activity of the sequence is required to protect endogenous H19 ICR from genome-wide reprogramming. We then reconstructed a synthetic DNA fragment in which the CTCF binding sites, Sox-Oct motifs, as well as the 118-bp sequence, were inserted into lambda DNA, and used it to replace the endogenous H19 ICR. The fragment was methylated during spermatogenesis; moreover, its allele-specific methylation status was faithfully maintained after fertilization, and imprinted expression of the both Igf2 and H19 genes was recapitulated. Conclusions Our results identified a 118-bp region within the H19 ICR that is required for de novo DNA methylation of the paternally inherited H19 ICR during pre-implantation period. A lambda DNA-based artificial fragment that contains the 118-bp sequence, in addition to the previously identified cis elements, could fully replace the function of the H19 ICR in the mouse genome.

The most common molecular mechanism for achieving genomic imprinting is allele-specific DNA methylation of the imprinting control regions (ICRs), which is frequently observed at imprinted gene loci. DNA methylation of the ICRs is generally acquired during either spermatogenesis or oogenesis; accordingly, ICRs are classified as germline differentially methylated regions (gDMRs). The allelic methylation pattern is maintained after fertilization, throughout the lifespan: the germline-methylated ICRs on one of the alleles are resistant to genome-wide demethylation activity, which is associated with epigenetic reprogramming during the pre-implantation period, and non-methylated ICRs on the other allele are protected from allele-nonspecific de novo methylation during cell differentiation in post-implantation embryos. In other words, differential methylation of the ICRs is regulated at three distinct stages: gametogenesis, pre-implantation, and post-implantation, ensuring monoallelic gene expression in somatic cells.
At the Igf2/H19 locus, Igf2 is expressed only from the paternal allele, whereas H19 is expressed only from the maternal allele [3,4]. The imprinted expression of both genes is governed by the concerted action of their shared enhancer, located downstream of the H19 gene, and a paternally methylated gDMR called the H19 ICR. On the maternal allele, the unmethylated H19 ICR recruits CCCTC binding factor (CTCF) to form an enhancerblocking insulator to interfere with distal Igf2 gene activation by the enhancer, resulting in exclusive H19 gene expression. By contrast, a hypermethylated paternal ICR silences nearby H19 gene transcription, but allows Igf2 gene expression by preventing CTCF from binding to the ICR [5][6][7][8][9]. Loss and gain of methylation at the H19 ICR has been reported in 30-60% and 5% of patients with SRS and BWS, respectively [2]; therefore, it is of considerable clinical importance to elucidate the allele-specific methylation mechanisms of the H19 ICR.
In previous work, we generated transgenic mice (TgMs) harboring either randomly integrated mouse H19 ICR fragments [10] or fragments of the H19 ICR embedded in human β-globin locus YACs (150 kb; [11]), and found that paternally inherited Tg fragments acquired DNA methylation after fertilization even though they were not methylated in sperm. In other words, our results demonstrated that two separable methylation acquisition processes occurred at the H19 ICR: one during spermatogenesis that depends on the activity of the surrounding sequence (i.e., outside the H19 ICR), and another in post-fertilization embryos that is governed by its intrinsic activity. The latter allele-specific, postfertilization de novo DNA methylation of the transgenic H19 ICR was also observed in the endogenous Igf2/H19 gene locus and was catalyzed by oocyte-derived de novo methyltransferases (Dnmt3a and Dnmt3L) [12]. We then determined that a 765-bp sequence in the 5′-portion of the H19 ICR was necessary for acquisition of methylation after fertilization: deletion of that sequence from the endogenous paternal ICR caused loss of methylation at the remaining H19 ICR in pre-implantation embryos, without changing its hypermethylation status in sperm. We concluded that paternal allele-specific de novo methylation activity maintains the imprinted methylation of the H19 ICR in pre-implantation embryos [12]. On the other hand, mutation of CTCF-binding sites [13] and Sox-Oct motifs [14] within the mouse H19 ICR caused aberrant gain of methylation on the maternal allele after implantation, indicating that these elements are required to protect the maternal, hypomethylated H19 ICR from allele-nonspecific de novo methylation.
The regulatory sequences we have identified thus far (a 478-bp segment of the 765-bp fragment mentioned above, the CTCF-binding sites and Sox-Oct motifs) in the H19 ICR are capable of transforming a normally nonimprinted λ DNA sequence into the DMR, when they are assembled together on a λ DNA fragment and assayed in TgM [15]. However, it remains to be determined whether the synthetic fragment can fully reproduce the genomic imprinting phenomena at endogenous mouse Igf2/H19 gene locus, as allele-specific methylation of the fragment was observed only during the post-fertilization period in the transgenic β-globin gene locus.
In addition, it remains unknown how paternal H19 ICR methylation is acquired through the 478-bp sequence after fertilization. The sequence could be recognized by a sequence-specific DNA-binding factor(s) in an allelespecific manner, so that the H19 ICR is distinct from the other genomic regions. Although ZFP57 maintains hypermethylation at multiple ICRs [16,17], that factor is not a plausible candidate for the regulator of H19 ICR de novo methylation for two main reasons. First, ZFP57 binds to DNA in a CpG methylation-dependent manner, whereas the transgenic H19 ICR in sperm is unmethylated. Consistent with this, in our previous work [15], we failed to demonstrate ZFP57 binding to the 478-bp sequence in gel-shift assays. In addition, the DNA methylation status of the endogenous H19 ICR is not affected in Zfp57-knockout mice [16,18]. Therefore, we assume that currently unidentified factors are responsible for allele-specific, post-fertilization methylation at the H19 ICR.
In this study, to clarify the mechanisms involved in the two separate mechanisms of gamete DNA methylation and post-fertilization methylation, we generated TgMs carrying a series of 5′-truncated H19 ICR fragments, with the aim of identifying the cis element(s) (and trans-acting factors that bind these elements, ultimately) responsible for the acquisition of post-fertilization methylation. We determined that a 118-bp sequence within the 478-bp region was essential for the activity in the TgM context. As anticipated, deletion of the sequence from the endogenous mouse H19 ICR decreased its methylation level in pre-implantation embryos, but not in sperm. The λ-based reconstituted fragment, including the 118-bp sequence, recapitulated both imprinted methylation and imprinted gene expression after fertilization in transgenic animals. Most importantly, the reconstituted fragment fully complemented the function of the endogenous H19 ICR, including acquisition of methylation in sperm.

Results
A 118-bp sequence at the 5′-segment of the H19 ICR is essential for acquisition of paternal methylation In previous work, we have narrowed down the sequence responsible for post-fertilization paternal methylation of the H19 ICR to a 478-bp region and demonstrated that it is required in vivo for normal development ( Fig. 1a; [12,15]). Furthermore, we showed that postfertilization, allele-specific methylation was recapitulated in a λ-phage-based synthetic DNA fragment in the TgM only when the 478-bp sequence was included [15]. To further define the responsible sequence in this study, we generated H19 ICR fragments with a series of 5′-deletions of the 478-bp region (~ 60 bp intervals) and inserted them into the human β-globin YAC to generate TgMs (Fig. 1b). To avoid position-of-integration site effects, which transgene fragments frequently incur, and to directly compare activity at a single genomic site, we combined two successive deletion fragments side-by-side and employed a transgene co-placement strategy [19] (Additional file 1: Fig. S1A). Four YAC constructs, each carrying a distinct set of deletions (fragments del-8/9 to 2/3), were used to generate TgMs, and at least two independent mouse lines were established for each construct. Long-range analysis of thymus genomic DNA demonstrated that all but two harbored an intact, single-copy transgene (Additional file 1: Fig. S1B) (lines 36 and 20 of the del-2/3 TgMs lacked sequence 5′ to the LCR and 3′ to the β-globin gene regions, respectively). Cross-mating of these TgMs with Cre-TgM caused in utero Cre-loxP recombination that generated daughter sublines carrying either of the H19 ICR deletion fragments, which was confirmed by Southern blot analysis of somatic cell DNA (Additional file 1: Fig. S1C and D).
Tail somatic cell DNA of animals inheriting the YAC transgenes either paternally or maternally was prepared, and the methylation status of their endogenous and transgenic H19 ICR sequences was determined by Southern blot analysis (Additional file 2: Fig. S2A). The appearance of digested and undigested endogenous fragments in equimolar ratios served as a control for complete genomic DNA digestion by methylationsensitive restriction enzymes. The results revealed that 5′-deletion fragments of the H19 ICR in animals inheriting the transgenes maternally were hypomethylated, as was the intact 2.9-kb fragment (mat. in Additional file 2: Fig. S2B-F). By contrast, while the paternally inherited transgenic H19 ICR in the del-9 and del-8 lines was hypermethylated (pat. in Additional file 2: Fig.  S2B and C), those in the del-7 lines (pat. in Additional file 2: Fig. S2D) exhibited partial methylation, and all others were hypomethylated (pat. in Additional file 2: Fig. S2E-I).
To precisely determine the sequence requirements for paternal H19 ICR methylation, we analyzed the methylation status of the Tg H19 ICR in lines del-8, -7, Fig. 1 Search for DNA sequences which are responsible for paternal methylation of the H19 ICR fragment. a Structure of the mouse endogenous Igf2/H19 locus. The expression of paternal Igf2 and maternal H19 genes depends on the shared 3′ enhancer. The H19 ICR, located approximately at − 4 to − 2 kb relative to the transcription start site of H19 gene is contained within a 2.9-kb SacI (Sa)-BamHI (B) fragment. DNA sequence (478 bp) shown to be sufficient for acquiring paternal methylation in TgM [15] is marked in gray. Dots (1-4) indicate CTCF-binding sites. G, BglII; H, HindIII sites. b Structure of the 150-kb human β-globin locus YAC. The LCR and β-like globin genes are denoted as gray and filled boxes, respectively. In our previous studies, the H19 ICR (2.9-kb) or the ICR4321S (766-bp shorter than the original 2.9-kb sequence) fragments were introduced 3′ to the LCR [11,12]. In this study, a series of 5′-truncated H19 ICR fragments (del-2-9) were inserted into the identical position of the YAC to examine their activities in TgM c Two-cell embryos that inherited the transgenes either paternally (pat.) or maternally (mat.) were embedded in agarose beads and treated with sodium bisulfite. The beads were used to amplify the region II in a by nested PCR. PCR products were individually subcloned and sequenced. The results from single beads are presented together in a cluster and -6 by bisulfite sequencing (Fig. 2a). Analysis of tail somatic cell DNA revealed that paternally inherited Tg sequences in lines del-8, del-7, and del-6 were hyper-, partially and hypo-methylated, respectively, whereas all of these sequences were hypomethylated when maternally inherited (Fig. 2b). To determine whether loss of methylation in the paternal del-7 and del-6 sequences was due to a lack of de novo DNA methylation activity in pre-implantation embryos or a lack of methylation maintenance activity after implantation, we analyzed two-cell embryos (Fig. 2c). The del-7 and del-6 sequences were partially hypomethylated, suggesting that the 118-bp sequence between the 5′ ends of the del-8 and del-6 fragments are essential for de novo methylation of the paternal H19 ICR during the preimplantation period.

Deletion of the 118-bp sequence results in loss of methylation at the H19 ICR during the pre-implantation period
To confirm our results in the 5′-deletion mutants, we next deleted the sequence from the 2.9-kb H19 ICR in YAC TgM by in vivo genome editing (Fig. 3a). To this end, we generated two pX330-based plasmids expressing the Cas9 nuclease and guide RNAs targeting the 5′ ends of the del-6 and del-8 sequences (Additional file 3: Fig.  S3A). Pronuclear co-injection of these plasmids into fertilized eggs recovered from the H19 ICR/human β-globin YAC TgM [11] led to generation of mutant transgenic loci. Two mutant lines with an identical 116-bp deletion (a bit shorter than 118 bp due to restriction by PAM motif locations) in the H19 ICR were generated (lines 226 and 247; Additional file 3: Fig. S3A). Analysis of tail somatic cell DNA by Southern blotting and bisulfite sequencing revealed that the mutant transgenic H19 ICR sequences were hypomethylated regardless of their parental origin (Additional file 3: Fig. S3B and C, and Fig. 3b). In addition, hypomethylation of the mutant paternal H19 ICR was also observed in two-cell embryos (Fig. 3b), indicating that the 116-bp sequence was required for post-fertilization methylation of the H19 ICR transgene.
When we generated the aforementioned mutation in the transgenic H19 ICR, endogenous H19 ICR was concomitantly mutagenized, and the 116-bp sequence was deleted (Fig. 4a). Bisulfite sequencing analysis revealed that the endogenous H19 ICR with the mutation was fully methylated in sperm (Fig. 4b), demonstrating that the 116-bp sequence was dispensable for its germline methylation. This result was consistent with our hypothesis that establishment of methylation at the 2.9kb H19 ICR sequence in sperm was under the control of its surrounding sequence, rather than its intrinsic activity [10]. By contrast, paternally inherited mutant H19 ICR exhibited significant loss of methylation in two-cell (Fig. 4c) and blastocyst stage embryos (Fig. 4d). This result was also consistent with our hypothesis that the post-fertilization methylation activity of the H19 ICR observed in the TgM context was responsible for protecting its germline-established paternal methylation (at the endogenous locus) against genome-wide reprogramming activity during pre-implantation period [12]. In addition, these results clearly demonstrated that the paternal H19 ICR methylation cannot be maintained solely by a ZFP57-mediated mechanism during the post-fertilization period, as the mutant sequence retained all binding sites for ZFP57 [15]. The 118-bp sequence was required for paternal methylation of the transgenic H19 ICR. a Structure of the transgenes. The 116-bp sequence, which was a part of the 118-bp region identified by 5′-truncation experiments (Fig. 2), was internally deleted from the 2.9-kb H19 ICR fragment in TgM by CRISPR/Cas9 genome editing (Tg-5′ICR-KO(116)). Regions I and III, indicated by gray bars below the map, were analyzed by bisulfite sequencing. b DNA methylation status of the mutant H19 ICR transgene in tail somatic cells (upper panel) or 2-cell embryos (lower) of TgM, that inherited the transgenes either paternally (pat.) or maternally (mat.), was analyzed by bisulfite sequencing

The reconstituted synthetic fragment recapitulates genomic imprinting in YAC-TgM
We previously showed that artificial DMR activity can be generated by assembling the sequences required for protecting the paternal H19 ICR against genome-wide demethylation (by simultaneous de novo DNA methylation) during the pre-implantation period (i.e., the 478-bp sequence), as well as those required for protecting the maternal, unmethylated H19 ICR from postimplantation de novo methylation (i.e, the CTCF and Sox-Oct motifs) on the λ DNA sequence [15]. To determine whether the shorter 118-bp sequence was sufficient to confer the same activity, we combined the LCb fragment (λ DNA fragment harboring the CTCF and Sox-Oct motifs; [14,15]) and the LCb fragment with the 118-bp sequence attached (termed the LCb118), and inserted them into human β-globin YAC, employing the transgene co-placement strategy to precisely compare their activities (Additional file 4: Fig. S4A). Following establishment of two intact, single-copy YAC TgM lines, confirmed by long-range Southern blot analysis of thymic DNAs (lines 28 and 890; Additional file 4: Fig. S4B), the mice were crossed with Cre-TgM to induce in utero Cre-loxP recombination. Tail DNAs from the offspring confirmed that Tg sublines harboring either LCb or LCb118 sequences were successfully The 118-bp sequence was also required for paternal methylation of the endogenous H19 ICR. a Map of wild-type and knockout alleles. The same 116-bp sequence that was deleted from the transgenic H19 ICR fragment in Fig. 3, was removed from the mouse endogenous locus by CRISPR/Cas9 genome editing (endo-5′ICR-KO(116)). b-d DNA methylation status of the mutant H19 ICR in sperm (b), 2-cell embryos (c), or blastocysts (d) of 116-bp KO mice, that inherited the KO allele either paternally (pat.) or maternally (mat.), was analyzed by bisulfite sequencing obtained from both parental lines (Additional file 4: Fig. S4C).
Methylation analysis of tail somatic cell DNA by Southern blotting revealed that the LCb fragments exhibited low-level methylation in more than half of the individuals inheriting the transgene paternally (lines 28 and 890; Additional file 5: Fig. S5A and B), consistent with previous data obtained at distinct integration sites of transgenes [15]. By contrast, the paternally inherited LCb118 fragments exhibited high-level methylation in all individuals analyzed (lines 28 and 890; Additional file 5: Fig. S5C and D), whereas maternally inherited fragments exhibited hypomethylation, which was also the case for the LCb (Additional file 5: Fig. S5B). Importantly, the methylation status of LCb118 transgenes was reprogrammable over generations depending on parental origin, which is an important feature of genomic imprinting (Additional file 5: Fig. S5D). Therefore, we concluded that the 118-bp sequence was sufficient for the acquisition of paternal methylation. In addition, LCb and LCb118 fragments were not methylated in the testis germ cells (Additional file 6: Fig. S6), indicating that the 118-bp sequence conferred post-fertilization acquisition of methylation.
To precisely evaluate the function of the 118-bp sequence in the context of artificially assembled LCb sequences embedded in the YAC Tg in mice (Fig. 5a), we analyzed their methylation statuses by bisulfite sequencing. The results revealed that the Tg-LCb sequence was not methylated in the paternal allele in tail somatic cells and in the sperm (Fig. 5b). By contrast, the Tg-LCb118 sequence in tail somatic cells was hypermethylated only after paternal transmission (Fig. 5c). In addition, such acquisition of allele-specific methylation was observed as early as the two-cell stage, whereas the sequence was not methylated in sperm (Fig. 5c). Thus, given that the temporal and allele specificity of methylation acquisition at the transgenic H19 ICR and transgenic LCb118 sequences were identical, we concluded that paternal allele-specific post-fertilization methylation was recapitulated by the synthetic LCb118 sequence in the TgM context.
The LCb118 sequence was placed between the LCR enhancer and the genes in the human β-globin locus YAC in TgM. Therefore, we anticipated monoallelic β-globin gene expression because of allele-specific enhancer-blocking activity due to methylation-sensitive CTCF binding to LCb118. To test the in vivo role of LCb118, we prepared chromatin from nucleated erythroid cells of adult animals inheriting the transgene either paternally or maternally; their hyper-and hypomethylation states, respectively, were confirmed by Southern blot analysis (Fig. 6a). ChIP analysis of the chromatin revealed that CTCF binding was enriched at a significantly higher level at the maternal LCb118 than at the paternal one (Fig. 6b), consistent with the respective methylation statuses of the sequences. Level of enrichment at the endogenous H19 ICR was similar regardless of whether the transgene was inherited paternally or maternally; this is as expected, as the value represents the sum of the two parental alleles. We then analyzed transgenic human β-globin gene expression in the nucleated erythroid Reconstitution of the differentially methylated region by a synthetic DNA fragment in TgM. a TgM lines were generated by using the 150-kb human β-globin locus YAC bearing either the LCb or LCb118 fragments introduced 3′ to the LCR within the YAC. Filled and gray boxes indicate the "b" region (which includes Sox-Oct motifs) and 118-bp sequence, respectively. b, c DNA methylation statuses of the LCb (b) and LCb118 (c) fragments in tail somatic cells, 2-cell embryos, and sperm of YAC-TgM, that inherited the transgenes either paternally (pat.) or maternally (mat.). Regions analyzed by bisulfite sequencing were indicated by gray bars above each map cells by RT-qPCR. The results revealed that the transgene was highly expressed only when paternally transmitted (Fig. 6c).

The reconstituted fragment is able to replace the function of the endogenous H19 ICR
We previously demonstrated that the post-fertilization de novo methylation activity of the H19 ICR protects its paternal methylation status against pre-implantation reprogramming of the whole genome. However, in the TgM context, acquisition of methylation in sperm took place neither in the wild-type nor in artificially assembled fragments. Based on our previous observations [10], we anticipated that gametic methylation of the H19 ICR was under control of the surrounding sequences somewhere within the Igf2/H19 gene locus. Therefore, we decided to test whether the LCb/LCb118 sequences could be methylated in sperm when they were inserted in place of the endogenous H19 ICR sequence, and if so, whether they could completely replace its function (Additional file 7: Fig. S7A).
To generate knock-in alleles, LCb or LCb118 targeting vectors harboring the H19 ICR flanking sequences, together with a genome editing plasmid targeting the H19 ICR region, were transfected into C57BL/6 (B6) mouse ES cells. Southern blot and sequencing analyses identified one and two ES cell clones, respectively, with their endogenous H19 ICR sequences correctly replaced with the LCb or LCb118 sequences (Additional file 7: Fig. S7B and data not shown). These ES cell clones were then used for co-culture aggregation to establish mouse lines. Correctness of mutagenesis was confirmed by Southern blot and sequencing analyses of the mouse tail tip DNA (Additional file 7: Fig. S7C and data not shown).
Next, the methylation status of the synthetic sequences knocked in at the endogenous Igf2/H19 locus was analyzed by bisulfite sequencing (Fig. 7a). As anticipated, the LCb sequence was almost fully methylated in sperm (Fig. 7b), in contrast to its unmethylated state in the transgenic environment (Fig. 5b). After fertilization, however, the methylation level of the paternal LCb sequence gradually decreased during the pre-implantation period (Fig. 7b), consistent with a lack of post-fertilization methylation activity of the LCb (Fig. 5b). The maternally inherited fragment was almost devoid of methylation at the blastocyst stage (Fig. 7b). By contrast, the LCb118 sequence, which was also methylated in the sperm, remained methylated inheriting the transgene either paternally (P) or maternally (M) were made anaemic and spleens were removed, from which one-quarter each was used for genomic DNA or total RNA preparation with the remaining half used for chromatin preparation. a DNA methylation status of the transgene was determined by Southern blot analysis using BamHI with (+) or without (−) BstUI (vertical lines) and a λ probe. *: methylated, uncut fragments in BstUI (+) lanes. b ChIP analysis of CTCF occupancy at the transgene. Chromatin was immunoprecipitated using either control IgG or anti-CTCF antibodies. Following qPCR analyses of three distinct genomic regions (Necdin; negative control, endogenous H19 ICR; positive control, and LCb118 transgene), relative enrichment values (CTCF/IgG signal ratio) were calculated. The average and standard deviation (S.D.), determined by three reactions, are depicted, as a signal for Necdin (M) was arbitrary set at 1.0. Statistical differences were determined using an unpaired t test (N.S., not significant). c The relative expression levels of the human β-globin gene, after normalization to that of the endogenous mouse α-globin gene were determined by RT-qPCR analysis. The average and standard deviation (S.D.), determined by three reactions, are depicted, as a value of No. 2640 animal was arbitrary set at 1.0 Fig. 7 Reconstitution of the differential methylation in the LCb118 sequence at the mouse endogenous Igf2/H19 locus. a Map of the wild-type and knock-in alleles. The endogenous mouse H19 ICR sequence was replaced by the LCb or LCb118 synthetic DNA fragments. Filled and gray boxes indicate the "b" region (which includes Sox-Oct motifs) and 118-bp sequence, respectively. DNA methylation status of the LCb (b) and LCb118 (c) sequence in sperm, 2-cell embryos, and blastocysts of knock-in mice, that inherited the mutant alleles either paternally (pat.) or maternally (mat.), was analyzed by bisulfite sequencing. The overall percentage of methylated CpGs is indicated next to each panel even after fertilization (Fig. 7c). This hypermethylation was also observed at the blastocyst stage, when the methylation level of the maternally inherited sequence was significantly lower (Fig. 7c). Taken together, these observations suggest that at the endogenous Igf2/H19 locus, hypermethylation of the paternal H19 ICR established in the sperm was maintained even after fertilization in a 118-bp sequencedependent manner. We then tested whether monoallelic expression (i.e., genomic imprinting) of the Igf2/H19 genes was recapitulated at the mutant locus. To discriminate allelic expression of the Igf2 and H19 genes, we took advantage of single-nucleotide polymorphisms (SNPs) between the B6 and JF1 inbred mouse strains. Male mice homozygous for the LCb118 allele, which was generated in the B6 background, were mated with wild-type JF1 female mice to derive a paternally inherited LCb118 allele in the offspring (Fig. 8a, left), whereas B6 female mice homozygous for the LCb118 allele were mated with wild-type JF1 male mice to derive a maternally inherited LCb118 allele in the offspring (Fig. 8a, right). Bisulfite sequencing analysis of fetal liver DNA (18.5 dpc) revealed that paternally inherited LCb118 was preferentially methylated (Fig. 8b; the methylated region common to both parental transmissions is outside of the DMR).
Next, we conducted a ChIP assay to analyze CTCF binding in the fetal liver (Fig. 8c, 18.5 dpc). When the LCb118 allele was maternally inherited, CTCF bound at significant levels to its sequence. By contrast, when the LCb118 allele was paternally inherited, CTCF was enriched at the maternally inherited WT H19 ICR. These results clearly demonstrate that CTCF bound the maternally inherited, hypomethylated sequences irrespective of whether they were H19 ICR or LCb118 (Fig. 8c).
Finally, we analyzed expression of the Igf2 and H19 genes by PCR amplification of cDNA prepared from the fetal liver RNA, followed by restriction enzyme digestion at sites containing strain-specific SNPs (Fig. 8d). The results revealed that Igf2 was expressed only from the alleles carrying either hypermethylated H19 ICR or LCb118 sequences in cis (Fig. 8d, upper), whereas H19 was active only when cis-linked H19 ICR or LCb118 sequences were hypomethylated (Fig. 8d, bottom). Similar results were obtained when liver RNA from E12.5 embryos was used (Additional file 8: Fig. S8A). By contrast, the H19 gene was aberrantly expressed from the paternally inherited LCb allele in multiple embryos (Additional file 8: Fig. S8B), in which LCb sequences exhibited lower methylation levels (Additional file 9: Fig. S9). In addition, the number of pups that inherited the LCb allele paternally was significantly smaller than expected (Additional file 10: Table S1). These results suggested that the LCb was insufficient to replace H19 ICR function in the regulation of genomic imprinting and proper development.
In summary, an artificially reconstituted LCb118 sequence knocked into the endogenous Igf2/H19 locus faithfully recapitulated the phenomenon of genomic imprinting, including establishment of methylation in the sperm, post-fertilization and post-implantation maintenance of differential methylation, allele-restricted CTCF binding, and control of monoallelic gene expression (Fig. 8e). In addition, our results clarify the role of the 118-bp sequence in the post-fertilization maintenance of paternally inherited endogenous H19 ICR.

Discussion
Allele-specific DNA methylation of ICRs plays a fundamental role in the regulation of genomic imprinting. Since most ICRs are differentially methylated during gametogenesis (i.e., gDMR), a great deal of attention has been focused on elucidating the molecular mechanism by which ICRs in primordial germ cells, where almost all pre-existing methylation is erased, eventually acquire asymmetric methylation during germ cell differentiation. Subsequently, genome-wide DNA methylation analysis identified many gDMRs that are methylated by the same mechanism as ICRs, but are unrelated to genomic imprinting [20,21]. Therefore, a critical difference between general gDMRs and ICRs is whether their (See figure on next page.) Fig. 8 Genomic imprinting recapitulated in LCb118 knock-in mice. a Breeding scheme. In order to distinguish parental origin of the alleles by using SNPs between inbred mouse strains, endo-LCb118 homo-knock-in mice (LCb118/LCb118; C57BL/6 J [B6] background) were mated with wild-type mice (H19 ICR/H19 ICR; JF1/Msf [JF1]), and offspring was obtained. b-d Livers from two pairs of E18.5 embryos, each inheriting the knock-in allele either paternally (pat.; P) or maternally (mat.; M) were used for genomic DNA, total RNA, and chromatin preparations, as in Fig. 6. b DNA methylation status of LCb118 region (the same position as in Fig. 7a) was analyzed by bisulfite sequencing. c ChIP analysis of CTCF occupancy at the LCb118 sequence. Chromatin was immunoprecipitated using either control IgG or anti-CTCF antibodies. Following qPCR analyses of three distinct genomic regions (Necdin; negative control, endogenous H19 ICR; positive control, and LCb118), relative enrichment values (CTCF/IgG signal ratio) were calculated. The average and standard deviation (S.D.), determined by three reactions, are depicted. Statistical differences were determined using an unpaired t test (N.S., not significant). d The allele-specific expression of the Igf2 and H19 genes was examined by RFLP analysis. Igf2 and H19 gene transcripts were amplified by RT-PCR followed by BstUI or Cac8I digestions, respectively. Parental origin of transcripts was discriminated by allele-specific restriction sites. The sites were also introduced into primer sequence so that complete digestion of PCR products can be concomitantly monitored. e Schematic representation of the genomic imprinting recapitulated in the LCb118 knock-in allele. f Hypothetical model for post-fertilization methylation maintenance mechanism at the Igf2/H19 locus Matsuzaki et al. Epigenetics & Chromatin (2020) 13:2 differential methylation statuses are maintained after fertilization. In other words, the mechanism that selectively maintains post-fertilization methylation at ICRs defines genomic imprinted regions. Sequence-dependent DNA-binding proteins are the most plausible candidates to support post-fertilization methylation maintenance at the specified ICRs. In fact, deficiency of ZFP57, one of the KRAB-zinc finger proteins (KZFPs), causes loss of methylation at multiple ICRs [16]. Through binding to its consensus DNA motif (5′-TGC CGC -3′) in the ICRs, ZFP57 maintains methylation via recruitment of a heterochromatic complex that contains KAP1, DNA methyltransferases, and histone methyltransferases [16,17,22]. ZFP57 binding to DNA depends on CpG methylation of the consensus motif, which in turn allows the maintenance of DNA methylation in an allele-specific manner [17]. Differential DNA methylation status was not affected at some ICRs, including the H19 ICR, in Zfp57-null mice [16,18], although ZFP57 binds these ICRs in ES cells [17], suggesting the existence of additional regulatory factors. Most recently, another KZFP, ZFP445, was reported to bind to methylated ICRs and participate in maintenance of their imprinted methylation status [23]. Allele-specific methylation at almost all ICRs was severely affected in Zfp57/Zfp445 double knockout mice, suggesting that these two proteins may coordinately maintain differential DNA methylation.
Despite the existence of recognition motifs for ZFP57, the LCb sequence and the H19 ICR sequence with a 116-bp deletion partly lost their DNA methylation at the paternal endogenous Igf2/H19 locus during the postfertilization period. Therefore, we propose additional mechanism(s) for maintenance of imprinted methylation. In our TgMs with the H19 ICR fragment, paternal allelespecific DNA methylation occurs at the transgene soon after fertilization [11,12], indicating that "de novo" methylation takes place in an allele-specific manner at the transgenic H19 ICR. Consistent with this, the paternally inherited H19 ICR fails to acquire methylation in early embryos when the supply of de novo methyltransferases, Dnmt3a and Dnmt3L, was eliminated by deletion of the corresponding genes in the oocyte [12]. We also demonstrated that this post-fertilization methylation activity existed at the endogenous H19 ICR as well [12]. Hence, we suggest that the maintenance of imprinted methylation during pre-implantation development is governed by de novo methylation activity mediated by paternal allele-and sequence-specific, yet DNA methylation-independent, DNA-binding factors. This notion is supported by our findings that a 118-bp sequence lacking any CpG motif (Additional file 3: Fig. S3A) is necessary and sufficient for post-fertilization imprinted DNA methylation.
It is unlikely that ZFP57 and/or ZFP445 act through the 118-bp sequence, as they recognize methylated DNA. In support of this hypothesis, we also failed to detect binding of the ZFP57 protein to the 118-bp sequence in gelshift assays [15].
How do these two seemingly independent mechanisms collaborate to maintain methylation imprinting? As mentioned earlier, deletion of the 116-bp sequence from the endogenous H19 ICR resulted in reduced methylation of this locus during the pre-implantation period, suggesting that the methylation maintenance activity of ZFP57/ ZFP445 during this period is insufficient. Due to predominant genome-wide demethylation activity during this period, additional maintenance involving the 118-bp sequence of the H19 ICR may be necessary. By contrast, Takahashi et al. reported that almost all methylation was lost at the endogenous ICRs in Zfp57/Zfp445 double mutant mice by around E11.5. Furthermore, we previously suggested that post-fertilization de novo methylation activity of the H19 ICR disappears sometime during early embryogenesis [12]. These results together imply that ZFP57/ZFP445-dependent activity is the sole mechanism responsible for post-implantation methylation maintenance at the H19 ICR.
We can envision two compatible mechanisms by which the 118-bp sequence could contribute to de novo methylation at the paternally inherited H19 ICR soon after fertilization (Fig. 8f ). First, since histones rather than protamine are retained at the H19 locus in sperm [24,25], the 118-bp sequence might be involved in the establishment of epigenetic modifications during spermatogenesis, either as a binding site for specific histone modification enzymes or as the deposition site for the marks. Such a non-methylation mark would then be utilized to distinguish the parental origin of the alleles and somehow be translated into differential DNA methylation after fertilization. Second, the sequence might act as a scaffold for recruitment of de novo DNA methyltransferases in pre-implantation embryos. Specific DNA-binding factors, which have not yet been identified, might recognize the 118-bp sequence associated with allele-discriminating signatures and recruit de novo DNA methyltransferases (i.e, Dnmt3A and 3L) in early embryos. Identification of the factors that bind the 118bp sequence should provide insight into the molecular mechanism of post-fertilization, allele-specific methylation at the H19 ICR.
IG (intergenic)-DMR of the Dlk1-Dio3 imprinted domain is one of the three ICRs that acquires DNA methylation in sperm. Recent work showed that deletion of a tandem repeat sequence (300-400 bp) from the paternal IG-DMR caused loss of methylation only after the fertilization period [26]. Since the murine repeat array of the IG-DMR contains several consensus binding sites for Zfp57, it is conceivable that the phenotype was caused by a loss of Zfp57-dependent methylation maintenance. Curiously, however, the consensus motifs are not present in the repeat arrays of the human and sheep sequences [27]. Therefore, it is conceivable that a Zfp57-independent mechanism that is common to both H19 ICR and IG-DMR is operating at these paternal gDMRs through the 118-bp and the repeat array sequences, respectively, although they do not share significant sequence homology. In addition, the corresponding region of the human H19 ICR sequence (hIC1) is not strongly similar to the mouse 118-bp sequence, and Hur et al. failed to recapitulate paternal methylation of the hIC1 (4.8 kb) when knocked into the mouse Igf2/H19 locus [28]. It remains an open question whether the mechanism of post-fertilization methylation maintenance we found in the mouse H19 ICR is conserved in other mammals, especially in humans, and whether it is also employed at other imprinted loci.

Conclusions
We showed that the 118-bp region of the H19 ICR is responsible for post-fertilization acquisition of DNA methylation at the paternal ICR in both transgenic and endogenous loci. The reconstituted LCb118 fragment not only exhibited methylation dynamics identical to that of the wild-type H19 ICR fragment in the transgenic context, but also recapitulated imprinted methylation and imprinted expression of the Igf2/H19 genes when used to replace the endogenous H19 ICR. These results demonstrated that the imprinted status in the mouse genome can be generated by an artificial fragment that includes a limited number of cis elements.

Preparation of co-placement yeast targeting vector for LCb/LCb118 sequences
Purified YAC DNA was microinjected into fertilized mouse eggs from CD1 (ICR) (for generation of 5′-del TgM) or C57BL/6 J (for generation of LCb/ LCb118 TgM) mice. Tail DNA from founder offspring was screened first by PCR, followed by Southern blotting. Structural analysis of the YAC transgene was performed as described elsewhere [29,30]. TgM ubiquitously expressing cre recombinase [31] or TgM carrying Zp3-Cre gene (Jackson Laboratory; [32]) were mated with parental YAC-TgM lines to generate sublines (i.e., each carrying one of the test fragments). Successful Cre-loxP recombination was confirmed by Southern blotting.
CRISPR/Cas9-assisted homologous recombination in ES cells B6 J-S1 ES cells derived from C57BL/6J mouse strain [34] were maintained in DMEM High Glucose Cells were then stripped, made single-cell suspension and seeded onto feeder-cell plates in the medium without puromycin. After 3 days culture, colonies were picked up, expanded and homologous recombination event was checked by PCR and Southern blotting with several combinations of restriction enzymes and probes. Chimeric mice were generated by a coculture method using eightcell embryos from CD1 mice (ICR, Charles River Laboratories). Chimeric males were bred with B6J females, and germ line transmission of the mutant allele was identified by Southern blot analysis.

Preparation of embryos
Female mice were super-ovulated via injection of pregnant mare serum gonadotropin, followed by human chorionic gonadotropin (hCG) (47-48 h interval). Two-cell embryos were flushed from oviducts by M2 medium at 44 h after hCG injection, and then washed by PBS. Embryos at E3.5 (blastocysts), E12.5, and E18.5 were obtained by natural mating.

DNA methylation analysis by southern blotting
Genomic DNA extracted from tail somatic cells was first digested by EcoT22I (for analysis of the 5′-deleted transgenic H19 ICR) or BamHI (for analysis of the 116bp deleted transgenic H19 ICR or the LCb and LCb118 transgenes) and then subjected to the methylationsensitive enzymes BstUI or HhaI. Following size separation in agarose gels, Southern blots were hybridized with α-32 P-labeled probes and subjected to X-ray film autoradiography.

DNA methylation analysis by bisulfite sequencing
Pre-implantation embryos were embedded in agarose beads and treated with sodium bisulfite as described previously [10]. Genomic DNA extracted from adult male sperm or the tail tips of ~ 1-week-old animals was treated with sodium bisulfite using the EZ DNA Methylation Kit (Zymo Research). Sperm and tail tip DNA was digested with XbaI prior to the treatment. Subregions of the transgenic H19 ICR and the transgenic, as well as the knock-in LCb/LCb118 fragments were amplified by nested PCR, while those of the endogenous H19 ICR with the 116-bpdeletion were amplified by single-round PCR. The PCR products were subcloned into the pGEM-T Easy vector (Promega) for sequencing analyses. PCR primers are listed in Tables 1 and 2.

Chromatin immunoprecipitation (ChIP) assay
The LCb118 YAC-TgM (2-4 months old) inheriting the transgene either paternally or maternally were made anaemic by phenylhydrazine treatment, and nucleated erythroid cells were collected from their spleens. Livers were obtained from E18.5 embryos inheriting the LCb118 knock-in allele either paternally or maternally. Cells were fixed in PBS with 1% formaldehyde for 10 min at room temperature. Nuclei (2 × 10 7 cells) were digested with 12.5 units/ml of micrococcal nuclease at 37 °C for 20 min to prepare primarily mono-to di-nucleosomesized chromatin. The chromatin was incubated with anti-CTCF antibody (D31H2; Cell Signaling Technology) or purified rabbit IgG (Invitrogen) overnight at 4 °C and was precipitated with preblocked Dynabeads protein G magnetic beads (Life Technologies, Carlsbad, CA). Immunoprecipitated materials were then washed extensively and reverse cross-linked. DNA was purified with the QIAquick PCR purification kit (Qiagen, Venlo, the Netherlands) and subjected to qPCR analysis. The endogenous H19 ICR and Necdin sequences were analyzed as positive and negative controls, respectively [35]. PCR primers were reported previously [35].   ICR-MA-5S4  5′-GAA TTT GGG GTA TTT AAA GTT TTG -3′   ICR-MA-5S13  5′-GGT GAT TTA TAG TAT TGT TAT TTG -3′   LCR-MA-5S1  5′-TAT AGA TGT TTT AGT TTT AAT AAG -

RT-qPCR
Total RNA was recovered from phenylhydrazine treated anaemic adult spleens (1-2 months old) of LCb118 YAC TgM using ISOGEN (Nippon Gene) and converted to cDNA using ReverTra Ace qPCR RT Master Mix with gDNA Remover (TOYOBO). Quantitative amplification of cDNA was performed with the Thermal Cycler Dice (TaKaRa Bio) using TB Green Premix EX TaqII (TaKaRa Bio). PCR primers were reported previously [15].

Allele-specific expression analysis
LCb118 knock-in mice (which has a genetic background of Mus musculus domesticus) were mated with wild-type JF1 mice (which was provided by the RIKEN BRC through the National Bio-Resource Project of the MEXT, Japan, and of which genome is basically from Mus musculus molossinus) to distinguish parental origin of the alleles in the offspring. Total RNA from livers of E12.5 or E18.5 embryos was converted to cDNA as described above. PCR was performed using AmpliTaq Gold 360 and PCR primers listed in Table 3 with (E12.5 cDNA) or without (E18.5 cDNA) α-32 P-dCTP. The amplified products were digested with BstUI or Cac8I, in order to discriminate the parental origin of the transcripts. The sites were also introduced into primer sequence so that complete digestion of PCR products can be concomitantly monitored.