Functional Characterization of a Novel Ku70/80 Pause Site at the H19/Igf2 Imprinting Control Region

Abstract

The imprinted expression of the H19 and Igf2 genes in the mouse is controlled by an imprinting control center (ICR) whose activity is regulated by parent-of-origin differences in methylation. The only protein that has been implicated in ICR function is the zinc-finger protein CTCF, which binds at multiple sites within the maternally inherited ICR and is required to form a chromatin boundary that inhibits Igf2 expression. To identify other proteins that play a role in imprinting, we employed electrophoresis mobility shift assays to identify two novel binding sites within the ICR. The DNA binding activity was identified as the heterodimer Ku70/80, which binds nonspecifically to free DNA ends. The sites within the ICR bind Ku70/80 in a sequence-specific manner and with higher affinity than previously reported binding sites. The binding required the presence of Mg²⁺, implying that the sequence is a pause site for Ku70/80 translocation from a free end. Chromatin immunoprecipitation assays were unable to confirm that Ku70/80 binds to the ICR in vivo. In addition, mutation of these binding sites in the mouse did not result in any imprinting defects. A genome scan revealed that the binding site is found in LINE-1 retrotransposons, suggesting a possible role for Ku70/80 in transposition.

The reciprocal imprinting of the maternally expressed H19 gene and the paternally expressed Igf2 gene in the mouse is controlled by a single imprinting control region (ICR) that lies 5′ of H19 (32, 48). The activity of the ICR is controlled by its methylation status, which is established differentially during oogenesis and spermatogenesis (6, 7). On the unmethylated maternal allele in the soma, the ICR acts as a chromatin boundary that silences Igf2 at a distance by blocking multiple tissue-specific enhancers that lie 3′ of H19 from activating the Igf2 gene (3, 24, 28). On the paternal allele, gametic methylation on the ICR spreads to the H19 promoter, thereby silencing the H19 gene (11). The methylation also inhibits the formation of the boundary, allowing Igf2 to be expressed (3, 24). That the ICR is both necessary and sufficient for imprinting is demonstrated by the transcriptional consequences of ICR deletions in mice (32, 48) as well as the ability of the ICR to establish parental-specific differences at ectopic sites (28, 41).

The chromatin boundary activity at the H19/Igf2 locus requires the zinc-finger binding protein CCTC binding factor (CTCF) that binds, in a methylation-sensitive manner, to four sites within the ICR (3, 24). Mutations in the four CTCF sites eliminate boundary function and result in the accumulation of methylation on the maternal chromosome after fertilization, suggesting that CTCF is one of the factors that maintain the maternal ICR in an unmethylated state (40, 43). However, the CTCF sites are not required to establish the maternal allele in an unmethylated state during oogenesis, implying that other factors may be involved (40, 43). Using a different approach, Fedoriw et al. (16) have recently shown that inhibiting CTCF with a transgene expressing CTCF-specific RNA interference in oocytes leads to the acquisition of methylation on the ICR. One unlikely possibility to explain the contradiction between the two findings is that the mutations that eliminate binding in vitro are less deleterious in vivo. A more likely alternative is that CTCF binds elsewhere in the region or has an indirect effect on the ICR that does not require binding to DNA.

On the paternal ICR, mutations in the CTCF sites have no effect on the accumulation of DNA methylation in testes or somatic tissues (40, 43). Although the simplest explanation for the accumulation of methylation during spermatogenesis is the absence of CTCF to prevent it, the protein has been detected by immunofluorescence in spermatogonia that are undergoing methylation, suggesting that additional regulation is required (33). This is consistent with the finding that when two CTCF sites from the chicken β-globin locus were inserted in place of the ICR at the H19 locus, paternal methylation did not occur and chromatin boundaries were formed on both alleles (46). More recently, Engel et al. (15) depleted the number of CpG residues in the ICR without disrupting the ability of CTCF to bind and showed that this had no effect on the establishment of paternal methylation in the germ line but resulted in a loss of maintenance methylation in the soma. Thus, sequences outside the CTCF sites themselves must play a role in methylation maintenance.

Taken together, these experiments suggest that there may be one or more proteins other than CTCF that play a role in the establishment and maintenance of parental-specific epigenetic differences at the ICR, and in vivo footprinting of the ICR has reinforced this conclusion (44, 45). Recently, two groups identified a series of sites within the ICR that bind specifically and exclusively to fetal and neonatal testes extracts derived from the stages of development when paternal methylation is being assembled on the ICR (5, 45). These sites contain a core that is similar to the consensus sequence for nuclear receptor extended half sites, and mutations in the half site eliminate binding. However, when Bowman et al. (5) mutated three of these sites in vivo, no effect on imprint establishment or maintenance was observed.

This report extends the search for additional proteins by using extracts from liver, a somatic tissue. We show that the ICR contains two sites that are very strong pause sites in vitro for the Ku70/80 heterodimer that has been implicated in nonhomologous recombination. Using chromatin immunoprecipitation (ChIP) assays and genetic mutations in mice, we conclude that these sites are not involved in the regulation of imprinting at the H19/Igf2 locus.

MATERIALS AND METHODS

Preparation of protein extracts.

Individual livers from adult mice were homogenized either by hand or with a motorized type B pestle for 10 to 15 strokes in 1.5 ml of NE2 extraction buffer (250 mM sucrose, 10 mM HEPES [pH 7.9], 450 mM KCl, 2 mM MgCl₂, 2 mM CaCl₂, 0.1% Triton X-100, 1 tablet of protease inhibitor [Roche]/10 ml). The homogenized tissue was centrifuged twice at 11,000 rpm for 3 to 5 min each at 4°C and then centrifuged at 180,000 × g at 4°C for 90 min. The supernatant was decanted and stored in liquid N₂ at −80°C.

Electrophoretic mobility shift assays (EMSAs).

DNA restriction fragments were radiolabeled with the Klenow fragment of DNA polymerase, while oligonucleotides were labeled with T4 polynucleotide kinase and annealed. The sequences of the oligonucleotide probes are shown in Table 1. All labeled probes were purified in 5% acrylamide gels. Mobility shift reactions were carried out in 40 μl at room temperature for 30 min in 20 mM HEPES-KCl (pH 7.9), 150 mM KCl, 5 mM MgCl₂, 1 mM dithiothreitol, 40,000 cpm radiolabeled DNA, 0.05 μg/μl poly(dI:dC) (0.015 μg/μl for recombinant human Ku70/80), 5% glycerol, 0.5% Triton X-100, and 1 to 5 μl protein extract (20 to 40 mg/ml). The reaction mixtures were analyzed without loading dye on 1.5-mm-thick 5% acrylamide gels.

TABLE 1.

Sequences of oligonucleotide probes

Probea	Sequence
HS1	TGCCGCGTGGTGGCAAAATCGATTGCGCCAAACCTAAAGAGCC
HS2	TCAGTCTTGCGCCCTTCACGATCGATCGGTTCACTCTCCACGCTGT
GRMMTV	AACTGAGAAAGAGAAAGACGACA
C3HMMTV	GTCTCAAGAAGAAAAAGACGACA
CMYCPRE	GTACAGAAAGGGAAAGGGACTAGC
U5HTLV	GAATGAAAGGGAAAGGGGTGGAAC
OCTAMER	AGCTTGCTTATGCAAATAAGGTG
U1PSE	AAGTGACCGTGTGTGTAAAGAGTAG
HSE	CCCGAAACTGCTGGAAGATTCTTGGCCC
IAPENH	CTGCGCATGTGCCAAGGGTATCTTATGACT
TCBE3A	CCAGGAGTCACAACAGGATGTGGTTTG
TCBBE1	TCTCACCCCAGGTCTGGCTGTTTATCTGTAAGTAACA

^aThe sequences of the EMSA oligonucleotides are HS site 1 (HS1), HS site 2 (HS2), mouse mammary tumor virus (GRMMTV), C3H element (C3HMMTV), c-myc (CMYCPRE), human T-cell leukemia virus I (U5HTLV), immunoglobulin octamer motif (OCTAMER), U1 snRNA (U1PSE), heat shock element (HSE), mouse retrovirus-like element (IAPENH), TCR-β enhancer E3 (TCBE3A), and TCR-β enhancer E4 (TCBBE1).

Protein purification.

A whole-cell extract was prepared from the livers of 83 adult mice. The final high-speed centrifugation was done at 70,000 rpm in a Ti70 rotor for 142 min. The supernatants were fractionated in batches on a 750-ml S300 gel filtration column in GF buffer (20 mM HEPES-KCl [pH 7.9], 150 mM KCl, 5 mM MgCl₂, 1 mM dithiothreitol). Active fractions were pooled and applied three times to a 6-ml DNA affinity column generated according to the methods of Marshak et al. (35) with a 46-bp oligonucleotide from HS site 2 (Table 1) with SalI and XhoI compatible ends. The affinity column was eluted with 1 M KCl in GF buffer, and the active fractions were pooled, dialyzed into 150 mM KCl in GF buffer, and applied to a 1-ml Mono Q cation-exchange column. The column was eluted with a 200 to 500 mM KCl gradient, and the active fractions were pooled and analyzed on a 7.5% acrylamide gel. The protein bands were excised and sequenced at the Harvard Microchemistry Facility by mass spectrometry.

Chromatin immunoprecipitation.

Adult liver (0.3 g) was sliced in 10 ml of phosphate-buffered saline with 1% formaldehyde and protease inhibitors (1 tablet/50 ml) at 25°C, homogenized for 10 strokes in a type B glass Dounce homogenizer, and incubated at 25°C for 10 min. The cross-linking was terminated with the addition of 500 μl of 2.5 M glycine. The samples were centrifuged at 4°C for 2 min at 2,000 rpm and resuspended three times in 10 ml of ice-cold phosphate-buffered saline. The final pellet was resuspended in 400 μl of lysis buffer (Upstate Biotechnology ChIP kit) and incubated on ice for 10 min, followed by the addition of 1.6 ml of ChIP dilution buffer (Upstate Biotechnology ChIP kit). The samples were sonicated three times with 10 pulses at 50% output, power 10, with a Branson Sonicator 2000 and centrifuged at 13,000 rpm at 4°C for 10 min. The immunoprecipitation was carried out on the supernatant using the Upstate Biotechnology ChIP kit and 10 μl of each antibody (CTCF polyclonal from Upstate [06-917], Ku70 polyclonal from Santa Cruz [sc-1487], Ku70 monoclonal from Neomarkers [Ab-4]). Quantitative PCR was performed on the precipitated DNA with a Bio-Rad iCycler. The A primers were F, 5′-TCTGCTGAATCAGTTGT, and R, 5′-GGGCTCTTTAGGTTTGG, and the B primers were F, 5′-AATCAACAAGGTCGGCTTACTC, and R, 5′-CTGCGATGTACGAGACTTCAC. The copy number was reported relative to that of CTCF, which was arbitrarily set at 100. In addition, all values were normalized to the copy number of each sample before immunoprecipitation.

Generation of H19^{Ku site−} mice.

Mutations in the two ATCGAT binding sites were introduced into a 4.7-kb XbaI genomic DNA fragment (from −5.5 to −0.8 kb relative to the H19 transcription start site) by site-directed mutagenesis (using the QuikChange site-directed mutagenesis kit from Stratagene) and integrated into the targeting construct of Schoenherr et al. (43). The vector was introduced into AB2.2 embryonic stem (ES) cells, and colonies were selected in the presence of G418 as described previously. Homologous recombination at the H19 locus was tested by Southern blot analysis with both external and internal probes. Detection of recombination between loxP sites to remove the neomycin selection cassette was achieved with the 3′ external probe and a SpeI genomic DNA digest. The presence of the mutant binding sites was confirmed by PCR followed by digestion with ClaI to verify elimination of the ATCGAT ClaI sites. The primers that were used for HS site 1 were F, 5′-CTACATTCACACGAGCATCCAGG, and R, 5′-GATGTACGAGACTTCACTGCCGC. The primers that were used for HS site 2 were F, 5′-GGCAAACTCATGGGTCACTCAAGGC, and R, 5′-GGTACTCAAGCTTTGTCACAGCGG.

Analysis of RNA.

RNase protection assays were performed using the RPAIII RNase protection assay kit from Ambion. The H19 and Igf2 probes were prepared as previously described (26) and incubated overnight with 10 μg of total neonatal liver RNA at 42°C. RNase digestion was carried out at 55°C for 1 h, and the protected fragments were separated on 5% acrylamide gels containing 8 M urea.

RESULTS

Identification of ICR binding activity.

To identify additional proteins that bind to the ICR, an EMSA screen was performed using nuclear extracts prepared from neonatal liver. Using a series of overlapping DNA probes, two regions of DNA binding were identified that coincided with the regions of the ICR that are hypersensitive to nuclease digestion in chromatin (HS1 and HS2) (Fig. 1A) (25, 31). Site 1 lies immediately 3′ of the first of two CTCF sites in HS1, and site 2 lies between the two CTCF sites in HS2.

FIG. 1.

Identification and characterization of the ATCGAT binding activity. (A) The top line depicts the H19 gene, with the horizontal arrow indicating the start of transcription. The two vertical arrows indicate the position of the DNase I-hypersensitive sites within the ICR. The second line indicates the positions of the four CTCF sites (closed circles) within the ICR and the DNA probes used in the EMSA analysis, with those exhibiting DNA binding activity drawn as black lines and those that did not bind protein shown by hatched lines. The vertical lines indicate the positions of the oligonucleotide probes for HS site 1 and site 2. (B) EMSA analysis of neonatal liver extracts using HS site 1 and HS site 2. Lanes 1 and 9, no extract; lanes 2 to 8 and 10 to 16, analysis in the presence of extract with unlabeled HS site 1, HS site 2, or a nonspecific competitor from the AFP promoter (NS) (18) added in 10 and 100 M excess. (C) EMSA analysis of HS site 2 DNA in the absence (lane 1) or presence of a 50-, 500-, or 1,000-fold excess of HS site 2, a mutant form of HS site 2 where the ATCGAT motif has been mutated to CACGCA (lane 3), and a series of oligonucleotides in which a single base in the sequence has been changed, as indicated below the figure (lanes 4 to 25). The bases that are critical to binding are shown in the shaded box.

Binding to HS site 1 and HS site 2 was specific, based on the inability of a nonspecific fragment derived from the α-fetoprotein (AFP) promoter (18) to reduce binding in a 100-fold excess (Fig. 1B). Furthermore, the two DNA binding sites were effective competitors against one another, suggesting that the same protein or protein complex bound each site (Fig. 1B). When the sequences of the sites were compared, a common ATCGAT motif was found. Base substitutions of each base pair within the ATCGAT motif severely inhibited the ability of HS site 2 to compete with itself for binding, while mutations of single base pairs outside of the ATCGAT motif had no effect (Fig. 1C). A search of the equivalent region in the human H19 gene revealed a single copy of the ATCGAT motif (data not shown).

To assess the distribution of the binding activity, protein extracts were prepared from a variety of tissues at different times during development. The activity was found in whole embryo extracts prepared at embryonic day 10.5 and embryonic day 16.5 and in adult liver, oocytes, and neonatal testes, implying that the activity is ubiquitously expressed in mice (data not shown).

Purification of ATCGAT binding activity.

To identify the ATCGAT binding activity, we prepared whole-cell extracts from 83 adult mouse livers, yielding 4.8 g of protein. The binding activity, which was followed with EMSA, separated on an S300 gel filtration column with an apparent molecular mass of 150 kDa (Fig. 2). The eluant was applied to a sequence-specific DNA affinity column containing multimerized copies of the ATCGAT binding site. The active fractions that eluted with 1 M KCl were pooled and applied to a Mono Q cation-exchange column. Two prominent bands of approximately 70 kDa and 80 kDa were visualized by silver staining in active fractions of the eluant (Fig. 2). These bands were sequenced by mass spectrometry at the Harvard Microchemistry Facility.

FIG. 2.

Biochemical purification of Ku70/80. The purification of the ATCGAT binding activity was carried out by sequential chromatography using gel filtration, DNA affinity, and Mono Q columns, with the binding activity monitored by EMSA. The extracts loaded onto each column and the eluant fraction exhibiting the highest level of binding were analyzed by acrylamide gel electrophoresis and Coomassie staining (A) or silver staining (B). The positions of Ku70 and Ku80 are indicated by arrows.

Sequence analysis identified the two polypeptides as Ku70 and Ku80, which are known to form a heterodimer, consistent with the sizes of the gel filtration complex and the individual bands on the silver-stained gel. Ku70/80 are ubiquitous nuclear proteins that bind to double-strand DNA breaks (9, 39, 51). Following a double-stranded break, the heterodimer is thought to function by holding the broken ends together and recruiting the DNA ligase IV-XRCC4-DNA-dependent protein kinase complex to repair the break (13, 21). Furthermore, this complex may have a role in signaling to the cell that a double-stranded break has occurred (14). Evidence for the role of the Ku complex in nonhomologous end joining (NHEJ) can be seen in the extreme radiation sensitivity of Ku-deficient cells as well as in the specific V(D)J recombination defects of the Ku70^−/− and Ku80^−/− mice (10, 22, 23, 38, 47, 52). In addition to the defects in V(D)J recombination, Ku70^−/− and Ku80^−/− mice also have severe growth retardation, a phenotype that is also observed in mice carrying a deletion of the Igf2 gene (8, 23, 38).

DNA binding specificity of the Ku70/80 complex.

Ku70/80 is thought to bind with high affinity to duplex DNA ends, but in a sequence-nonspecific manner. As such, its purification as the ATCGAT binding activity was unexpected. To verify that the ATCGAT binding activity is composed of Ku70 and Ku80, commercially available human recombinant Ku70 and Ku80 proteins were incubated in the presence of radiolabeled HS site 2. A retarded mobility complex indistinguishable from that obtained with the neonatal liver extract was observed (Fig. 3A). Furthermore, the recombinant Ku-DNA complex was competed specifically with site 2, but not with the AFP-nonspecific competitor. Finally, the human Ku heterodimer exhibited the same sequence specificity as the ATCGAT activity when challenged to bind the oligonucleotides containing base substitutions in site 2 (Fig. 1C and data not shown).

FIG. 3.

Binding properties of the Ku complex. (A) The mobility of the Ku70/80 complex bound to HS site 2 in neonatal liver extracts was compared to that of recombinant human Ku70/80. The complexes were formed in the absence (-) or presence of a 100-fold molar excess of HS site 2 or the AFP nonspecific competitor (NS). (B) Gel mobility complexes were formed with neonatal liver extract or recombinant human Ku70/80 in the absence (-) or presence (+) of Mg²⁺. Extracts with Mg²⁺ were dialyzed in a similar fashion to those without Mg²⁺ to control for the effects of dialysis. (C) Human recombinant Ku70/80 was incubated with radiolabeled fragments from HS site 2 (lanes 1 and 2), mouse mammary tumor virus (lanes 3 and 4), the C3H element (lanes 5 and 6), c-myc (lanes 7 and 8), the human T-cell leukemia virus I (lanes 9 and 10), the immunoglobulin octamer motif (lanes 11 and 12), the U1 snRNA (lanes 13 and 14), the heat shock element (lanes 15 and 16), a mouse retroviral-like element (lanes 17 and 18), the T-cell receptor β (TCR-β) enhancer E3 (lanes 19 and 20), and the TCR-β enhancer E4 (lanes 21 and 22), with (+) or without (-) a 500-fold excess of unlabeled HS site 2. (D) Human recombinant Ku70/80 was incubated with radiolabeled HS site 2 in the absence (lane 1) or presence of a 50- or 500-fold excess of unlabeled fragments derivedfrom HS site 2 (lanes 2 and 3), an AFP nonspecific fragment (lanes 4 and 5), mouse mammary tumor virus (lanes 6 and 7), the C3H element (lanes 8 and 9), c-myc (lanes 10 and 11), the human T-cell leukemia virus I (lanes 12 and 13), the immunoglobulin octamer motif (lanes 14 and 15), the U1 snRNA (lanes 16 and 17), the heat shock element (lanes 18 and 19), a mouse retrovirus-like element (lanes 20 and 21), the TCR-β enhancer E3 (lanes 22 and 23), and the TCR-β enhancer E4 (lanes 24 and 25).

The sequence-nonspecific binding of the Ku heterodimer to free DNA ends does not require Mg²⁺. However, the complex has been shown to translocate along DNA, and this activity requires Mg²⁺ (20). To test whether the specific binding of Ku70/80 to the ATCGAT binding site involves translocation, we removed Mg²⁺ from the liver extract and the recombinant protein by dialysis, and in both cases the ATCGAT binding activity was abolished (Fig. 3B). This suggests that the specific binding to ATCGAT requires translocation to the site from a free DNA end rather than direct binding.

Relative specificity of DNA binding by Ku70/80.

A number of other sequence-specific Ku70/80 binding sites have been reported in the literature (14, 19, 42). Some of these have been shown to require free DNA ends and thus are thought to be strong internal pause sites, while others have been proposed to be sites of direct binding, based on the ability of Ku70/80 to bind to the site on circular DNA (19, 20). To compare the relative affinity of Ku70/80 for the ATCGAT binding site versus that of other sequences, we monitored the ability of sequences in both of these classes to compete for binding with the ATCGAT-containing probe (HS2). The best-characterized example of the direct binding class is the NRE1 element of the mouse mammary tumor virus long terminal repeat (20), a tandem repeat of GAAAGA that bears no sequence similarity to the ATCGAT sequence. As shown in Fig. 3C, lanes 3 and 4, an oligonucleotide containing the NRE1 sequence binds to Ku70/80, but the binding is effectively competed by HS site 2. Likewise, ATCGAT competes very effectively for pause sites, as illustrated in lanes 13 and 14 by the binding of Ku70/80 to an element in the U1 snRNA promoter. Indeed, of the 10 sequences tested, all but one bound Ku70/80 and was effectively competed with HS site 2. In contrast, none of the sequences competed with HS site 2 for binding, even at 500-fold excess (Fig. 3D). Thus, under the conditions of our assay, HS site 2 displays the strongest affinity for Ku70/80 of the sequence-specific binding sites identified to date.

Genome distribution of Ku binding sites.

Given the observation that ATCGAT is a high-affinity site for Ku70/80, we used BLAST (1) to determine the distribution of this 6-mer in the mouse genome. We found that the sequence is found 104,497 times, which is very close to the expectation of 109,997 times based on the dinucleotide composition of the genome. The LINE-1 retrotransposon, which represents approximately 19% of the mouse genome, accounts for 26% of the ATCGAT hexamers in the genome. In this respect, the LINE-1 element is unique, as other classes of mouse repeats, with the exception of the IAPLTR1a I MM subfamily of IAP retrotransposons, did not contain this site (Table 2) (27). Within the 6.4-kb sequence of the consensus LINE-1 retrotransposon, ATCGAT is found twice, 1,233 bp apart.

TABLE 2.

Distribution of ATCGAT sites in mouse repeats^a

Mouse repeat	Repeat class	No. of repeats		No. of ATCGAT sites		Avg no. of sites
		Recent	Diverged	Recent	Diverged	Recent	Diverged
B1	SINE B1	61,301	19,160	32	7	0	0
MTA	MaLR LTR	12,524	6,602	8	9	0	0
L1_MM	LINE1	6,052	10,026	13,040	14,751	2.15	1.47
ORR1A0	MaLR LTR	3,102	6,505	1	3	0	0
MTB	MaLR LTR	1,897	16,434	9	25	0	0
MTAI	MaLR LTR	1,666	1,117	13	12	0.01	0.01
MERVL_LTR	MERVL LTR	1,493	872	9	1	0.01	0
IAPLTR1a_MM	LTR	993	3,390	0	0	0	0
IAPEZI	LTR	726	1,008	1	1	0	0
PMR89	PMR	719	129	0	0	0	0
IAPLTR1a_I_MM	LTR	707	1,014	243	191	0.34	0.19
IAPEY2_LTR	LTR	589	1,197	0	0	0	0
B2	SINE B2	574	47,316	6	169	0.01	0
BC1	SINE	560	732	0	0	0	0
RLTRETN_MM	LTR	532	459	0	0	0	0
MUSID6	SINE	521	331	0	0	0	0
RLTR10C	LTR	443	1,211	0	4	0	0
LTRIS	LTR	420	484	0	1	0	0
MERVL	MERVL LTR	412	291	2	0	0	0
ORR1A	MaLR LTR	391	15,683	1	16	0	0
RLTR10A	LTR	380	1,761	0	0	0	0
MUSID2	SINE	347	1,777	0	0	0	0
LTRIS_Mus	LTR	316	720	1	0	0	0
LTRIS2	LTR	259	889	1	0	0	0
IAPEY3_LTR	LTR	255	1,443	0	0	0	0
RLTR13C2	LTR	251	564	4	9	0.02	0.02
RLTR1C_MM	LTR	242	390	0	0	0	0
LTRIAPEY	LTR	225	424	0	3	0	0.01
RLTR13B1	LTR	219	460	4	6	0.02	0.01
ORR1AI	LTR	201	450	3	11	0.01	0.02
IAPA_MM	LTR	199	228	0	0	0	0
RLTR13B2	LTR	172	229	0	1	0	0
RLTR1B_MM	LTR	161	606	0	0	0	0
RLTR13B3	LTR	125	164	0	0	0	0

^aBLAST (1) was used to find all occurrences of known mouse repeats (27) in the mouse genome. We divided these matches into more-recently duplicated repeats (>60% sequence identity to the consensus) and more-ancient repeats (30 to 60% sequence identity). All repeats with more than 200 recent matches by this criterion are shown in Table 1.

In vivo binding of Ku to ATCGAT binding sites.

To test whether Ku70/80 binds in vivo to the ATCGAT sequences in the H19 ICR, we performed ChIP assays with adult liver extracts (Fig. 4). Due to a lack of confirmed in vivo binding sites for Ku in the mouse, we used the two CTCF sites in HS1 as control. In addition, we employed multiple Ku antibodies to decrease the possibility of an antibody-specific failure in the immunoprecipitation. Primer set A amplifies a region containing the most-5′ CTCF binding site and an ATCGAT binding site, while primer set B contains only a CTCF binding site. Using commercially available antibodies to CTCF and Ku70/80, we observed enrichment of both CTCF sites, relative to a control with no antibody (Fig. 4). In contrast, no enrichment was observed with either of two Ku antibodies tested. While the lack of a true positive control must be considered, this is consistent with the requirement of translocation from a free DNA end rather than direct binding.

FIG. 4.

ChIP of the ICR. Primers derived from HS1 were used to amplify DNA that was immunoprecipitated from chromatin using antibodies to CTCF and Ku70/80. Primer set A contains both a CTCF binding site (filled circle) and an ATCGAT binding site (open circle), while primer set B contains only a CTCF binding site. Each set of bars depicts the copy number, relative to CTCF, of the amplified DNA immunoprecipitated with CTCF antibody (Upstate Biotechnology), no antibody, or Ku antibodies purchased from Santa Cruz (SC) and Neomarkers (N).

Mutating the ATCGAT binding sites in mice.

Although the ChIP assays argue that Ku70/80 does not bind, at least to the site in HS1, in somatic tissues, it is possible that the sites play a role in imprinting at some other time in development or in another tissue. To explore this possibility, we used site-directed mutagenesis to change the ATCGAT sites to CACGCA, which eliminated Ku70/80 binding in vitro without changing the CpG content of the ICR. The modified fragment was incorporated into a targeting vector that had been used previously to modify the ICR in mice (43).

Two correctly targeted ES cell lines were identified and confirmed by Southern analysis with DNA probes that lie outside the vector (Fig. 5A). In addition, since the mutations eliminate ClaI recognition sites, the presence of the mutations could be verified by performing PCR followed by digestion with ClaI (data not shown). The two correctly targeted ES cell lines were used to generate mice, and the neomycin selection cassette was excised upon transmission from the male chimeras due to the expression of cre recombinase in testes (Fig. 5A). The resulting mice were designated H19^{Ku site−}.

FIG. 5.

Effects of mutating the Ku70/80 binding sites in vivo. (A) Targeting strategy for substituting mutations in HS site 1 and 2 into the wild-type ICR (WT). The structures of the modified chromosome before (targeted) and after germ line transmission in mice are indicated. The vector contains the mutant Ku70/80 sites (white X), the cassette containing the neomycin resistance gene, and cre recombinase flanked by loxP sites (triangles), along with flanking DNA. Below, genomic DNA from ES cells (lanes 1 to 4) or F₁ progeny (lanes 5 and 6) was subjected to Southern analysis. The 5′ external probe (3.7-kb BamHI/XbaI fragment) was used with a Bstz171/EcoRV digest. The wild-type allele generates a 12.5-kb band, while the targeted allele generates a 7.7-kb band. The 3′ external probe (1.8-kb HindIII/SmaI fragment) was used with a SpeI digest. The wild-type allele generates a 22.3-kb band, while the targeted allele generates an 8.1-kb band. Analysis of the recombination between loxP sites to remove the neomycin selection cassette was done with the 3′ external probe on a SpeIdigest. The targeted allele generates an 8.1-kb band, while the floxed allele generates a 22.3-kb band. (B) Methylation analysis was performed by Southern blotting on genomic DNA isolated from livers of adult C57BL/6J (WT) and H19^{Ku site−} (D^M) mice. The DNA was digested with StuI, Bstz171, and BamHI (SBB) in the absence and presence of the enzyme indicated above the lanes and probed with a 1.8-kb-ClaI/BglII fragment containing the ICR (the relevant restriction sites are shown below). The relevant enzymes and probe are shown below relative to the H19 start of transcription. Lanes 1 to 5, C57BL/6J DNA (WT); lanes 6 to 10, maternal inheritance of H19^{Ku site−}; lanes 11 to 15, paternal inheritance. M, maternal inheritance; P, paternal inheritance. (C) The allele-specific expression levels of H19 and Igf2 mRNAs in neonatal liver of progeny of H19^{Ku site−} females crossed to M. castaneus males (D^M × C) and H19^{Ku site−} males crossed to M. castaneus females (C × D^M) were assessed using allele-specific RNase protection assays (26). These were compared to reciprocal crosses between C57BL/6J (D) and M. castaneus (C) mice. D and C arrows indicate the migration of the protected fragments for the C57BL/6J and M. castaneus bands, respectively.

Effect of mutating the Ku70/80 binding sites.

To assess the consequences of eliminating the Ku70/80 binding sites on the parental-specific methylation of the ICR, Southern analysis was performed using the methylation-sensitive and insensitive isoschizomers HpaII and MspI, respectively, as well as the methylation-sensitive enzymes HhaI and BstUI. Genomic DNAs were first digested with StuI, Bstz171, and BamHI, which allowed the mutated allele to be distinguished from the wild-type allele as the result of the introduction of a Bstz171 site at the 5′ end of the ICR. As shown in Fig. 5B, the mutations of the Ku70/80 sites did not prevent the methylation-sensitive enzymes from digesting the unmethylated maternal allele (M) in the middle panel. Likewise, the mutant paternal allele (P) in the third panel was resistant to digestion, indicating that it had retained its fully methylated identity.

To determine whether imprinting was affected by the mutations, we examined the expression of both H19 and Igf2 in reciprocal crosses between Mus domesticus mice carrying the H19^Ku site− mutation and Mus castaneus mice. Figure 5C illustrates that the M. domesticus-derived H19 mRNA was transcribed from the mutant maternally inherited chromosome at levels comparable to a wild-type chromosome (lanes 1 and 2) and was appropriately silenced on the paternal chromosome (lanes 3 and 4). Likewise, the expression of the M. domesticus Igf2 gene displayed normal levels of expression on the paternal mutant chromosome (lanes 7 and 8) and no expression from the maternally inherited mutant chromosome (lanes 5 and 6). Thus, we could find no evidence for a role for the Ku70/80 binding sites in the establishment or maintenance of imprinting at the H19/Igf2 locus.

DISCUSSION

The molecular basis for the establishment and maintenance of parental-specific differences in the expression of the H19 and Igf2 genes has been the subject of intense study for the last decade. To date, a single protein, CTCF, has been convincingly shown to play an essential role in the soma, where it mediates the activity of a chromatin boundary on the maternal allele (3, 24). The evidence is less clear that CTCF plays a role in the germ line to establish the unmethylated state of the ICR. Mutations in the four well-characterized CTCF sites within the ICR have no effect on its unmethylated state in oocytes (40, 43), yet RNA interference inhibition of CTCF expression in oocytes leads to the inappropriate accumulation of methylation on the ICR (16). Further experiments will be necessary to reconcile this paradox. On the other hand, CTCF appears to play no role in the acquisition of methylation and its maintenance, on the paternal chromosome, suggesting that additional factors remain to be identified (16, 43).

This study was initiated in order to identify additional factors that might play a role in the complex regulation of these imprinted genes. In vivo footprinting findings had suggested that such factors were present on the maternal ICR in the soma (44, 45). Using extracts prepared from neonatal liver as well as other tissues, we identified a ubiquitous protein complex that bound specifically to two identical binding sites in HS1 and HS2. Upon purification these proteins were identified as the heterodimer composed of Ku70 and Ku80. The two Ku70/80 sites in the H19 ICR do not appear to play an important role in either the establishment or maintenance of imprinting, based on our finding that mutations that eliminate in vitro binding have no physiological effect in vivo. This is consistent with chromatin immunoprecipitation results that suggested that Ku70/80 is not bound to the sites, at least in adult liver.

The binding site we have identified is a novel example of a high-affinity and sequence-specific Ku70/80 pause site. Ku70 and Ku80 are highly conserved proteins that play a central role in repairing double-strand breaks by the nonhomologous end-joining pathway and in V(D)J recombination (13, 21-23, 38, 47, 52). The complex has also been implicated in genome surveillance, the maintenance of telomeres, DNA replication, and the regulation of transcription (14). In some, but not all, of these processes Ku70/80 associates with the catalytic subunit of DNA-dependent protein kinase.

The role of Ku in nonhomologous recombination is thought to occur via its ability to bind with high affinity to free DNA ends and recruit DNA-dependent protein kinase and the rest of the complex required for end joining and repair (13, 21). Ku's affinity for free ends is rationalized by the crystal structure of the Ku complex, which predicts a preformed circle that can surround the DNA end and, in the presence of Mg²⁺, translocate along it (49). The C terminus of Ku70 contains a SAP DNA binding domain that packs against the base of Ku80 in the absence of DNA. However, when the crystal structure was solved with DNA, the SAP domain was disordered and in close apposition to the DNA (49). This suggests that during translocation, the SAP domain may be in a position to make contacts with individual base pairs. Our finding that the ATCGAT binding activity is dependent upon Mg²⁺ supports the view that the ATCGAT binding site is a high-affinity pause site.

Given the observation that ATCGAT is a high-affinity site for Ku70/80, we used BLAST to determine the distribution of this site in the mouse genome. We found that the ATCGAT sites are frequently found in LINE-1 retrotransposons. A role for Ku70/80 in transposition has been proposed for the P transposable element (P element) in Drosophila melanogaster (2, 37). P elements transpose through a cut-and-paste mechanism whereby the P element leaves its mobilization site and inserts itself into the new position as a double-stranded DNA molecule (30). Transposition requires the presence of 31-bp inverted repeats that flank the mobilized DNA and are also left behind at the site of mobilization (37). These inverted repeats contain sequence-specific binding sites for the Drosophila homolog of Ku70/80 (2). The repair of double-strand breaks after P element excision was severely affected in a strain with a mutation in the Ku70 gene (2). Thus, it is likely that Ku70 plays a role in NHEJ at the site of mobilization as well as the site of insertion.

In contrast, LINE-1 transposition in the mouse occurs through a copy-and-paste model where a single-stranded RNA copy of the element is transcribed at the mobilization site (34). This single-stranded RNA encodes an endonuclease-reverse transcriptase that creates a nick at the insertion site and uses the exposed 3′ end to prime the synthesis of cDNA (17, 34, 36). Following the conversion of the cDNA to double-stranded DNA, a double-strand break is generated at the insertion site and the ends are joined, likely through the NHEJ pathway (34). It is at this step that the Ku70/80 binding sites in the LINE-1 element might play a role by sequestering the complex on the double-stranded DNA before end joining.

Once ligation has occurred, the Ku70/80 circular complex is trapped on the DNA (39). This complex, which would only occur after nonhomologous recombination, could serve as a signal to the cell that a new transposition event has occurred. Retrotransposons threaten the host genome by increasing the frequency of rearrangements via homologous recombination between nonallelic repeats, production of chimeric transcripts that combine retrotransposon and cellular sequences, and transcriptional interference from transcripts originating from retrotransposon promoters (12). In response, the mammalian genome attempts to shut down the process by methylating retrotransposon sequences (4, 50). In this cycle, it is crucial that the genome be able to detect new retrotranspositions so that they can be inactivated by methylation. It is possible that the binding of Ku70/80 to internal ATCGAT pause sites serves as such a mark.