Genomic imprinting results in expression of some autosomal genes from one parental allele only. Human chromosome 11p15, and the syntenic region on mouse distal chromosome 7, contain several imprinted genes, including p57KIP2(CDKN1C) and IGF2. These two genes, which are separated by >700 kb, are both implicated in the pathogenesis of Beckwith-Wiedemann syndrome. We have shown previously that an Igf2/H19 transgene is expressed appropriately and can imprint at ectopic chromosomal locations. To investigate the p57KIP2 region, we similarly tested the imprinting and function of a 38 kb human genomic fragment containing the p57KIP2 gene in transgenic mice. This transgene showed appropriate tissue-specific expression and transgene copy number-dependent expression at ectopic sites. However, the levels of expression are reminiscent of that found for the paternal allele in humans (10%). There was no change in expression levels when the transgene was inherited from the maternal germline. These results suggest that the cis-elements required for enhanced expression of the maternally inherited p57KIP2 allele lie at a distance from the gene. This finding has important implications for the role of this gene in the human disease, in particular with respect to the translocation breakpoints identified in some patients.
The p57Kip2 gene lies within a cluster of imprinted genes in the distal region of mouse chromosome 7 and within the syntenic region of human chromosome 11p15 (1,2). The p57Kip2 gene encodes a cyclin-dependent kinase inhibitor whose expression in post-mitotic differentiated cells and ability to inhibit the cell cycle suggest a role in the decision-making process of cellular differentiation (2,3). The mouse gene is expressed from the maternal allele only, and the paternal allele is methylated and repressed in all tissues examined (1). In humans, ∼90% of the expression comes from the maternal allele, but no differential methylation between the two parental alleles has been reported (4–7).
Within the 1 Mb region of mouse distal chromosome 7, eight genes are known to be expressed monoallelically; both their gene order and their imprinting are conserved in humans (reviewed in refs 8,9). Disruptions of 11p15 are associated with Beckwith-Wiedemann syndrome (BWS), and the parent of origin-dependent inheritance of this disease suggests the involvement of imprinted genes (10–13). Since some BWS patients show biallelic IGF2 expression (14,15) and other, familial, cases have germline mutations in the p57KIP2 coding sequence (16–18), BWS may be a contiguous gene syndrome involving more than one imprinted gene. Furthermore, in the mouse, both overexpression of Igf2 (19–22) and deletion ofp57Kip2 (23,24) result in developmental abnormalities with similarities to BWS. In humans, five germline balanced chromosomal translocations fall within the 11p15 domain but they do not physically disrupt either the p57KIP2 gene or the IGF2 gene (25) but another imprinted gene within the region, KvLQT1 (26). There may be elements within this gene, which are required for correct imprinting of the whole 11p15 region, which resemble the Prader-Willi/Angelman syndrome ‘imprinting centres’ proposed to direct imprinting at human chromosome 15q11–q13 (27,28). Recently, a novel, paternally expressed transcript has been reported which lies within KvLQT1 and which may be disrupted by these translocation events (29).
Despite their relative proximity and their proposed involvement in BWS, there is no direct evidence for a mechanistic link between the imprinting of p57Kip2 and Igf2/H19.Wehave shown previously that a 130 kb mouse yeast artificial chromosome (YAC) transgene containing Igf2 and H19 can imprint independently of the upstream distal chromosome 7 region (30). In addition, the timing and direction of imprinting of p57Kip2 and H19 differ in the germline. The H19 paternal allele is active until late in spermatogenesis, but the p57Kip2 maternal allele is silent and is activated by passage through the female germline late in oogenesis (31–33). These data led us to suggest that p57Kip2 and H19/Igf2 constitute two separate imprinted domains (8). Recently, two biallelically expressed genes have been identified in the region between Kvlqt1 and Mash2 (34). If the domain which contains p57Kip2 is separate from that which contains H19 and Igf2, a DNA clone containing the p57Kip2 gene may imprint at ectopic sites and by a mechanism different from that of H19.
To address this question and to identify the control cis-elements involved in expression and imprinting of p57Kip2,we have tested the ability of a 38 kb human p57KIP2 transgene to imprint in the mouse. If imprinting of the humanp57KIP2 gene occurred in the mouse, it would indicate conservation of regulatory elements involved in the imprinting process and facilitate their identification by sequence comparison. We report the generation of transgenic mice which show appropriate tissuespecific expression of the p57KIP2 transgene. However, the levels of expression are low and reminiscent of that found for the paternal allele in humans; passage of the transgene through the female germline did not affect its expression. These results suggest that the cis-elements required for enhanced expression normally seen on maternal transmission of p57KIP2 are absent from the transgene and may lie at some distance from the gene.
Sequence comparison of the human and mouse p57KIP2 loci
We first compared the sequence, methylation status and genomic organization of the mouse and human p57Kip2 genes. The humanp57KIP2 gene spans a small genomic region of <2.5 kb and is contained within four exons (35). We isolated a mouse p57Kip2 bacterial expressed chromosome (BAC) genomic clone using a 0.53 kb fragment from the p57Kip2 cDNA. We sequenced a 3 kb fragment containing the gene and, by alignment with the published p57Kip2 cDNA sequence, determined that the exon-intron boundaries were highly conserved (Fig. 1A; GenBank accession no. AF160190). The human and mouse p57Kip2 predicted proteins share amino acid sequence conservation in the cyclin-dependent kinase inhibitory domain and in the QT domain, but the two internal proline-rich domains and an acidic repeat domain found in the mouse sequence are replaced by a single PAPA repeat in the human sequence (2). The maintenance of the genomic organization of the homologues, in particular at the exon-intron boundaries, would suggest that the internal sequence has diverged between the two species rather than representing distinct events such as alternative splicing or insertion/ deletion. The overall level of identity over the sequence we compared is 73%, with 91% identity over 146 bases around the TATA and CAAT boxes which include the predicted promoter sequence. Within intron II there are two regions with a striking degree of homology: Region A is 90% identical over 144 bases and 80% G+C, and Region B is 98% identical over 154 bases and 72% G+C.
CpG-rich sequences and differential methylation are known to be associated with imprinted genes, and CpG analysis (CpGPLOT) of the two genes confirms that both are exceptionally GC- and CpG-rich. The predicted promoter region and all four exons are contained entirely within a CpG island-like region with a GRAIL CpG island/GC score of 75.6 for the human sequence and 69.0 for the mouse (Fig. 1B). As previously reported (1), in the mouse, the paternal allele is methylated and repressed but no methylation differences have been reported between the parental alleles in humans. We analysed the methylation status across the CpG-rich region of both genes using the methylation-sensitive restriction enzymes (Fig. 1C). We have shown previously that the mouse maternal allele is unmethylated at several EagI sites (36). The continued presence of the 2.9 kb gene-containing BamHI fragment after digestion with BssHII, EagI andNotI indicates that there is methylation of the paternal allele. However, some CpG dinu-cleotides within the gene on the paternal allele are unmethylated as the BamHI fragment is absent in the SacII digest and present at <50% with the other enzymes. At this level of analysis, the human gene is unmethylated on both alleles including at the NotI sites within the predicted promoter region. This absence of methylation at the human locus may explain why the gene is only partially repressed after paternal inheritance in humans, since methylation is thought to be an important mechanism by which transcriptional silencing is maintained (37).
Characterization of transgene and generation of p57KIP2 embryonic stem (ES) cell lines
A number of large genomic clones had been identified as part of a project to physically map human chromosome 11p15 (25). Our restriction analysis placed the p57KIP2 gene at the centre of a cosmid clone, Q6, with 20 kb of genomic sequence upstream and 15 kb of sequence downstream (Fig. 2A). We linearized the cosmid clone with £coRV and co-electroporated it into ES cells with a linearized neomycin-selectable marker driven by the constitutively active phosphoglycerate kinase promoter (pPGK neo bpA). We analysed DNA from resistant ES cell colonies by digestion with EcoRI, Southern blotting and hybridization with radioactively labelled Q6 DNA and compared the banding pattern with that of the EcoRI-digested construct (Fig. 2A). Eighteen of the 36 clones contained intact integration events. We used a probe from the right arm of the cosmid vector to determine copy number, and there were integrations of between one and at least 20 copies of the transgene in the different lines (data not shown).
Expression of the humanp57KIP2 transgene during ES cell differentiation
Withdrawal of leukaemia inhibitory factor (LIF) in the absence of an adherent feeder layer of embryonic fibroblasts induces the differentiation of ES cells and the formation of embryoid bodies (38). Undifferentiated ES cells express very low levels of two cyclin-dependent kinase inhibitors, p21 and p27, and withdrawal of LIF results in their up-regulation (39). We initially determined whether this was also true for p57Kip2.We prepared RNA samples at day 0 and day 5 of differentiation from 1B2, an ES cell line with three copies of the transgene. By RT-PCR analysis, using primers R27 and R28 which span intron I of the mouse gene, we were able to detect a transcript in the differentiated cells but not in the undifferentiated cells, indicating that p57Kip2 is up-regulated during differentiation (Fig. 2B). RT-PCR across intron I also amplified a larger RNA species suggesting that, like the human gene (35), the mouse gene also undergoes alternative splicing at the 5′ endtoinclude intron I. Using primers R13 and R14 which span intron III of the human gene, we were able to detect expression in both undifferentiated and differentiated cells, and the level of expression in differentiated cells appeared higher. We confirmed these results by northern analysis of poly(A)+ RNA (data not shown).
Transmission of the transgene
We obtained live born male chimeras for five ES cell clones and mated these with 129/Sv females to generate two lines of mice carrying the human transgene: 1B2 (three copies) and 2B1 (one copy). The consequence of overexpression of p57Kip2 in a whole organism has not been tested previously, but studies in cell lines suggest that inhibition of the cell cycle in some tissues could lead to observable developmental defects (2,3). Transgenic pups obtained from the 1B2 and 2B1 chimeras were indistinguishable from their wild-type littermates at birth, indicating that paternal inheritance of this transgene had no overt phenotypic consequence.
We analysed expression of the 1B2 transgene by northern analysis of poly(A)+ RNA after male and female transmission (G1)inkidneyandbrainfromnewbornmice(Fig.3).We detected the human p57KIP2 transcript after both male and female transmission at similar levels (1:1.2 and 1:1.1 ratio male to female, respectively), suggesting that the transgene was not subject to imprinting in these tissues. For line 2B1, we were able to amplify the human transcript by RT-PCR with equal efficiency (data not shown). In addition, the transgene remained unmethylated after both male and female transmission in both lines (data not shown).
By in situ hybridization on saggital sections of 13.5 day post-coitum (d.p.c.) 1B2 embryos, we determined that the human transgene was expressed in all the tissues which express endogenous p57Kip2 (Fig. 4A and B). Specifically, transgene expression is detectable in the smooth muscle layer of the intestine (Fig. 4D), the developing pituitary (Fig. 4F), the muscle of the tongue (Fig. 4H), the epithelium of the lung, the cartilage of the developing ribs and in skeletal muscle (Fig. 4J) and also in the spongiotrophoblast layer of the placenta (Fig. 4L) with a similar pattern to the endogenous gene (Fig. 4C, E, G, I and K, respectively). No signal was observed with the sense probe (Fig. 4M), and the human antisense RNA probe did not cross-hybridize with any mouse transcripts (Fig. 4N). Expression at 10.5 d.p.c. was also consistent with the endogenous mouse gene (data not shown).
We were able to detect expression from the human transgene by RT-PCR, northern and in situ hybridization, and the pattern of expression from the transgene was similar to that of the endogenous p57Kip2. However, our analysis suggested that the level of expression from the transgene was low. In line with three copies of the human transgene, the human in situ signal was barely detectable at a time when a strong signal was seen for the mouse RNA probe. To quantify the level of expression of the human transgene accurately, we developed an RNase protection assay using mouse and human antisense RNA probes incorporating a similar number of [α-32P]CTPs. This allowed us not only to compare directly the levels of expression of the human transgene between lines but also to compare them with the endogenous mouse transcript. We detected the maternally inherited human transcript in total brain RNA from neonates for the 1B2 line at ∼14% of the level of the endogenous gene per copy and at 13% for the single copy line, 2B1 (Fig. 5A).
One possibility is that the transgene in both lines may be silenced predominantly. We assessed the level of expression of the transgene in the parent ES cell lines and in three additional, independent lines using the RNase protection assay (Fig. 5B). In the low copy lines, the protected transcript was barely visible, while in the two high copy lines, 2A1 and 2A5, the protected transcript was clearly detectable (Fig. 5B). When adjusted for copy number, the level of expression from the transgene was between 9 and 36% of the endogenous mouse gene expression in all the lines. For lines 1B2 and 2B1, this level was similar to the levels we detected in vivo.The consistent results between the five lines exclude position effect and high copy number-dependent silencing as explanations for low levels of expression.
Our studies suggest that this human transgene may behave as the paternal allele in humans. It lacks the regulatory elements required for enhanced maternal expression which therefore lie at a distance from the gene. In conclusion, although the human transgene is able to direct correct spatial and temporal expression, it fails to achieve the level of activation of the maternal allele.
Large transgenes have been used in the past to study gene regulation and to rescue lethal phenotypes (40,41). More recently, similar studies have been carried out to characterize imprinted genes and their cis-acting control sequences (30,42). In this study, we have used a human p57KIP2 sequence to examine expression and imprinting of this gene. The use of human transgenes provides direct sequence comparison that potentially allows rapid detection of sequences conserved between mouse and man which may be an indication of function.
In this study, we have used a large human p57KIP2 transgene in mice with 20 kb of upstream sequence and 15 kb of downstream sequence. We obtained temporal and spatial patterns of expression that were indistinguishable from the endogenous mouse p57Kip2 gene. However, the levels of expression were 10-fold less than normal. The transgene expression was unaffected by its parental origin, unlike the endogenous gene which shows high levels of expression after maternal inheritance. The position of transgene integration can affect transcription, although this is more common with smaller transgenes (43). High copy concatameric arrays of transgene copies at a single locus can also have an adverse effect on their expression (44). However, we can rule out both these possibilities as being responsible for the low levels of p57KIP2 expression. This is because we observed only the appropriate expression and no unexpected patterns that can occur with smaller transgenes in ectopic sites (45). Furthermore, our data indicate that although the levels were low, we did obtain copy number-dependent levels of expression, suggesting that all the copies were being expressed similarly at low levels. There is another possibility that some of the human control elements such as enhancers and/or promoters were not recognized in the mouse, resulting in the low levels of expression. However, this seems unlikely based on previous successful studies with human transgenes which have been shown to function appropriately in the mouse (examples are refs 41,46–48).
A plausible alternative explanation for the spatially appropriate but low level expression may be that the regulatory elements responsible for enhanced expression of the human p57KIP2 gene lie outside the 38 kb transgene we used in this study. The complete regulatory sequences of a gene can be composed of a number of autonomous enhancer modules, each of which directs transcription in some specific tissue at a specific time. These enhancers are generally near their cognate core promoters, although they can also act over long distances. It is not exceptional for some of them to lie many kilobases away from their target gene (49).
It is important to recall that the humanp57KIP2 gene does not exhibit differential methylation seen in the mouse where the silent paternal copy is methylated and shows no expression. In contrast, the human gene shows low levels of expression from the paternal allele. The absence of methylation of the paternal allele in humans leaves it potentially active, but the imprint, whatever its nature, prevents its full expression. The absence of appropriate enhancers on the transgenes is therefore consistent with the tissue-specific but low levels of expression we detected, which is similar to the human p57KP2 paternal allele.
Given that p57KIP2 is an imprinted gene, another interpretation of our results is that the transgene lacks a regulatory element which is necessary for imprinting. More particularly, the transgene may lack an element that is capable ofconferring high levels of expression on maternal inheritance. Evidence for the imprinting of p57Kip2 in the mouse demonstrates that the erasure of the parental imprints in primordial germ cells causes silencing of the p57Kip2 gene; this could be considered as the default state for the gene prior to imprinting (31,33). The maternal allele is made active only following its passage through the growing oocyte (32). Therefore, the inherent state of p57Kip2 is to be silent. What we observe with the human p57KIP2 transgene in mice is perhaps a reflection of its behaviour in its silent state. It follows that a control element, perhaps an activator sequence, is required to convert the locus from an inactive chromatin state to an active state after passage through the female germline. If such an element exists, it must also lie outside the transgene which we used in this study. Conclusive proof that the regulatory elements lie outside the vicinity of the gene will require similar data from an equivalent sized mouse p57Kip2 transgene.
Studies on a number of chromosomal domains have demonstrated that imprinted genes tend to occur in clusters where they exhibit a complex sharing of control elements (42,50). p57KIP2 lies at the distal end of the 11p15 imprinted domain and 65 kb away from the biallelically expressed NAP-2 gene (51,52). This may mark the boundary between imprinted and non-imprinted chromatin. The cis-elements required for full expression may lie within this 45 kb sequence not covered by the 5′ region of our transgene. However, the translocation breakpoints in BWS lie downstream of p57KIP2 (25) and it seems more likely that the putative activator/enhancers lie >15 kb downstream of the p57KIP2 gene, perhaps within KvLQT1.
Recently, in humans, a novel paternally expressed transcript, LIT1, was identified within KvLQT1 (29). LIT1 is associated with a maternally methylated CpG island, and loss of imprinting (LOI) is observed in 58% of the BWS patients tested. The authors suggest that LOI of LIT1 in some way affects transcription of p57KIP2. Their preferred model is that the unmethylated CpG island of LIT1 acts as an insulator between the p57KP2 gene and its enhancer(s). However, in mice lacking DNA methyltransferase (Dnmt1−/−), where presumably both CpG islands are unmethylated, p57Kip2 is active (53), which makes the insulator model untenable. By a similar argument, their second model, where transcription from the KvLQT1 promoter displaces a repressor of p57KIP2, is also less likely as Kvlqt1 is not expressed in Dnmt1−/− mice. Transcription from the KvLQT1 promoter, at least in the mouse, is therefore not involved directly or indirectly in p57Kip2 activation. Their third model, where the oppositely imprinted genes compete for shared enhancers, is also hard to reconcile with data from the mouse. p57Kip2 is expressed biallelically in the Dnmt1−/− mouse, and loss of methylation of the LIT1 CpG island in BWS patients results in biallelic expression of LIT1 and this presumably also may be the case in Dnmt1−/− mice. If the absence of methylation is an indication of an active promoter, LIT1 and p57KIP2 cannot be competing for the same enhancers. In some respects, the behaviour of p57Kip2 resembles that of the Igf2r gene. Both genes are expressed from the maternal allele, both genes are silent after imprint erasure and both require passage through the female germline for activation (31,32). However, unlike p57Kip2, Igf2r is not expressed in the Dnmt1−/− mouse. None of the existing models can therefore fully explain how the imprinting and regulation of p57Kip2 occurs. It is possible that there is a novel mechanism involved in the imprinting of the p57Kip2 gene. Whatever the relationship is between these genes, our data suggest that there may be elements controllingp57Kip2 expression within KVLQT1 which can be directly addressed in vivo by testing transgenes (with appropriate modifications) spanning the two regions.
The observation that the cis-elements required for enhanced maternal expression of p57KIP2 are not closely linked to this gene raises a separate but very important issue concerning BWS. Despite striking similarities in the phenotype of the null p57Kip2 mouse with BWS, this gene has often been thought to be a minor player in this disease because mutations in the gene are rare. However, our prediction is that one or more of the enhancers for p57KIP2 or an activating element lie within KvLQT1. In BWS patients, the translocation events may separate p57KP2 from these regulatory elements or disrupt their function. This would result in a loss of expression, perhaps in specific tissues which might not have been examined previously but crucially, without a detectable mutation at the p57KIP2 locus. These findings might explain the very low mutation frequency of p57KIP2 seen in BWS patients and can be addressed by examining p57KIP2 expression in patients with translocations or LOI of LIT1.
Materials and Methods
BAC library screen
A 129 gridded pBeloBAC library was purchased from Genome Systems (St Louis, MO) and hybridized with a 0.5 kb XhoI-EagI fragment from thep57Kip2 cDNA (GenBank accession no. U22399; a kind gift of S. Elledge, Baylor College of Medicine, Houston, TX). One positive clone, 144D14, was identified.
DNA and RNA preparation
BAC and plasmid DNA were isolated from liquid culture by alkaline lysis (54). RNA was prepared from ES cell lines or ground up tissue using RNAzol B (AMS Biotechnology, Witney, UK). Total RNA was reverse transcribed with random hexamers (Life Technologies, Paisley, UK) using M-MuMLV reverse transcriptase (Boehringer Mannheim, Mannheim, Germany). Poly(A)+ RNA was prepared using the MicroFast-Track system (Invitrogen, San Diego, CA). Genomic DNA was prepared as described in Hogan et al. (55).
Southern and northern analysis and genomic probes
DNA and RNA blots were prepared by standard procedures (54) using Hybond N+ membrane (Amersham, Little Chalfont, UK). ES cell lines were characterized using the whole 38 kb cosmid as a probe and with a right vector arm probe of 2.6 kb. Methylation status was determined using a 0.53 kb XhoI-EagI fragment of the p57Kip2 cDNA and a 1.1 genomic NotI fragment of the human gene.
Primers R27 (5′-ACTGAGAGCAAGCGAACA-3′)and R28(5′-AAGCGTTCCATCGCTGTTCTG-3 ′), which span intron I of the mouse p57Kip2 gene, were used to detect a 120 bp product. Primers R13 (5′-GATTTCTTCGCCAAGCGC-3′)and R14(5′-GCACTGAGTTTCAGCAGAG-3′), which span intron III of the human p57KIP2 gene, were used to detect a 420 bp product. Primers AC1 (5′-GCTGTGCTATGTTGCTCTAG-3′)andAC2(5′-ATCGTACTCCTGCTTGCTGA-3′) were used to detect a 400 bp product of the mouse β-actin gene. PCR was performed using the Expand Long Template PCR system (Boehringer Mannheim). PCR of mouse p57Kip2 consisted of an initial denaturation step at 96°C for 3 min followed by 25 cycles of 94°C for 30 s, 65°Cfor 30 s, 68°C for 30 s. PCR of humanp57KKP2 was performed at an annealing temperature of 55°C for 35 cycles and of mouse β-actin at 58°C.
RNase protection assay
Antisense [α-32P]CTP-labelled RNA probes were synthesized from linearized mouse (GenBank accession no. U22399, nucleotides 105–216) and human (GenBank accession no. U22398, nucleotides 146–276) 5′-untranslated region sequences, gel eluted, hybridized and digested according to the protocol supplied by Ambion (Austin, TX). Products were resolved on a 6% denaturing acrylamide gel.
Embryonic stem cell lines
ES cells were maintained in the undifferentiated state in ES cell medium [Dulbecco's modified Eagle's medium-12 supplemented with 20% fetal calf serum (FCS), glutamine, non-essential amino acids, sodium pyruvate, sodium bicarbonate, antibiotics and β-mercaptoethanol] on feeder layers ofprimary embryonic fibroblasts in the presence of LIF-conditioned medium. pPGK neo bpA was a gift ofS. Aparacio (Wellcome/ CRC Institute, Cambridge, UK).
On day 3.5 of pregnancy, blastocysts were flushed from the uteri of MF1 females naturally mated with MF1 males. Approximately 10 ES cells were injected into the blastocoel cavity in PB1 medium containing 10% FCS. After injection, the blastocysts were cultured for up to 2 h and transferred into the uteri of 2.5 d.p.c. psuedopreganant (C57BL/6xCBA)F1 females.
In situ hybridization of p57Kip2/KIP2
A 1.4 kb fragment of the mouse cDNA and a 1.1 kb NotI fragment spanning exons I and II of the human gene were used to prepare sense and antisense RNA probes by in vitro transcription using the DIG RNA labelling kit (Boehringer Mannheim). Saggital sections from normal and transgenic mouse embryos at 10.5 and 13.5 d.p.c. were used for in situ hybridization. Briefly, embryos were fixed in 4% paraformaldehyde at 4°C overnight, sections made and hybridized with the probes overnight at 65°C. The sections were washed at 65°C and incubated with pre-adsorbed alkaline phosphatase-conjugated anti-DIG antibody overnight at 4°C. Alkaline phosphatase activity was detected using BM purple AP substrate (Boehringer Mannheim) and counterstained with 0.5% eosin.
We would like to thank all members of the laboratory for their support, in particular J. Ainscough, L. Lefebvre, M. Narasimha and K. Otte for their help and advice and J. Ainscough and M. Narasimha for critical reading of the manuscript. This work was supported by a Wellcome Trust grant (036481).