A differentially methylated region within the gene Kcnq1 functions as an imprinted promoter and silencer

Abstract

The imprinted gene cluster on mouse distal chromosome 7 contains a differentially methylated CpG island that maps within the Kcnq1 gene that has been shown to be required for the imprinting of multiple genes. To evaluate models for how this imprinting control region (ICR) regulates imprinting, we have characterized it structurally and functionally. We show that the region contains a promoter for a paternally expressed anti-sense transcript, Kcnq1ot1, and we define the extent of the minimal promoter. We describe three paternal-specific nuclease hypersensitive sites immediately upstream from the start site and show that they are required for full promoter activity. The expression of Kcnq1ot1 during pre- and postnatal development is compared to that of other imprinted genes in its vicinity, Cdnkn1c and Kcnq1. The lack of coordination in their expression tends to rule out an enhancer competition model for the action of the ICR in imprinting control. Using a stable transfection assay we show that the region contains a position-independent and orientation-independent silencer. We propose, on the basis of these findings, that the Kcnq1 ICR functions as a silencer on the paternal chromosome to effect the repression of neighboring genes.

INTRODUCTION

Genomic imprinting is an epigenetic process by which the expression of a gene is governed by its parent of origin. Of the more than 35 imprinted genes identified in mammals, many map in close proximity to other imprinted genes (see www.mgu.har.mrc.ac.uk/research/imprinted/imprin.html). This clustering has been attributed to the action of imprinting control regions (ICRs) within each cluster that coordinate the imprinting of multiple genes, often at distances greater than 100 kb.

One of the most intensively studied imprinted gene clusters resides within an 800 kb region on the distal end of mouse chromosome 7 and the homologous region on human chromosome 11p15.5. It contains at least nine imprinted genes, including the maternally expressed Impt1, Ipl, Cdkn1c (p57^Kip2), Kcnq1 (Kvlqt1), Ascl2 (Mash2) and H19 genes and the paternally expressed Igf2 and Ins2 genes (Fig. 1A). The Ins2, Igf2 and H19 genes are coordinately imprinted through an ICR located immediately 5′ of H19 (1,2). On the maternal chromosome, the ICR acts as a chromatin enhancer blocker to prevent the interaction between Igf2 and enhancers that lie 3′ of the ICR (3–6). On the paternal chromosome, allele-specific methylation of the ICR suppresses both H19 transcription and enhancer blocking activity, thus enabling Igf2 transcription. The H19 ICR, however, has no role in the imprinting of the more telomeric Ascl2, Kcnq1 and Cdkn1c genes (7), suggesting that an additional ICR is present in the region.

A plausible candidate for that ICR was a ∼2 kb CpG island, the Kcnq1-associated differentially methylated region (DMR), within intron 10 of Kcnq1 (Fig. 1A and C) (8–10). This region is methylated in oocytes, but not in sperm, a germline-specific pattern that is maintained after fertilization in all somatic cells (11). It is also associated with a large non-coding RNA that is transcribed exclusively from the paternal allele in the opposite orientation to Kcnq1 in both humans (KCNQ1OT1 or LIT1) (8,9) and mice (Kcnq1ot1 or Lit1) (10). Mice carrying a targeted translocation between the CpG island and Cdkn1c display loss of expression and imprinting of Cdkn1c, while the imprinting of the genes centromeric to the breakpoint that remain contiguous with the CpG island are not affected (12). Horike et al. (13) have recently generated a deletion of the CpG island on a paternally inherited human chromosome in human : chicken hybrid cells, and the deletion increases the expression of both CDKN1C and KCNQ1, implying a loss of imprinting. More recently, Fitzpatrick et al. (14) and Steele et al. (manuscript in preparation), have demonstrated that the CpG island is required for the imprinting of multiple linked genes.

In this report we have characterized the sequences surrounding the Kcnq1 ICR. We have identified the minimal promoter and transcription start site for the Kcnq1ot1 gene, and mapped three paternal-specific regions of DNase I hypersensitivity in chromatin immediately 5′ of the site. We have determined the temporal and spatial expression pattern of Kcnq1ot1 and have compared it to that of Kcnq1 and Cdkn1c. Finally we show that the region does not function as an enhancer blocker in vitro, but rather harbors a strong silencer. We suggest that the mechanism by which the Kcnq1 ICR acts to regulate imprinting may be distinct from the mechanism uncovered at the H19 ICR.

RESULTS

Parental-specific nuclease hypersensitivity in the Kcnq1 ICR

Although a large CpG island is present in the 10th intron of Kcnq1 in both humans and mice, its primary sequence is poorly conserved (Fig. 1B and C). However, within both the mouse and human CpG islands is a number of conserved 30 bp repeats (with 6 bp sub-repeats) that map just downstream of a single mouse expressed sequence tag (EST) (Fig. 1D).

Recently, Kanduri et al. (15) have shown that the CpG region is hypersensitive to nuclease digestion in chromatin. To test whether the hypersensitivity is allele-specific, as has been shown for the H19 ICR, we prepared nuclei from neonatal liver and incubated with DNase I for increasing lengths of time. Following purification, the DNA was digested to completion with EcoRI or EcoRI in combination with the methylation-sensitive enzyme EagI. Because the maternal chromosome is differentially methylated at two EagI sites within the EcoRI fragment (10 and data not shown), it is resistant to EagI digestion, and thus we can use this resistance to distinguish between the parental chromosomes. As shown in Figure 2, using a probe that lies at the 3′ end of the EcoRI fragment, we detected three strong hypersensitive sites, resulting in bands that are 1.7, 1.6 and 1.5 kb in size. Digestion with EagI fully reduced these fragments to the expected sizes of 1.2, 1.1 and 1.0 kb, indicating that they were generated on the unmethylated paternal chromosome. Further analysis with probes at the 5′ end of the EcoRI fragment confirmed these results, and positioned the hypersensitive sites just 5′ of the EST (data not shown). These hypersensitive sites were also identified in samples from brain nuclei (data not shown).

Functional analysis of the Kcnq1ot1 promoter

To ask whether the hypersensitive sites coincide with the promoter for the Kcnq1ot1 transcript, we tested several overlapping fragments for promoter activity using the luciferase-based pGL3-Basic vector (Promega), which lacks both promoter and enhancer sequences. The fragments tested are depicted in Figure 3A, together with the relative luciferase activities in transient transfection assays into Hep3B cells. The mean activities of the test constructs were calculated relative to the pGL3-Basic vector without an insert, designated as 1 relative light unit (RLU).

pGL2 and pGL3, encompassing ∼600 bp of the region immediately upstream of the Kcnq1ot1 CpG island and containing the three hypersensitive sites (nt 101453–102050; GenBank accession number AP001295) displayed the highest promoter activity. This activity was greatly reduced when the most distal hypersensitive site was deleted (pGL1) and modestly reduced in the absence of the two most proximal hypersensitive sites (pGL4). This activity was orientation-dependent, as shown by the lack of promoter activities of the pGL3-rev and pGL4-rev fragments cloned in the reverse orientation. Thus the region containing the hypersensitive sites within the Kcnq1 CpG island is critical for promoter activity. The conserved repeated sequences have no promoter activity on their own, based on the low activity of pGL5, which contains five copies of the repeats, and their removal from pGL3 (resulting in pGL2) had no effect as well.

A DNA sequence homology search within the ∼600 bp minimum promoter using the transcription factor database, TFSEARCH (www.cbrc.jp/research/db/TFSEARCH.html) revealed the presence of a putative CREB binding site, three putative CCAAT boxes, three NF-Y (CCAAT-binding) sites and a putative TATA box (data not shown).

Identification of the Kcnq1ot1 transcription start site

We used RNase protection analysis to precisely position the transcription start site of the Kcnqt1ot1 transcript. As shown in Figure 4A, we tested a series of probes spanning much of the Kcnq1ot1 CpG island. Figure 4B shows that RPA1 and RPA2 were completely protected from RNase digestion, indicating that they were contained entirely within the transcript, whereas probe RPA5 was 5′ of the start site, as suggested from its complete digestion. The sizes of two probes, RPA3 and RPA4, were reduced upon treatment with RNase, indicating that they overlapped the start site. Based on the sizes of the protected bands, we mapped the start site 40 bp downstream of HS3, and 20 bp upstream of the 5′ end of the mouse Kcnq1ot1 EST.

The expression of Kcnqt1ot1

The precise role of the Kcnq1 ICR in the imprinting of genes at the distal end of mouse chromosome 7 is unknown. It is possible that the DMR functions like the H19 ICR as an enhancer blocker to modulate the access of genes to enhancers distributed on either side of the DMR (Fig. 7A). In that case, the promoter activity and the Kcnq1ot1 transcript may not be relevant to the activity of the DMR in imprinting. Alternatively, as was suggested initially for the H19 ICR, the DMR may compete with imprinted genes for shared enhancers (Fig. 7B), in which case the expression of the Kcnq1ot1 transcript should coincide with the expression of the imprinted genes. Thirdly, it has recently been shown that non-coding transcripts may play a role in both imprinting and in X chromosome inactivation (Fig. 7C), as has been shown for the anti-sense transcript at the imprinted Igf2r locus (16). In that case, one would again expect coincident expression of the transcript and the imprinted genes. Finally, the Kcnq1 ICR may function as a silencer element that establishes a repressive domain propagating bi-directionally along DNA over long distances. This element could serve to actively silence both promoter and enhancer regions 5′ and 3′ of the DMR (Fig. 7D). In this last model, the promoter activity and the Kcnq1ot1 transcript may not play important roles in the activity of the DMR in imprinting.

To begin to consider whether one of these models pertains to the Kcnq1 ICR, we characterized the tissue-specific and temporal pattern of Kcnq1ot1 expression. It has been previously demonstrated by RT–PCR that KCNQ1OT1 is expressed in a variety of human fetal tissues, such as liver, kidney, heart, spleen, cerebrum, thymus, muscle and lung (8,10). Furthermore, it was determined in mouse that the anti-sense transcript was expressed paternally in e14.5 lung, gut, heart and kidney (10). In Figure 5 we employed RNase protection analysis to compare the expression pattern of Kcnq1ot1 in various neonatal and adult mouse tissues with that of Kcnq1 and Cdkn1c. While Cdkn1c is imprinted in all tissues and developmental stages examined, Kcnq1 is only imprinted in mouse during the earliest stages of fetal development, and is bi-allelic by e14.5 (17–19).

The data illustrate that Kcnq1ot1 is broadly expressed in both neonatal and adult tissues, whereas the expression of Cdkn1c is highest in neonatal brain, kidney and lung, and placenta, with lower levels in adult kidney. Kcnq1 expression, on the other hand, overlaps with that of Cdkn1c in neonatal brain, and kidney, but it is absent in placenta. In neonatal brain and adult kidney, the Kcnq1 transcripts are of a different size than those found in the other tissues. These bands probably represent isoforms of the Kcnq1 transcript (20). The patterns of expression of the three genes are not consistent with models that require coordinate transcriptional regulation through shared enhancers. Indeed Cdkn1c and Kcnq1 display patterns that are typical for tissue-specifically regulated genes, while Kcnq1ot1 behaves like a constitutively expressed gene. These data would tend to render unlikely models involving enhancer competition. The data do not rule out the transcriptional-dependent silencing model, however, as that model only requires that Kcnq1ot1 be transcribed wherever the genes are imprinted.

Enhancer blocking assays

To determine if the Kcnq1 ICR functions by enhancer blocking, as has been shown for the H19 ICR, we tested multiple fragments of the ICR in a stable transformation assay. This well-established assay tests the ability of a sequence to block activation of a neomycin resistance reporter gene by a strong enhancer in an erythroleukemia cell line (3,21). In the first set of experiments, we tested for enhancer blocking by placing fragments of the Kcnq1 ICR between the neo gene driven by the mouse globin promoter and its enhancer (‘blocking position’ in Fig. 6A and B). As a control, the H19 ICR reliably gave a 10-fold decrease in the number of neo-resistant colonies compared to the pNI vector without an insert (H19 ICR-B in Fig. 6B) (3). Insertion of the 1.9 kb Full Kcnq1 ICR fragment in the blocking position in either orientation resulted in an approximately 20-fold decline in colony number.

To assess whether the Kcnq1 ICR was functioning as an enhancer blocker or a silencer, we then tested the Full fragment in the ‘silencing position’, in which it is no longer between the enhancer and the gene (Fig. 6C). In that position the Full fragment caused an equally dramatic decrease in colony number in both orientations. Thus in this system the Kcnq1 ICR is acting as a silencer, not an enhancer blocker.

One explanation for the silencing observed in this assay is the presence of the promoter within the full fragment, which could be competing with or occluding the globin promoter of the neo gene. To help address this issue, we tested the ability of the SV40 promoter to reduce colony numbers in the ‘silencing position’. This construct resulted in a minimal decrease in colony number, indicating that a strong promoter, in this case the SV40 promoter, is not sufficient to elicit silencing activity in this assay. This is further supported by comparison of the HSS and HS fragments in the luciferase promoter assay (Fig. 3B). While both display strong promoter activity, there is a significant difference in their silencing activity (Fig. 6C). This would also tend to suggest that the promoter and silencer are separable.

To localize the silencing activity within the Full fragment, we generated a series of additional constructs that contained subsets of the ICR, and tested them in both the silencing and blocking positions. Neither the hypersensitive sites (HS) nor the CpG island alone had any effect on either silencing or blocking. The region overlapping the conserved repeats led to a 50% reduction in the silencing position, but had no effect in the blocking position. When a fragment containing both the hypersensitive sites and the transcriptional start site (HSS) was tested in the silencing position, it displayed strong silencing activity. These results suggest that the hypersensitive sites together with the transcription start site are sufficient for silencing activity.

DISCUSSION

Previous experiments had established that the H19 ICR was not required for the imprinting of the Ascl2, Kcnq1 and Cdkn1c genes, which lie ∼300–800 kb telomeric to the H19/Igf2 locus (7), and suggested that at least one more ICR resided within the imprinting domain on mouse chromosome 7. The identification of a conserved differentially methylated CpG island within intron 10 of Kcnq1 in both human (8) and mouse (10) and the demonstration that it acquired its differential methylation in the germline (11), made it an excellent candidate for an ICR.

Several observations of the DMR in patients with Beckwith–Wiedemann syndrome (BWS), a congenital disorder associated with fetal overgrowth, macroglossia and anterior abdominal defects (22), have supported the idea that the DMR has an important role in imprinting in humans. BWS can result from either paternal duplications of 11p15.5, maternally inherited mutations in CDKN1C, paternal uniparental disomy (UPD), or maternal chromosome rearrangements and translocations (17,23–25). A rare subset of patients have maternally inherited chromosomal rearrangements that contain breakpoints within the KCNQ1 gene (20). Although it has not been experimentally verified, it is thought that the disease phenotype results from the loss of maternal CDKN1C expression rather than a reduction in KCNQ1, a conclusion that implicates sequences within KCNQ1 in CDKN1C expression and imprinting. This supposition has been greatly strengthened by studies in mice that contain an engineered translocation similar to ones in BWS, where loss of Cdkn1c expression was confirmed (12). Furthermore, the frequent loss of maternal methylation at the KCNQ1 CpG island within BWS patients also points to the importance of maintaining the differential methylation of this region in the coordinate control of the imprinting cluster (9–11). Definitive evidence that the DMR regulates imprinting comes from analyses of mice in which the region was deleted (14, Steele et al., manuscript in preparation), and where the normal repression of multiple genes, including Kcnq1, Cdkn1c and Ascl2 was relieved on the paternal chromosome.

The goal of the present study was to functionally characterize the Kcnq1 DMR as a prelude to ascertaining its potential mechanism of action as an ICR. We show that the DMR contains within it a functional promoter that directs the transcription of a non-coding RNA, Kcnq1ot1. The promoter was localized to a region containing three strong paternal-specific DNase I hypersensitive sites, and we showed that the presence of those sites was required for full promoter activity. Nuclease hypersensitivity, associated with a variety of cis-acting regulatory elements, including promoters, enhancers and chromatin insulators, is thought to arise from specific protein–DNA interactions that disrupt the normal nucleosomal array. It is likely in this instance that the hypersensitive sites mark the position of protein binding required for the activity of the promoter. By using RNase protection we delineated the start of transcription, and showed that it is consistent with the functional mapping of the promoter and the 5′ end of a rare mouse EST for Kcnq1ot1.

There have been several models proposed to explain the action of ICRs. One model of ICR function proposed initially for the H19 ICR was enhancer competition between the maternally expressed H19 gene and the paternally expressed Igf2 gene (Fig. 7A) (26–28). This model was based on the remarkable similarities in the temporal and spatial expression patterns of the two genes, which suggested the possibility that they competed for the same enhancers (28–30). We compared the expression pattern of the Kcnq1ot1 and Cdkn1c genes during embryogenesis as well as within various neonatal and adult mouse tissues in order to explore the likelihood that the enhancer competition model could apply in this case. Our results demonstrate that the Kcnq1ot1 pattern of expression is distinct from that of Cdkn1c, the gene whose imprinting is maintained throughout pre- and post-natal development, unlike the other genes in the cluster whose imprinting is lost during embryogenesis. These findings do not lend any support to an enhancer competition model to explain the reciprocal imprinting of Cdkn1c and Kcnq1ot1 (Fig. 7B).

Another model that might be invoked to explain the action of the Kcnq1 ICR in imprinting comes from studies of the Xist and Igf2r genes, where non-coding RNAs have been implicated in long-range chromosomal silencing. In the latter case, it has been proposed that an anti-sense transcript within the Igf2r gene mediates repression of Igf2r in cis. The authors proposed a two-step model where the Air transcript directly silences the Igf2r promoter, thereby altering its chromatin configuration and leading to the spread of the silenced state to flanking genes (16). Such a model could also apply to the Kcnq1ot1 transcript (8–10), although several adaptations would be required to explain how the Kcnq1 gene loses its imprinting by e14.5, whereas the Cdkn1c gene is constitutively imprinted. It is conceivable, for example, that once it has been established early in development, the silent chromatin state induced by the anti-sense transcript is propagated faithfully at Cdnk1c, while other regulatory elements act to overcome the silencing at Kcnq1.

The most well studied ICR is at the H19 locus, which has been shown to function as a strong enhancer blocker both in tissue culture assays and in vivo (3–5). However, our findings for the Kcnq1 ICR are not consistent with this model. The enhancer blocker model makes strong predictions about the positions of enhancers relative to genes, such that enhancers required for Cdkn1c expression must lie on one side of the ICR, while those for Kcnq1 and Ascl2 must lie on the other (Fig. 7A). Thus it is difficult to reconcile this model with recent findings of enhancers that act in tissues relevant to Cdkn1c expression between the gene and the ICR (31). If those enhancers are required for Cdkn1c expression in vivo, their position cannot be reconciled with the enhancer blocker model.

Our findings contradict those of Kanduri et al. (15), who found that a 3.6 kb PstI fragment containing the Kcnq1 ICR displayed position-dependent and orientation-dependent enhancer blocking activity using two alternative in vitro assays. The authors concluded that the Kcnq1 ICR probably functions through an enhancer blocking model to regulate genes in the cluster. The basis for this discrepancy is unclear at this point. There are several differences in the experimental approaches, including the constructs, cell lines utilized and the sizes of the ICR fragments tested. In their experiments, they tested a larger 3.6 kb fragment, which may have led to different results in the silencer assay.

Our analysis indicates that the Kcnq1 ICR functions on the paternal chromosome as an orientation-independent silencer, not an enhancer blocker (Fig. 7D). The silencing activity of the Kcnq1 ICR was localized to the hypersensitive sites and the transcription start site. Silencers are regulatory elements that act in a position and orientation-independent manner to silence genes. It has been proposed that silencer elements recruit protein complexes to set-up a repressive chromatin domain (32–34). This repressive domain may propagate bi-directionally along DNA over long distances, actively silencing both promoter and enhancer regions until it encounters a discrete boundary.

It is presently unclear how the Kcnqot1 transcript itself might escape this silencing. There are several examples of genes escaping repressive heterochromatin. In the case of X-inactivation, several genes on the X chromosome escape inactivation and are expressed on both the active and inactive X chromosome (35–38). Furthermore, in Drosophila melanogaster, there are well studied examples of genes whose expression actually requires that they are situated within heterochromatin (39). There may be features intrinsic to promoters and the proteins that bind to them that allow them to escape this silencing. A silencer on the paternal allele at the Kcnq1 ICR is a very attractive model, as it satisfies several requirements that the boundary model fails to meet. Most notably, it sets no restrictions on the positions of enhancers relative to the imprinted genes, and thus allows them to be regulated independently of one another.

METHODS

Transcription assays

Reporter constructs (pGL3, 3-rev, 4, 4-rev, 5) were generated by PCR using primers designed with SacI and HindIII restriction sites at their ends. The PCR primers are listed in Table 1. A proofreading Taq (Platinum Taq; Gibco-BRL) was used to minimize the occurrence of amplification errors. PCR amplification took place over 35 cycles, under the following conditions: 94°C for 30 s, 55°C for 58 s, and 72°C for 1 min. Constructs pGL1 and 2 were generated by restriction enzyme digestion of the region. All Kcnq1ot1 promoter fragments were ligated into the pGL3-Basic luciferase reporter vector (Promega), which lacks both promoter and enhancer sequences. Following transformation, the inserts were sequenced to confirm the fidelity of the amplification.

Hep3B cells were grown in Dulbecco's modified Eagle's medium (DMEM) supplemented with penicillin (100 µg/ml), streptomycin (100 µg/ml) and 10% fetal bovine serum. Transfections were carried out in six-well 35 mm² tissue culture plates in triplicate using 3 µl lipofectamine 2000 reagent (Gibco-BRL) and 1 µg of plasmid per well. Each plasmid was co-transfected with 1 µg of pRL-SV40 Renilla luciferase vector to control for differences in transfection efficiency. Cells were harvested 48 h after transfection, and cell lysates were assayed for both Firefly and Renilla luciferase activities using the protocols from the Dual-Luciferase Reporter Assay System (Promega). Mean activities of the test plasmids were standardized to the pRL-SV40 control, and expressed relative to pGL3-Basic vector activity (designated as 1 relative light unit or RLU).

Enhancer-blocking assays

All fragments of the Kcnq1 ICR were generated by PCR on cloned genomic DNA. Primer pairs were designed with either AscI or NdeI sites on the ends of each primer depending on whether the fragment was cloned into the blocking or silencing position, respectively (see Table 1). The 1.9 kb Full fragment was generated with primer 1 and primer 2. The 390 bp HS fragment was generated using primers 1 and 3. Fragment HSS has an additional 70 bp extending from the 3′ end of HS and was generated using primers 1 and 4. The 700 bp fragments containing the repeats and the remainder of the CpG island were generated using primers 5 and 6 and primers 2 and 7, respectively. An SV40 minimal promoter fragment of ∼200 bp was generated using the pGL3-Promoter Vector (Promega) as template for amplification with the primers 5′-CTGCATCTCAATTAGTCAGCAACC and 5′-CCAACAGTACCGGAATGCCAAG. The construct DMD-B contains the 1.6 kb H19 DMD fragment cloned into the blocking position and was generously provided by the Felsenfeld lab. The DMD fragment was subsequently subcloned by blunt-ended ligation into the silencing position to create DMD-S. All fragments were ligated into the pNI construct (3) carrying a neomycin resistance gene driven by the human ^Aγ-globin promoter and an enhancer derived from a mouse β-globin LCR (a gift of A.G. West and G. Felsenfeld). A 1.2 kb chicken insulator element was also present 3′ of the promoter in all constructs to block the effects of the enhancer on an adjacent tandemly integrated promoter (21,40). Following transformation into E.coli, the inserts were sequenced to confirm both orientation and the fidelity of the amplification.

Assays were performed as described (40) with the following modifications: 1.0 µg of DNA was used for the electroporation into the human erythroleukaemia cell line K562 (ATCC) and the cells were allowed to recover for 36 h prior to selection. The number of G418-resistant colonies was counted after 3 weeks of growth in soft agar.

DNase I hypersensitivity analysis

Nuclei were isolated from adult tissues as described by Bartolomei et al. (27) and washed once in RSB (50 mm Tris pH 8.0, 100 mm NaCl, 3 mm MgCl₂, 0.1 mm PMSF, 5 mm sodium butyrate), resuspended in RSB containing 0.1 mm CaCl₂ and 1 µg/ml DNase I (Worthington) and incubated at 37°C for the times indicated.

RNase protection assay

Total RNAs were prepared using Trizol (Sigma) from mouse neonatal and adult tissues. All RNase protection assays were carried out using the RPAIII kit (Ambion). To delineate the Kcnq1ot1 transcription start site, restriction fragments were subcloned into the pBluescript II KS+ vector, and radiolabeled anti-sense probes were synthesized using either T3 or T7 polymerase in the presence of [³²P]CTP at 37°C for 1 h. The probes were treated with RNase-free DNase for 15 min at 37°C and purified on an 8 m urea/5% polyacrylamide gel. Probes (20 000–80 000 cpm) were each hybridized to 10 µg of total RNA for at least 16 h at 42°C. Following treatment with ribonuclease, the samples were separated on a 7 m urea/7.5% polyacrylamide gel. The expression of Kcnq1 and Cdkn1c were assessed in an identical manner using probes 438 bp (nt 1095–1533; GenBank accession number NM008434) and 125 bp (nt 1357–1482; GenBank accession number U22399) in length, respectively.

ACKNOWLEDGEMENTS

We would like to thank B. K. Jones for her gift of RNAs used for the RNase protection analysis and advice on the DNase I hypersensitivity analysis; A. West and C. Schoenherr for advice on the enhancer blocking assays; M.A. Cleary and J.V. Schmidt for advice and reagents; and members of the Tilghman lab for their helpful comments and suggestions on the manuscript. D.M.D. is supported by the Cancer Research Fund of the Damon Runyon Foundation Fellowship, DRG no. 1675. S.J.S.S. is supported by the New Jersey Commission on Cancer Research. This work was supported by a grant from the NIH. S.M.T. was an investigator of the Howard Hughes Medical Institute.

†

The authors wish it to be known that, in their opinion, the first two authors should be regarded as joint First Authors.

Figure 1. The structure of the Kcnq1ICR. (A) The organization of the imprinted gene cluster on the distal end of mouse chromosome 7. Genes are depicted by boxes with maternally expressed genes in red, paternally expressed genes in blue and non-imprinted genes in white. The second line illustrates a 12.6 kb HindIII fragment within intron 10 of Kcnq1 that contains the Kcnq1ot1 ICR and promoter (horizontal arrow). Three DNase I hypersensitive sites are indicated as vertical arrows and conserved 30 bp repeats as vertical bars. (B) Dot plot comparison of the human and mouse genomic sequences spanning the HindIII fragment. (C) CpG dinucleotide distribution for the HindIII fragment. Numbers refer to those in accession no. U90095 (human) and no. AP001295 (mouse). (D) DNA sequence alignment of the 30 bp sequences within the human and mouse Kcnq1 ICRs. The conserved bases are highlighted.

Figure 2. Allele-specific DNase I hypersensitivity of the Kcnq1 ICR. (A) Liver nuclei were digested for increasing time with DNase I. Following purification, DNA was digested with EcoRI(R) or EcoRI and EagI (E) and subjected to Southern blot analysis using a 700 bp PCR fragment that detects a 6.4 kb EcoRI fragment. The R/Pv, R/Ba and R/Bm lanes contain liver DNA digested with EcoRI, and PvuII (Pv), BsaHI (Ba) or BsmBI (Bm) to serve as size markers. (B) The top line shows the 12.6 kb HindIII fragment, with the start of transcription indicated by the horizontal arrow and vertical bars representing the repeats. Below is the 6.4 kb EcoRI fragment, and the sizes of the smaller fragments generated by DNase I digestion. The vertical arrows indicate the positions of the hypersensitive sites. The position of the EagI sites that cleave the unmethylated paternal fragments and the position of the probe are indicated. R, EcoRI; E, EagI; P, PvuII; Ba, BsaHI; Bm, BsmBI.

Figure 3. Functional mapping of the Kcnq1ot1 promoter. The top line depicts the ICR, with the transcription start site indicated by the horizontal arrow and the DNase I hypersensitive sites indicated by vertical arrows. The location of repetitive elements are depicted as black vertical bars. (A) The positions of the several overlapping fragments derived from the Kcnq1ot1 ICR that were cloned into the pGL3-Basic luciferase reporter vector are displayed below. The level of luciferase activity generated by transient transfection of the constructs, as measured by quantity of luminescence relative to the vector without insert, is indicated to the right of each construct. (B) Some of the constructs used in the enhancer blocking assay (Fig. 6) were placed in the basic-pGL3 vector and tested for luciferase activity. The resulting levels of luciferase activity are indicated to the right of each construct. These constructs were tested at a different time than those described in (A), resulting in different absolute values in the assay.

Figure 4. Mapping the transcriptional start site of Kcnq1ot1. (A) The top line depicts the ICR, with the transcription start site indicated by the horizontal arrow and the 5′ end of the mouse EST indicated by the arrowhead. The DNase I hypersensitive sites are indicated by vertical arrows. The repetitive elements are depicted as black vertical bars. Below are depicted the probes used in RNase protection assays to determine the location of the Kcnq1ot1 transcription start site. The region of each probe protected upon RNase digestion is displayed in white. (B) Each probe was hybridized to neonatal mouse brain (B) and day 15.5 mouse embryo (E) total RNA. + lanes, no RNA control; − lanes, undigested probe.

Figure 5. Expression of Kcnq1ot1, Kcnq1 and Cdkn1c. RNase protection analysis of 10 µg of total RNAs were simultaneously performed with three probes: a 343 bp Kcnq1ot1 probe (102063–102406; GenBank accession number AP001295), a 438 bp Kcnq1 probe (1095–1533; GenBank accession number NM_008434) and a 125 bp Cdkn1c probe (1357–1482; GenBank accession number U22399). BN, neonatal brain; BA, adult brain; KN, neonatal kidney; KA, adult kidney; LN, neonatal liver; LA, adult liver; TA, adult testis; PL, placenta; d7–d17, total embryonic RNA from embryonic day 7 to embryonic day 17.

Figure 6. Silencing activity of the Kcnq1ot1 ICR. (A) The top line represents the 3.6 kb PstI fragment encompassing the ICR. Regions of note include the DNase I hypersensitive sites (HS and vertical arrows), repeats (vertical bars) and the Kcnqot1 transcription start site (horizontal arrow). The regions of the Kcnq1 ICR tested for activity are indicated below, along with the enhancer-blocking/silencing construct. The position of the enhancer is indicated as an oval (E). (B) Enhancer blocking and (C) silencing activities of fragments of the Kcnq1 ICR inserted at either position in constructs stably integrated into K562 cells. For each construct, colony number was normalized to no insert control, pNI. Data are the average of at least three independent assays.

Figure 7. Models to account for Kcnq1ot1 ICR Activity. (A) The enhancer blocker model—the DMR region behaves as a boundary on the paternal allele, blocking access of Cdkn1c to putative enhancer elements 3′ of Kcnq1ot1 and blocking access of Kcnq1 to putative enhancer elements 5′ of Kcnqot1. (B) Enhancer competition model—on the unmethylated paternal allele, the stronger Kcnq1ot1 promoter utilizes enhancer elements located at any position on the chromosome. The Kcnq1ot1 promoter is methylated on the maternal allele and is subsequently silenced. The Cdkn1c gene is thus able to utilize the enhancer elements, leading to the maternal expression of this gene. (C) Transcriptional-dependent silencing model—the production of the Kcnq1ot1 transcript is required to initiate silencing on the paternal allele. Silencing of Kcnq1 and the subsequent induction of a silent chromatin state (represented by grey ovals) could then lead to the loss of imprinting of other flanking genes, such as Cdkn1c and Ascl2. (D) Silencer model—the Kcnq ICR itself acts as a silencer, establishing a repressive chromatin domain that spreads bi-directionally from the ICR (represented by gray ovals).

Table 1.

List of PCR primers for plasmid construction

Primer	Sequence
Primers used in luciferase assay
Luc1-for	5′-GGAGGCTCTCCCTCCCCTTCCCCTAAATCCTG-3′
Luc2-for	5′-GGAGCTCGCTAGGAGGAACAGTTGCCTCAG-3′
Luc1-rev	5′-CCAAGCTTGAGCAAAGCACACTGAGGATGGC-3′
Luc2-rev	5′-CCAAGCTTGCGACTTGTGCCGTGCTGACTC-3′
Primers used in enhancer blocking assay
Primer 1	5′-CAGGGGCATACTCATCTTTGGC-3′
Primer 2	5′-GCAACCACTGCGGCTTCCAC-3′
Primer 3	5′-AGCGACAACGGGTAGGCCAC-3′
Primer 4	5′-CCACCATCATAGACCACGC-3′
Primer 5	5′-GCGTGGTCTATGATGGTGC-3′
Primer 6	5′-GACTTGTGCCGTGCTGACTC-3′
Primer 7	5′-GCGCTCATCATAGCCTCC-3′