Transgenic RNAi Reveals Essential Function for CTCF in H19 Gene Imprinting

See allHide authors and affiliations

Science  09 Jan 2004:
Vol. 303, Issue 5655, pp. 238-240
DOI: 10.1126/science.1090934


The imprinted regulation of H19 and Insulin-like growth factor 2 expression involves binding of the vertebrate insulator protein, CCCTC binding factor (CTCF), to the maternally hypomethylated differentially methylated domain (DMD). Howthis hypomethylated state is maintained during oogenesis and the role of CTCF, if any, in this process are not understood. With the use of a transgenic RNA interference (RNAi)–based approach to generate oocytes with reduced amounts of CTCF protein, we found increased methylation of the H19 DMD and decreased developmental competence of CTCF-deficient oocytes. Our results suggest that CTCF protects the H19 DMD from de novo methylation during oocyte growth and is required for normal preimplantation development.

Genes that are subject to genomic imprinting in mammals exhibit differential expression of the parental alleles (1). The transcriptional machinery is thought to differentiate between the two alleles through heritable DNA and histone modifications. Although recent work has characterized epigenetic marks in somatic lineages, the mode by which this epigenetic control is established in the germ line remains poorly understood.

The maternally expressed H19 gene lies about 100 kb from the paternally expressed Insulin-like growth factor 2 gene (Igf2) (2). The imprinted expression of both genes depends on a 2-kb differentially methylated domain (DMD) located 2 kb upstream from the H19 promoter (3, 4) (Fig. 1A). The DMD is postulated to function as a methylation-sensitive insulator through the binding of the insulator protein CTCF to conserved upstream sequence elements (59). Binding of CTCF to the hypomethylated maternal DMD allows H19 exclusive access to downstream enhancers. Mutation of CTCF binding sites causes acquisition of methylation during postimplantation development, suggesting that CTCF also maintains maternal allele–specific hypomethylation (10, 11).

Fig. 1.

Targeting of CTCF mRNA by transgenic RNA interference. (A) Model for imprinted regulation of H19 and Igf2. Enhancer blocking is established on the maternal allele, allowing H19 exclusive access to shared enhancers (indicated with an “e”). Igf2 is solely expressed from the paternal chromosome, where hypermethylation of the DMD abrogates CTCF binding. (B) Typical reverse transcription PCR result for each transgenic line, comparing CTCF transcripts in oocytes from transgenic and nontransgenic littermates (16). GAPD was used to normalize expression levels between samples. Amounts of CTCF mRNA in transgenic oocytes, compared to that of nontransgenic oocytes, are shown below gel images.

Most imprinted genes examined to date harbor maternal-specific hypermethylated imprinting control regions (12). In contrast, the H19 DMD possesses one of the few maternal hypomethylation marks, implying that mechanisms may exist that not only confer germline methylation but also protect it from such modifications. It has been suggested that CTCF may be responsible for this protective role (13). To determine whether CTCF is required to maintain hypomethylation of the H19 DMD during this crucial period, we selectively ablated CTCF mRNA in growing oocytes.

Long, double-stranded RNAs are an effective and specific means of targeting genes in oocytes and preimplantation embryos (14). A system has been described that expresses an RNA hairpin under the control of a promoter specific to growing oocytes [Zona pellucida 3 (15)]. We generated mice harboring such a transgene, which was designed to target CTCF mRNA (16) (fig. S1A). Five transgenic founders transmitted the construct to their offspring and were chosen for further study (16) (fig. S1B).

To assess the ability and extent of targeting in oocytes, we used a polymerase chain reaction (PCR) assay that permits the comparison of relative amounts of CTCF transcript between transgenic and nontransgenic littermates (16). Each transgenic line reduced CTCF mRNA amounts, but the degree of targeting was family-specific (Fig. 1B). Line 12 exhibited almost complete loss of CTCF mRNA, and lines 1 and 21 also had very little CTCF mRNA. The other two transgenic families, 4 and 56, had more modest decreases. In independent experiments, glyceraldehyde-3-phosphate dehydrogenase (GAPD) levels were unaffected in all lines examined (17), suggesting that the effect was specific.

Amounts of protein were also diminished in transgenic oocytes. Immunostaining of nontransgenic oocytes showed that CTCF was strongly localized to the germinal vesicle (GV) (Fig. 2, A and F). Oocytes from transgenic lines exhibited moderate to barely detectable levels of CTCF protein, correlating strongly to the degree of transcript reduction (Fig. 2 and fig. S2). Reduced mRNA and protein levels allowed us to determine whether CTCF reduction affected methylation of the H19 DMD in this epi-allelic series.

Fig. 2.

CTCF protein levels in nontransgenic and transgenic oocytes. CTCF was visualized with monoclonal antibody against CTCF and Cy3-conjugated antibodies against mouse immunoglobulin G (IgG) (A to E). Oocytes from nontransgenic and transgenic females were stained with 4′,6′-diamidino-2-phenylindole (DAPI) (F to J). (A and F) Nontransgenic oocytes showing CTCF localized to the germinal vesicle (white arrowheads). (B and G) Nontransgenic oocytes stained with Cy3-conjugated antibodies to mouse IgG alone. Oocytes are shown for transgenic lines 12 (C and H), 1 (D and I), and 4 (E and J). Specificity of CTCF targeting in oocytes is confirmed by CTCF positive staining in the cumulus cells (asterisk) in (A) and (C). The scale bar represents 40 μm.

Oocytes from all transgenic lines were analyzed for methylation in the 5′ portion of the DMD with the use of a bisulfite mutagenesis and sequencing assay. As previously shown, the H19 DMD was hypomethylated in the oocytes of nontransgenic females (Fig. 3A) (18). In contrast, the H19 DMD was hypermethylated in the three lines showing the strongest decrease of CTCF mRNA and protein (Fig. 3B and fig. S3). DNA strands exhibited a wide variety of methylation patterns, ranging from the typical wild-type hypomethylated state to complete methylation. The DMD remained hypomethylated in oocytes from lines 4 and 56 (Fig. 3B and fig. S3), consistent with their higher levels of CTCF protein. These data suggest that substantial loss of CTCF protein is associated with hypermethylation of the H19 DMD.

Fig. 3.

Methylation of the H19 DMD in nontransgenic and transgenic oocytes. (A) Schematic of the H19 and skeletal α-actin genomic loci, highlighting the areas assayed by bisulfite mutagenesis (16). The CTCF sites in the H19 DMD are indicated with asterisks. The methylation profiles from nontransgenic oocytes are shown above each locus; each line denotes an individual strand of DNA with open and solid circles corresponding to unmethylated and methylated CpGs, respectively. The number of strands observed with a given methylation profile (if greater than one) is indicated to the left of each line. Black bars represent CTCF binding sites. (B) Methylation pattern of the H19 DMD in oocytes from transgenic (tg) lines 12, 1, and 4. (C) Methylation profile of skeletal α-actin promoter in transgenic oocytes from line 1.

To investigate the possibility that the observed DMD hypermethyation represented nonspecific effects, we examined other regions of the genome. If hypermethylation of the DMD were a result of global increases in methylation, other regions may also acquire similar levels of methylation. The promoter region of the Skeletal α-actin gene is hypomethylated in oocytes (Fig. 3A) (19). Our results demonstrate that, even in transgenic lines where the H19 DMD acquired methylation, the Skeletal α-actin region remained free of methylation (Fig. 3C). The differentially methylated region 1 of the imprinted gene Snurf-snrpn is maternally hypermethylated (20). Transgenic oocytes showed no differences in the hypermethylation of this locus as compared to nontransgenic littermates (17). Therefore, it is unlikely that H19 hypermethylation in transgenic oocytes is a result of global hypermethylation or disruption of imprinting patterns.

Transgenic females produced fewer offspring than their nontransgenic littermates (Fig. 4A); pups born from transgenic females appeared normal (17). The degree of this effect correlated with reduction in CTCF protein level, because line 12 showed the greatest decrease in litter size, whereas line 4 showed a modest, yet significant, effect. To determine whether the transgene caused infertility or defects in viability of the zygotes, we examined eggs from line 12 by in vitro fertilization. The results demonstrated that fertilization incidence between transgenic and nontransgenic eggs was identical (Fig. 4B). In contrast, the zygotes from the transgenic females were unable to develop to blastocyst-stage embryos as successfully as control zygotes (Fig. 4C). Therefore, knock-down of CTCF in the oocyte resulted in increased zygotic lethality, demonstrating a requirement for this protein in early embryos. Because appropriate imprinting or expression of H19 is not required for normal development (4, 21), depletion of CTCF is likely affecting other essential processes.

Fig. 4.

Effect of reduced CTCF on fertility and preimplantation development. For each graph, open bars represent transgenic lines, whereas solid bars are nontransgenic littermates. (A) Average litter sizes for transgenic and nontransgenic females. P values for lines 12, 1, and 4 are 0.001, 0.036, and 0.031, respectively. (B) Incidence of fertilization and development to blastocyst stage among eggs from line 12 transgenic and nontransgenic females. Incidence of fertilization is expressed as the percentage of eggs forming pronuclei, whereas blastocyst development reflects the proportion of fertilized eggs developing blastocyst morphology 96 hours after fertilization. The mean and standard error for three experiments are shown. P values for pronuclear formation and blastocyst development are 0.11 and 0.01, respectively. (C) Developmental stage of embryos derived by in vitro fertilization 96 hours after fertilization (line 12). Values are percentages of total embryos derived from transgenic (n = 111) and nontransgenic (n = 202) eggs.

CTCF is required to maintain the hypomethylated maternal H19 DMD. We show that this requirement for CTCF is during a critical period in which methylation marks are being set in the growing oocyte. Our results also indicate that imprinting control regions are recognized by the methylation machinery of the female germ line (22, 23) and that hypermethylation at loci such as the H19 DMD must be actively blocked. Precedence for this mechanism exists in the embryo, where Sp1 elements protect CpG islands from de novo methylation (24, 25). Recently, Schoenherr et al. mutated the CTCF sites in the endogenous H19 DMD and observed that, although this allele acquired methylation during postimplantation development, it remained hypomethylated in oocytes (10). These contradictory results may be explained by a modulation in CTCF binding by cofactors and/or posttranslational modifications in the female germ line (26). Thus, mutations to the DMD that abrogate CTCF binding in the soma may not be as effective in the germ line. Together, these data suggest that imprinted states may be erased, established, and maintained by the precise regulation of epigenetic modifiers and protective factors during germline development and gametogenesis.

Supporting Online Material

Materials and Methods

Figs. S1 to S3

Table S1

References and Notes

View Abstract

Navigate This Article