Human chromosome 11p15.5 harbors an intriguing imprinted gene cluster of 1 Mb. This imprinted domain is implicated in a wide variety of malignancies and Beckwith–Wiedemann syndrome (BWS). Recently, several lines of evidence have suggested that the BWS-associated imprinting cluster consists of separate chromosomal domains. We have previously identified LIT1, a paternally expressed antisense RNA within the KvLQT1 locus through a positional screening approach using human monochromosomal hybrids. KvLQT1 encompasses the translocation breakpoint cluster in BWS and patients exhibit frequent loss of maternal methylation at the LIT1 CpG island, implying a regulatory role for the LIT1 locus in coordinate control of the imprinting cluster. Here we generated modified human chromosomes carrying a targeted deletion of the LIT1 CpG island using recombination-proficient chicken DT40 cells. Consistent with the prediction, this mutation abolished LIT1 expression on the paternal chromosome, accompanied by activation of the normally silent paternal alleles of multiple imprinted loci at the centromeric domain including KvLQT1 and p57KIP2. The deletion had no effect on imprinting of H19 located at the telomeric end of the cluster. Our findings demonstrate that the LIT1 CpG island can act as a negative regulator in cis for coordinate imprinting at the centromeric domain, thereby suggesting a role for the LIT1 locus in a BWS pathway leading to functional inactivation of p57KIP2. Thus, the targeting and precise modification of human chromosomal alleles using the DT40 cell shuttle system can be used to define regulatory elements that confer long-range control of gene activity within chromosomal domains.
Received 28 April 2000; Revised and Accepted 10 July 2000.
Beckwith–Wiedemann syndrome (BWS) is the most common overgrowth syndrome manifested as numerous growth abnormalities especially exomphalos, macroglossia and gigantism (1,2). Children with BWS also show an increased susceptibility to a variety of childhood tumors including Wilms’ tumor, adrenocortical carcinoma and hepatoblastoma (3). BWS is a complex, multigenic disorder that includes familial, sporadic cases and those with chromosomal anomalies (4). Although the causative defects of BWS in all three subgroups are not fully understood, it is apparent that multiple imprinted genes on human chromosome 11p15.5 are involved in the pathology of BWS.
To date it is known that at least eight imprinted genes are located within the 11p15.5 region (see http://cancer.otago.ac.nz/IGC/Web/home.html ). Since both duplication of the paternally derived 11p15 and paternal uniparental isodisomy of 11p are found in sporadic BWS with chromosomal anomalies (5,6), it has been suggested that the overexpression of a paternally expressed gene and/or the deficiency of a maternally expressed gene could contribute to the molecular pathogenesis of BWS. Indeed, most sporadic patients exhibit abnormal biallelic expression of IGF2, a polypeptide growth factor that is normally transcribed only from the paternal allele (7,8). Consistent with this, overexpression of Igf2 in mouse embryos leads to the dosage-dependent appearance of some, but not all, of BWS phenotypes (9–11).
Maternally inherited germ line mutations in the coding region of p57KIP2 are found in other familial cases of BWS (12,13). p57KIP2 is a maternally expressed gene that encodes a potent inhibitor of cyclin-dependent kinases (14,15), mice lacking p57Kip2 also exhibit many features that model BWS, although most of these phenotypes do not overlap those seen in transgenic mice overexpressing IGF2 (16,17). Recently, Caspary et al. (18) produced double mutant mice that both carried a null mutation of p57Kip2 and exhibited loss of Igf2 imprinting. This demonstrated that these two oppositely imprinted genes could act in an antagonistic manner during mouse embryogenesis. These observations and the fact that these genes are expressed in all of the tissues that are most affected in BWS (14,15,19) imply certain interaction between the pathways involving IGF2 and p57KIP2. In addition, IGFII protein down-regulates the transcriptional activity of p57Kip2, suggesting that the trans effect of increased IGFII protein is also implicated in the etiology of BWS (20).
We have previously identified LIT1, also known as KvLQT1-AS/KCNQ1OT1, a paternally expressed antisense RNA within the KvLQT1 locus in a systematic screen using human monochromosomal hybrids (21). The KvLQT1 locus spans >400 kb and contains multiple balanced chromosomal translocation breakpoints associated with BWS, which are proposed to disrupt an imprinting control element (22). A subset of BWS patients exhibited a complete loss of maternal methylation at the LIT1 CpG island which demonstrated that this region was closely linked to imprinting of LIT1 (21,23–25). Intriguingly, the LIT1 CpG island is not necessarily associated with maternally inherited chromosomal translocations or loss of IGF2 imprinting in BWS. In addition, our finding that LIT1 imprinting is not disrupted in Wilms’ tumor, which shows frequent loss of IGF2 imprinting, is consistent with several lines of evidence suggesting that the BWS-associated imprinting cluster at 11p15.5 harbors two or more independent chromosomal domains (26–29).
Although these observations raise the possibility that the LIT1 locus mediates imprinting of neighboring genes within the chromosomal domain, it has remained unknown how disrupted imprinting of the non-coding LIT1 RNA contributes to the development of BWS. Here we generated modified human chromosomes carrying a targeted deletion of the LIT1 CpG island using the chicken DT40 cell shuttle system. Consistent with our prediction, lack of LIT1 expression on the targeted chromosome was accompanied by activation of the normally silent paternal alleles of multiple imprinted loci at the centromeric domain including KvLQT1 and p57KIP2. We further showed that the deletion had no effect on imprinting of H19, which is located at the telomeric domain. These results provide insight into a novel BWS pathway that leads to functional inactivation of p57KIP2 and show that the targeted modification of human chromosomal alleles using DT40 cells should contribute to identify regulatory elements that define epigenetic states of gene activity within chromosomal domains.
Targeted deletion of the human LIT1 CpG island in DT40 cells
We created modified human chromosomes carrying a targeted deletion of the LIT1 CpG island with the use of the chicken pre-B cell line DT40 (30). DT40 cells exhibit high frequency homologous recombination between exogenous DNA templates and their chromosomal counterparts (31). This enables the efficient targeted modification of individual human chromosomes, which have been transferred from mammalian cells via microcell-mediated chromosome transfer (32–34). We have previously established monochromosomal hybrids containing individual human chromosomes of defined parental origin (35). Human chromosomes in the monochromosomal hybrids were tagged with pSV2bsr, which confers blasticidin S (BS) resistance, so that the marked chromosomes could be successfully transferred into mammalian cells (36). In the present study the paternal and maternal copies of human chromosome 11, hChr.11P and hChr.11M, were transferred independently from A9 cells into DT40 cells by microcell fusion. DT40 microcell hybrids which retained a single intact copy of human chromosome 11 were obtained from each chromosome transfer.
A targeting construct that removed 2.6 kb of the KvLQT1 genomic sequence, encompassing the bulk of the LIT1 CpG island (Fig. 1A), was introduced into DT40 hybrid cells containing hChr.11P. Using a PGK-puroR cassette with 4.0 kb of 5′ and 3.0 kb of 3′ homologous sequences, 7 of 33 (21%) drug-resistant clones selected in the presence of both puromycin and BS were targeted correctly as shown in Figure 1B. Genomic DNA from drug-resistant DT40 clones was analyzed by Southern blotting using a probe internal to the targeting construct. Homologous insertions of the targeting construct at the LIT1 locus yielded a 10.5 kb EcoRV fragment, but instead random integrations resulted in a wild-type 20.9 kb fragment plus one or more fragments of unspecified size. The absence of the wild-type allele in homologous recombinants is confirmed by long-range PCR analysis. Cytogenetic analysis demonstrated a targeting construct with the integration site assigned to 11p15.5 (Fig. 2) and this modified chromosome was termed hChr.11PΔCpG. To assess the effect of the targeted alteration at the LIT1 locus, hChr.11P, hChr.11PΔCpG and hChr.11M were subsequently transferred into Chinese hamster ovary (CHO) cells by microcell fusion. The composition of individual human chromosomes in CHO hybrid cells was determined by cytogenetic analysis (Fig. 2). Two independent CHO hybrid cells obtained from each chromosome transfer were selected for further expression and methylation analyses.
Effect of the CpG island deletion on LIT1 RNA expression
To address whether targeted deletion of the CpG island abolished LIT1 RNA expression, RT–PCR analysis was performed on RNA obtained from CHO hybrids containing hChr.11PΔCpG, as well as control CHO hybrids containing hChr.11P or hChr.11M. We have previously shown that transcripts defined by 10 expressed sequence tags (ESTs) distributed over 60 kb of the LIT1 locus exhibited exclusive paternal expression in human monochromosomal hybrids (21). In this study, we examined the expression of transcripts defined by six ESTs across the 86 kb LIT1 locus (Fig. 3). Exclusive paternal expression was exhibited at all these sites in the wild-type controls. None of these transcripts were expressed in hChr.11PΔCpG hybrid cells, whereas the expression of the non-imprinted genes TSSC4 and TSSC6 was detected as in the control hybrid cells. This indicates that the deletion of the 5′ CpG island impaired LIT1 the expression on the active paternal chromosome and further supports the idea that LIT1 is expressed as a continuous antisense RNA across the KvLQT1 locus, as is the case for Tsix RNA in the Xist locus and Air RNA in the Igf2r locus (37,38).
Effect of the CpG island deletion on the imprinting cluster
If the LIT1 locus can act as a local regulator of the imprinting cluster, lack of LIT1 expression from the targeted chromosome would be accompanied by activation of the normally silent paternal alleles of multiple maternally expressed genes at the centromeric domain. As expected, KvLQT1 isoform 1 was expressed in CHO11PΔCpG hybrid cells, whereas the preferential maternal expression was maintained in the control hybrid cells without the deletion (Fig. 3C). This is consistent with the prediction that the maternal allele of Kvlqt1 is repressed through the regulation of a paternally expressed transcript in Dnmt1-deficient mice (26). Activation of the silent paternal allele of SMS4/KCNO1DN, a recently identified maternally expressed RNA transcript which is located centromeric to KvLQT1 (Z. Xin and T. Mukai, personal communication), was also exhibited in CHO11PΔCpG hybrid cells. A striking observation was made for p57KIP2, which is frequently mutated in familial cases of BWS (13). In CHO11PΔCpG hybrid cells, the p57KIP2 transcript was readily amplified, but expression was not detectable in the control hybrid cells containing either hChr.11P or hChr.11M, indicating that the p57KIP2 allele was hyperactivated on the targeted chromosome. Although the mechanisms of this hyperactivation are currently unknown, this is consistent with recent transgenic experiments suggesting that enhanced expression of p57KIP2 requires an activating element which may lie at a distance from the gene (39). p57KIP2 transcripts show highly tissue-specific distributions (15), in contrast to apparently ubiquitous expression of KvLQT1 isoform 1 and LIT1 (21,22). Therefore, the observed differences in the effects of the deletion on different loci surrounding LIT1 could relate to the obviously distinct tissue distributions of these transcripts. This may also explain the lack of p57KIP2 expression from the intact maternal chromosome in CHO hybrid cells. In contrast to the activation of the paternal allele of KvLQT1, SMS4 and p57KIP2, maternal expression of H19 was unaffected in CHO11PΔCpG hybrid cells.
Although there are no data currently available for CD81 imprinting in humans, the mouse Cd81 has been reported to exhibit biallelic expression in most embryonic tissues (26). Contrary to the previous finding of paternal expression of human MTR1 using somatic cell hybrids (40), there is no evidence for the imprinted expression of Mtr1 in mice (41). In this study, we examined the allelic expression of CD81 and MTR1 genes, which are linked to apparently non-imprinted genes TSSC4 and TSSC6 in humans, and detected RT–PCR products from both parental alleles in CHO hybrid cells. This is consistent with our observation that the CpG island deletion did not alter the expression pattern of the two genes. Thus, it is suggested that MTR1 does not show a significant allelic expression bias at least in some human tissues, in contrast with the initial report by Prawitt et al. (40). For IGF2, the expression was readily detected in CHO hybrid cells with hChr.11PΔCpG as well as in those with hChr.11P, so that we could not assess the effect of the CpG island deletion on IGF2 imprinting (data not shown). Because of comparatively low levels of HASH2 transcripts in CHO hybrid cells, we excluded HASH2 from this analysis.
We further examined the CpG methylation of imprinted loci in the 11p15.5 cluster. To analyze methylation of the CpG-rich promoter region of the human H19 gene, methylation-sensitive SacII digestion was used in Southern blot experiments (Fig. 4). Genomic DNA was digested with either RsaI alone or double digestion with RsaI and SacII. Normal human DNA showed partial digestion of the 2.3 kb fragment in the double digest, consistent with paternal methylation at the promoter region of H19 (42). Analysis of the transferred maternal human chromosome 11 in CHO cells produced completely digested 1.0 and 1.3 kb fragments, whereas both control and modified paternal chromosomes showed intact 2.3 kb fragments in double digests. These results indicate that paternal methylation at the H19 locus was maintained appropriately in CHO11PΔCpG hybrid cells. The full methylation observed at this site correlated with the lack of H19 expression in CHO hybrid cells containing either hChr.11P or hChr.11PΔCpG. The 5′ CpG island at the SMS4 locus was analyzed using the methylation-sensitive restriction enzyme BssHII (Fig. 4). The double digestion of genomic DNA with HindIII plus BssHII generated intact 4.9 kb fragments in control hybrid cells containing hChr.11P, showing contrast with partial digestion of the targeted paternal allele. These results demonstrate that deletion of the LIT1 CpG island compromised the maintenance of differential methylation at this locus, causing the targeted paternal allele to resemble the wild-type maternal allele, though the completely digested 3.4 kb fragment was undetectable in CHOΔCpG hybrid cells. No differential methylation was observed at the p57KIP2 and HASH2 CpG islands (data not shown), consistent with previous reports (26,29,39).
One of the main purposes of this study was to investigate the significance of the maternally methylated CpG island at the human LIT1 locus for the pathogenesis of BWS. Here we made use of recombination-proficient chicken DT40 cells to generate modified human chromosomes carrying a targeted deletion of the human LIT1 CpG island. Using this chromosome shuttle system, homologous recombinants were obtained at frequencies of >1 in 5 puromycin-resistant clones and this positive selection was as efficient as that described previously (32–34). Subsequently, we performed imprinting analysis of multiple imprinted loci at the 11p15.5 cluster in CHO microcell hybrids. Our finding that the LIT1 CpG island can act as a negative regulator of the p57KIP2 allele on the paternal chromosome has significant implications for the pathogenesis of BWS. A substantial portion of BWS patients are attributed to complete loss of maternal methylation at the LIT1 CpG island (21,23–25). Consistent with the fact that germ line mutations of the maternally inherited p57KIP2 allele are observed frequently in familial BWS patients (13), mice carrying a loss-of-function mutation in p57Kip2 show a phenotype characteristic of some aspects of BWS (16,17). Together these observations suggest that loss of maternal methylation at the LIT1 CpG island may result in the biallelic inactivation of p57KIP2 in BWS patients. This predicts that disrupted imprinting of LIT1 and germ line mutations in p57KIP2 would have some common effects in BWS. In particular, p57KIP2 mutations have been reported in conjunction with a high incidence of exomphalos, which is linked to abdominal wall defects in p57Kip2-deficient mice (13,16). Genotype–phenotype studies of BWS patients suggesting that exomphalos is frequently observed in the patients with imprinting defects at the LIT1 locus are compatible with this assumption (4). Additionally, the results also lead us to suggest that recent observations implying a repressed state of the mouse p57Kip2 allele in potentially imprint-free primordial germ cells (43–45) could be explained by biallelic expression of the reciprocally imprinted Lit1 RNA.
On the basis of expression and methylation studies, we conclude that targeted deletion of the LIT1 CpG island affects imprinting in cis at the centromeric domain including KvLQT1, LIT1 and p57KIP2, but not imprinting of H19 located at the telomeric end of the imprinting cluster. This fact provides the initial direct demonstration that imprinting at the centromeric domain on 11p15.5 is regulated in cis by an imprinting center and that methylation at the LIT1 CpG island can activate the KvLQT1 and p57KIP2 alleles, as proposed by others (4). Thus, the LIT1 locus is essential on the paternal chromosome for the long-range silencing of maternally expressed loci spanning a distance of >440 kb in the 11p15.5 cluster. This leads to the suggestion that the local control of imprinting at the centromeric domain involves transcription competition mechanisms, as demonstrated for the telomeric domain including the Igf2/H19 gene pair (9,46,47). Although the detailed extent and directionality of the local control exerted by the LIT1 locus remains elusive, the preponderance of imprinted antisense RNAs suggests a possible involvement in the regulation of the imprinted domains (48–51). It is noted that evolutionarily conserved maternal methylation at the imprinted 11p15.5 region has been described only for the LIT1 CpG island (26–29). LIT1 is closely linked to BWS-associated translocation breakpoints, which have been proposed to disrupt an imprinting control element. These observations imply the possibility that evolutionarily conserved domains of the LIT1 RNA that are located downstream of the CpG island (24,41) could be involved in imprint initiation, as proposed for the H19 RNA and SNRPN BD transcripts (52). Recently, large-scale sequencing of the mouse 7F4/F5 region revealed a striking sequence conservation around the KvLQT1 locus between human and mouse (41). Multiple conserved intronic sequences dispersed over 160 kb around the LIT1 locus suggest that this region is evolutionarily under certain functional constraints and that additional regulatory elements or genes within the KvLQT1 locus could be involved in coordinate regulation of the centromeric domain.
The organization of the imprinted gene cluster into independent domains implies the existence of boundary elements or chromatin insulators that define separate domains of distinct functional states, although this remains to be proven. In this context, recent targeting experiments in mice have suggested strongly that the unmethylated upstream region of the H19 gene on the maternal chromosome insulates the Igf2 gene from tissue-specific enhancers that are located downstream of the genes (53,54). Detailed analyses of the chromatin configuration and replication timing at the mouse H19 locus support this idea (55–58). By analogy with the telomeric domain, it is plausible that the human LIT1 CpG island functions as a boundary element within the centromeric domain, since this region also contains G-rich direct repeat clusters and is methylated specifically on the inactive maternal chromosome throughout development, suggesting that the LIT1 CpG island is a direct target for de novo methylation during gametogenesis (21,23,41). This model implies that an unmethylated active boundary element at the paternal LIT1 locus could insulate multiple maternally expressed genes from putative enhancers. Conversely, this insulating activity would be impaired on the methylated maternal chromosome, because methylation might block access of certain specific proteins to the target sites at the LIT1 locus, such as the DNA-binding protein CTCF which mediates insulator activity (59).
Similar to the domain organization of imprinted genes on particular chromosomal regions, multiple clusters of genes that escape X chromosome inactivation have been identified on the short arm of the human X chromosome using a somatic cell hybrid system (60–62; and see http://mediswww.meds.cwru.edu/dept/genetics/willard/data.htm ). In addition, it is of special interest that single genes that escape X inactivation lie within inactive chromosomal domains, which implies additional gene-specific regulatory elements that define epigenetic states of gene activity (63). This parallels autosomal genes that are located within the 11p15.5 imprinting cluster but apparently escape genomic imprinting (64). In conclusion, the imprinted gene cluster at 11p15.5 may provide an attractive model region for a detailed understanding of the mechanismsof long-range control of gene activity within chromosomal domains. Towards the dissection of the molecular basis of the imprinted gene cluster, it will be important to define chromosomal regulatory elements using multiple diverse genetic approaches including targeted modifications of human chromosomes in DT40 cells as presented in this study.
MATERIALS AND METHODS
Mouse A9 cells containing a human chromosome 11 that were used as fusion donors for chromosome transfers were established and cultured in the presence of BS essentially as described (35). Chicken DT40 cells were obtained from Dr Shunichi. Takeda (Kyoto University, Japan) and maintained at 37°C in RPMI 1640 medium supplemented with 10% fetal bovine serum, 1% chicken serum and 50 µM 2-mercaptoethanol. CHO cells used as fusion recipients for chromosome transfers were maintained at 37°C in F-12 nutrient mixture supplemented with 10% bovine calf serum.
Microcell-mediated chromosome transfer
Introduction of human chromosome 11 into DT40 cells was performed via microcell-mediated chromosome transfer as described previously (33), with the following modifications. Microcells purified from 1 × 108 A9 cells containing a human chromosome 11 were fused with DT40 cells propagated as an adherent monolayer on poly-l-lysine-coated dishes, using 50% (w/v) polyethylene glycol 1500 (Boehringer Mannheim, Indianapolis, IN). The fusions were diluted in 48 ml medium and divided into two 24-well plates. DT40 microcell hybrids were then selected in 15 µg/ml BS and used as fusion donors for subsequent chromosome transfers. CHO microcell hybrids containing a human chromosome 11 were obtained by polyethylene glycol-mediated fusion of CHO cells with microcells purified by centrifugation from 1 × 109 DT40 hybrid cells maintained on poly-l-lysine-coated flasks. The fusions were then selected in 3 µg/ml BS and picked for expansion.
Targeted deletion of the LIT1 CpG island in DT40 cells
The targeting vector was designed to replace the bulk of the LIT1 CpG island sequence with a puromycin resistance gene in the opposite transcriptional orientation as the endogenous locus. The 5′ region of homology was generated as a 4.0 kb fragment (–4.3 to –0.3 kb upstream of the CpG island) by long-range PCR of human genomic DNA, and was inserted into a SacII site of pBluescript II SK– (Stratagene, La Jolla, CA) carrying a 1.7 kb PGK-puroR cassette, in which a hybrid gene consisting of the phosphoglycerate kinase I promoter drives the puromycin resistance gene. A 3.0 kb KpnI fragment (+0.6 to +3.6 kb downstream of the CpG island) corresponding to the 3′ region of homology was inserted into a KpnI site of the vector. DT40 cells (1 × 107) were collected in 0.8 ml phosphate-buffered saline with 25 µg of NaeI-linearized targeting vector and electroporated at 550 V and 25 µF using a Gene Pulser apparatus (Bio-Rad, Hercules, CA). Double drug-resistant DT40 clones were then selected in 15 µg/ml BS plus 0.3 µg/ml puromycin using similar procedures as described above. Homologous recombination in DT40 hybrid clones was identified by PCR and Southern blot analysis using a PCR-generated probe. Primer sequences and reaction conditions described here are available from the corresponding author.
To identify a morphologically intact human chromosome 11 in metaphase spreads following chromosome transfer, conventional chromosome preparations were obtained as described previously (36). Chromosomes were visualized by Quinacrine plus Hoechst 33258 staining and at least 20 well-banded metaphases were analyzed. Fluorescence in situ hybridization (FISH) analyses were performed on fixed metaphase spreads of each cell hybrid using digoxigenin-labeled (Boehringer Mannheim) human COT-1 DNA (Gibco BRL, Grand Island, NY) and biotin-labeled PGK-puroR plasmid DNA essentially as described (65). Chromosomal DNA was counterstained with DAPI (Sigma, St Louis, MO). Image capture was archived using a CytoVision Probe system (Applied Imaging, Santa Clara, CA) and figures depicting chromosome spreads were prepared for reproduction using Adobe Photoshop version 5.5 (Adobe Systems, San Jose, CA).
Total RNA was extracted using RNeasy columns according to the manufacturer’s instructions (Qiagen, Chatsworth, CA) and treated with RNase-free DNase I (Takara, Kyoto, Japan). First strand cDNA synthesis was carried out with an oligo(dT)15 primer and M-MLV reverse transcriptase (Gibco BRL). PCR was performed on the cDNA with AmpliTaq Gold (Perkin Elmer, Foster City, CA) using a step-down protocol as described previously (21). Amplified fragments were resolved on a 3.0% agarose gel followed by staining with SYBR Green I (Molecular Probes, Eugene, OR). All PCR amplifications were carried out at least three times. Primer sequences were as follows:
for HRAS, 5′-TGAGGAGCGATGACGGAATA-3′
for H19, 5′-TACAACCACTGCACTACCTG-3′
for TSSC6, 5′-GACCGGGAAGCTGGAGAG-3′
for TSSC4, 5′-ACTGAGGTCCTGCCTGTGTT-3′
for KvLQT1 isoform 1, 5′-CGCGTCTACAACTTCCTCG-3′
for SMS4, 5′-GCTTTCTTTCCAGTGGCTCA-3′
for p57KIP2, 5′-GAGCGAGCTAGCCAGCAG-3′
for MTR1, 5′-AGACCCTCACCCAAAGGAAC-3′
for CD81, 5′-GACCCCACCGCGCATCCT-3′
Genomic DNA was extracted by standard phenol–chloroform extraction methods and digested with the appropriate restriction enzymes. The resulting fragments were fractionated on a 0.8% agarose gel and analyzed by Southern hybridization as described previously (66). Southern blots were imaged using a BAS-2500 bioimaging analyzer (Fuji Film, Tokyo, Japan). The probe was generated by PCR from human genomic DNA. Primer sequences were as follows:
for the H19 5′ probe, 5′-AACAACCCTCACCAAAGGCC-3′
for the SMS4 3′ probe, 5′-ACTCCCATCTGCTTCCCTTT-3′
We thank Dr Shunichi Takeda for the initial supply of DT40 cells and for valuable advice on gene targeting and Drs Minoru Koi and Kazuma Tomizuka for helpful advice and comments. This work was supported by CREST of Japan Science and Technology Corporation (JST), and grants from the Mitsubishi Foundation, the Ministry of Health and Welfare of Japan and the Human Frontier Science Program Organization (HFSPO).
To whom correspondence should be addressed. Tel: +81 859 34 8260; Fax: +81 859 34 8134; Email: firstname.lastname@example.org