Elucidation of the Minimal Sequence Required to Imprint H19 Transgenes

Genomics 73, 98 –107 (2001) doi:10.1006/geno.2001.6514, available online at http://www.idealibrary.com on

Elucidation of the Minimal Sequence Required to Imprint H19 Transgenes Melanie J. Cranston, 1 Tracy L. Spinka, David A. Elson, 2 and Marisa S. Bartolomei 3 Howard Hughes Medical Institute and Department of Cell and Developmental Biology, University of Pennsylvania School of Medicine, Philadelphia, Pennsylvania 19104 Received December 6, 2000; accepted January 23, 2001

modification alone is responsible for the observed imprinting patterns. Characterization of the cis-acting sequences required to imprint an endogenous gene is necessary before one can fully elucidate the mechanism by which imprinting is determined. Accordingly, the employment of mouse transgenes to recapitulate the endogenous pattern of imprinting has proven particularly useful in the study of the mouse H19 and Igf2r genes (Ainscough et al., 1997; Bartolomei et al., 1993; Brenton et al., 1999; Elson and Bartolomei, 1997; Pfeifer et al., 1996; Wutz et al., 1997). Specifically, we have shown that a 2-kb region from ⫺2 to ⫺4 kb relative to the start of H19 transcription [designated the differentially methylated domain (DMD)], which is proposed to harbor the allele-specific imprinting mark (Tremblay et al., 1997), was necessary for appropriate transgene imprinting: maternal expression and paternal hypermethylation (Elson and Bartolomei, 1997). Deletion of this element from an imprinted transgene resulted in a derivative transgene that was expressed and hypomethylated regardless of parental origin. Additional H19 transgenic studies revealed that a G-rich repetitive element located between the DMD and the promoter was not required for transgene imprinting (Stadnick et al., 1999). Importantly, related deletions at the endogenous locus, which were generated by homologous recombination in embryonic stem cells, exhibited behavior comparable to the transgenes (Thorvaldsen et al., 1998, and unpublished), thereby validating the transgenic model for investigating the minimal sequences that are required for imprinting. The first H19 transgene that displayed imprinted expression was 14 kb in length and included 4 kb of 5⬘ sequence, an internally deleted structural gene, and 8 kb of 3⬘ sequence within which the two endodermal enhancers are located (Bartolomei et al., 1993). To test the hypothesis that the differentially methylated sequence within this 14-kb transgene provided the crucial cis-acting imprinting signal, the differentially methylated sequence alone was used to generate a new transgene. The resulting transgene (RRSdBam), which harbored the 4 kb of upstream sequence and the inter-

The imprinted mouse H19 gene exhibits maternal allele-specific expression and paternal allele-specific hypermethylation. We previously demonstrated that a 14-kb H19 minitransgene possessing 5ⴕ differentially methylated sequence recapitulates the endogenous H19 imprinting pattern when present as high-copy arrays. To investigate the minimal sequences that are sufficient for H19 transgene imprinting, we have tested new transgenes in mice. While transgenes harboring limited or no 3ⴕ H19 sequence indicate that multiple elements within the 8-kb 3ⴕ fragment are required for appropriate imprinting, transgenes incorporating 1.7 kb of additional 5ⴕ sequence mimic the endogenous H19 pattern, including proper imprinting of low-copy arrays. One of these imprinted lines had a single 15.7-kb transgene integrant. This is the smallest H19 transgene identified thus far to display imprinting properties characteristic of the endogenous gene, suggesting that all cis-acting elements required for H19 imprinting in endodermal tissues reside within the 15.7-kb transgenic sequence. © 2001 Academic Press

INTRODUCTION

Genes that are subjected to genomic imprinting in mammals are expressed preferentially from a single parental allele (Bartolomei and Tilghman, 1997; Constaˆncia et al., 1998). While over 40 mammalian imprinted genes have been identified, the mechanism by which the parental alleles are distinguished and appropriately expressed remains poorly understood. Various elements, including replication timing, chromatin structure, DNA methylation, and repetitive sequences, have been proposed as conveying the allele-specific imprinting mark. Although differential DNA methylation is one of the stronger candidates, it is unlikely that this 1

Present address: Wellcome/CRC-Institute, Cambridge CB2 1QR, UK. 2 Present address: UCSF Cancer Center, Box 0808, San Francisco, CA 94143. 3 To whom correspondence should be addressed. Telephone: (215) 898-9063. Fax: (215) 573-6434. E-mail: [email protected].

0888-7543/01 $35.00 Copyright © 2001 by Academic Press All rights of reproduction in any form reserved.

98

H19 TRANSGENE IMPRINTING

nally deleted structural gene and was not expressed, was tested for methylation imprinting and found to be hypomethylated regardless of the parental origin of the transgene (Elson and Bartolomei, 1997). The lack of imprinting observed for this transgene has at least four plausible explanations. Two possibilities reflect the absence of the 3⬘ sequence harboring the endodermal enhancers: either expression of the transgene or a sequence in the 8 kb 3⬘ fragment may be required for appropriate transgene imprinting. The remaining two possibilities are an issue of size: the 6-kb RRSdBam transgene may be too small to establish an independent imprinting domain, suggesting that more generic H19 sequence or additional 5⬘ sequence may be required in the context of this smaller transgene to establish appropriate imprinting. Consistent with this last idea, the 5⬘ end of the original imprinted 14-kb transgene is 100 bp and 5 CpG dinucleotides shy of the 5⬘ border of the DMD (Tremblay et al., 1997). Thus, while the 14-kb transgene does not harbor the entire DMD, those sequences that are included appear to be sufficient to confer imprinting to this transgene. However, in the context of the smaller RRSdBam transgene, it is conceivable that additional 5⬘ sequence is required. To address the question of the minimal sequence that is sufficient for H19 imprinting we have tested new transgenic constructs. To assay the requirement for 3⬘ sequence we have derived two new constructs and found that at least a subset of elements 3⬘ to the H19 transcription unit can restore imprinting to the transgene. Furthermore, two transgenes that include additional 5⬘ flanking sequence closely approximate the imprinting pattern of the endogenous gene. Specifically, the majority of lines are imprinted, including a line with only one copy of the transgene, a finding that has not previously been demonstrated with minitransgenes. One construct harboring additional 5⬘ sequence but lacking 3⬘ sequence revealed further complexity in imprinting. While low-copy lines were imprinted under all conditions, high-copy lines failed to reverse a hypermethylation imprint when passed through the female germ line. We conclude from this analysis that multiple sequences at the H19 locus are required for faithful recapitulation of the endogenous imprinting pattern. MATERIALS AND METHODS Preparation and microinjection of transgene DNA. The transgenic DNA was purified and prepared for microinjection as previously described (Elson and Bartolomei, 1997). Microinjection was performed by Dr. Jean Richa at the University of Pennsylvania Core Transgenic Facility. DNA was injected into one of the pronuclei of fertilized one-cell mouse eggs derived from (C57BL/6 ⫻ SJL) F 1 intercrosses. At least 150 eggs were injected with RRSdBam/AFP and XXRSdBam and at least 300 eggs were injected with RRS2e and XXRSRdBam. Founder animals were identified by Southern blot analysis of tail DNA and outbred to DBA/2J mice (The Jackson Laboratory) to maintain the transgenic lines. The number of transgenic founders obtained relative to live born animals are as follows (construct indicated in parentheses): 7/56 (RRSdBam/AFP), 10/52

99

(RRS2e), 7/67 (XXRSRdBam), and 8/41 (XXRSdBam). All animals that transmitted the transgene were analyzed and are shown in Table 1. Preparation of genomic DNA and total RNA. Liver or tail tissue samples were incubated overnight at 55°C in 500 ␮l of digestion buffer (50 mM Tris–HCl, pH 8.0; 100 mM EDTA; 0.5% SDS) with 0.25 ␮g proteinase K (Boehringer Mannheim). NaCl was added to a final concentration of 150 mM, and the mixture was twice extracted with phenol:chloroform:isoamyl alcohol (25:24:1). DNA was ethanol precipitated and resuspended in TE (10 mM Tris–HCl, pH 7.6; 1 mM EDTA). Sperm were isolated from the cauda epididymis and incubated at 55°C in 500 ␮l of sperm digestion buffer (10 mM Tris–HCl, pH 8.0; 10 mM EDTA; 2% SDS) with 0.25 ␮g proteinase K and 144 mM ␤-mercaptoethanol. Sperm DNA was prepared as for genomic DNA. Animals were genotyped by PCR as previously described (Elson and Bartolomei, 1997). RNA from 5-day-old mice was prepared by the lithium chloride method as previously described (Auffray and Rougeon, 1980). S1 nuclease protection assay. The S1 nuclease assay probe was isolated and labeled as previously detailed (Elson and Bartolomei, 1997). The assay was performed as previously described (Malim, 1995) with the following modifications. Prior to hybridization, samples were denatured at 75°C (PP probe). Hybridization was carried out at 55°C. Samples were electrophoresed on a 1.0-mm-thick denaturing acrylamide gel containing 7 M urea, between 11.5 and 13 V/cm. Gels were fixed for 15 min in a solution containing 10% methanol and 10% acetic acid (v/v), dried, and exposed Kodak XAR-5 X-ray film or Phosphor Image screens (Molecular Dynamics). Methylation analysis. To assess methylation status at the 5⬘ end of the RRSdBam/AFP and RRS2e transgenes, each sample of liver DNA was digested with BamHI to liberate a 2-kb fragment from the 5⬘ end of the transgene and with BamHI plus HpaII. The 5⬘ 4.4 kb of the XXRSRdBam and XXRSdBam was assayed by digesting with BstXI and with BstXI plus HpaII. To assess methylation within the gene body, DNA was digested with PvuII (RRSdBamAFP, RRS2e, and XXRSRdBam), to release a 3.4-kb fragment encompassing the H19 coding region, and with PvuII and HpaII. For XXRSdBam, the gene body methylation was appraised by digesting with HindIII and with HindIII plus HpaII (the transgene-specific band is 2.9 kb). Ten micrograms of each DNA sample was digested and run on a 1% agarose gel with 1⫻ TBE (0.089 M Tris–HCl, 0.089 M boric acid, 0.004 M EDTA). DNA was transferred to nitrocellulose via Southern blotting (Southern, 1975) and hybridized to a 1.8-kb EcoRI–HindIII probe from the 5⬘ end of the transgene or a 3.0-kb gene body probe (RH and RS, respectively, Fig. 1). After being washed (Wahl et al., 1979), filters were exposed to Kodak XAR-5 X-ray film and Phosphor Image screens (Molecular Dynamics). Quantification of Southerns and S1 nuclease assays. Fixed and dried gels (S1 nuclease assays) and hybridized filters (Southern blots) were exposed to storage phosphor screens which were scanned on a Phosphor Imager Storm 840 (Molecular Dynamics). Gels were analyzed and relative band intensities were calculated using ImageQuant (Molecular Dynamics). For S1 nuclease assays, the volume of the band representing transgene expression for each sample was normalized to the volume of the band representing endogenous expression. For Southerns, the intensities of full-length bands in the lanes digested with HpaII were compared directly to the intensities of the same size bands in the non-HpaII-digested DNA lanes to determine the percentage of DNA molecules that were fully methylated. Note that the Phosphor Imager is sensitive through 5 orders of magnitude. Determination of transgene copy number. Ten micrograms of genomic tail or liver DNA was digested to completion with the appropriate enzyme to liberate a fragment that distinguishes the endogenous and transgenic copies (HindIII for RRSdBam/AFP, XbaI for RRS2e, PvuII for XXRSRdBam, XbaI and XhoI for XXRSdBam). The DNA was analyzed as described above, except that the 3.0-kb RS probe was used for hybridization.

100

CRANSTON ET AL.

RESULTS

H19 Transgene Expression Is Not Sufficient for Imprinting We previously identified H19 transgenes that approximated the endogenous H19 imprinting pattern; they were maternally expressed and exhibited paternal-specific methylation of the 5⬘ flanking sequence and the transcription unit (Bartolomei et al., 1993; Elson and Bartolomei, 1997). We have also previously shown that an H19 transgene (RRSdBam) that was lacking enhancers failed to display methylation imprinting; it was hypomethylated regardless of parental origin (Elson and Bartolomei, 1997). This transgene contained all of the differentially methylated sequence present in the original imprinted H19 transgene (RRSRdBam) but lacked 100 bp at the 5⬘ end of the H19 DMD, including 5 differentially methylated CpGs and the 3⬘ 8-kb fragment housing the endodermalspecific enhancers (Fig. 1). This loss of methylation imprinting could be attributed to a lack of specific 5⬘ or 3⬘ sequences, the size of the transgene, or absence of transgene expression. In the present study, we first tested the latter two possibilities with a transgenic construct that employed enhancers from the similarly expressed mouse ␣-fetoprotein gene (AFP). The AFP gene harbors three enhancers within a 7.5-kb XbaI fragment that are responsible for its high level transcription in fetal and neonatal liver (Godbout et al., 1988). The entire 7.5-kb fragment was cloned adjacent to the upstream EcoRI site, resulting in a derivative that is similar in size to the RRSRdBam transgene (RRSdBam/AFP, Fig. 1). Four independent transgenic lines with copy numbers ranging from 2 to 32 were identified (Table 1), bred to DBA/2J mice, and tested for allele-specific expression. The assay of RNA from two sequential generations of mice revealed that all lines exhibited expression of the transgene regardless of whether it was maternally or paternally inherited (Fig. 2A and data not shown). The methylation status of the RRSdBam/AFP transgenes was also tested. For all four lines, both the 5⬘ flanking sequence and the transcription unit were hypomethylated regardless of parental origin (data not shown). One conclusion from the analysis of the RRSdBam/AFP transgene is that it is not imprinted and expression of the transgene alone is not sufficient for imprinting. These results are consistent with previously described H19 derivative transgenes that were hypomethylated and expressed regardless of parental origin (Bartolomei et al., 1993; Brenton et al., 1999; Elson and Bartolomei, 1997; Pfeifer et al., 1996). Additionally, the absence of imprinting of RRSdBam/AFP transgenes also suggests that transgene size is not a critical determinant of imprinting since RRSdBam/ AFP is similar in size to the imprinted RRSRdBam transgene.

A Transgene Harboring a Truncated Enhancer Fragment Exhibits Variable Imprinting Another possible explanation for the lack of imprinting and differential methylation observed for the RRSdBam transgene could be the absence of critical sequence in the 3⬘ 8-kb SalI to EcoRI fragment where the endodermal-specific enhancers are located. To test this possibility, we generated transgenic mice with a new transgene containing a 2.5-kb NsiI–BglII fragment with the two endodermal enhancers but lacking 4.5 kb 5⬘ and 1 kb 3⬘ of the enhancers (RRS2e, Fig. 1). Six lines were identified with 2 to 30 copies of the transgene (Table 1). When assayed for RNA expression and differential DNA methylation, these lines exhibited variable expression and methylation patterns. The two high-copy lines were imprinted: the maternal allele was expressed and the paternal allele was hypermethylated (Figs. 2C, 3C, and 3D). Line 201-22 was imprinted in all analyzed N 2 and N 3 progeny. In contrast, line 293-122 was imprinted in the N 3 but not the earlier N 2 generation. Additionally, upon paternal transmission, approximately 25% of the N 3 animals expressed the transgene (Fig. 2C, lane 20 and data not shown), which was also hypomethylated (Fig. 3D, lanes 7 and 8 and data not shown). The reason for the difference in imprinting between these two lines is unclear but the results suggest that the RRS2e transgene can attract the appropriate modifications that cause it to mimic the endogenous pattern of H19 imprinting. Examination of the four lower copy RRS2e transgenic lines revealed that these transgenes were variably expressed (three of the lines expressed the transgene) (Table 1). While two of the lines exhibited no imprinting, a third line (line 293-5) was imprinted opposite to the endogenous H19 gene (Table 1 and data not shown). Since this transgenic line behaved opposite to all other imprinted H19 transgenic lines, its imprinting was probably not a consequence of specific H19 sequence. Rather, this line is likely to fall into the class of transgenes that is imprinted due to the site of integration (Allen and Mooslehner, 1992; Chaillet, 1994). When all RRS2e lines are considered together, the results indicate that sequence located between the transcription unit and the enhancers and/or the 1 kbfragment 3⬘ of the enhancers is important for imprinting. In the absence of this sequence, however, multiple copies of the transgene facilitate imprinting. Transgenes with Additional 5⬘ Sequence Are Imprinted Absence of critical 5⬘ sequence may also prevent the RRSdBam transgene from acquiring parental-specific methylation as well as the lower copy RRSRdBam and RRS2e transgenic lines from being appropriately imprinted (Elson and Bartolomei, 1997). To test this hypothesis, two new transgenic constructs were cloned that contained an additional 1.7 kb of 5⬘ H19 sequence immediately adjacent to the EcoRI site that delimits

H19 TRANSGENE IMPRINTING

101

FIG. 1. Endogenous H19 gene locus and transgenes. (A) The top line corresponds to 16.1 kb of the endogenous H19 gene locus. The five-exon endogenous H19 structural gene, spanning ⬃2.8 kb, is represented by the black box and the 3⬘ endoderm-specific enhancers (Yoo-Warren et al., 1988), situated at ⫹8 and ⫹9.5 kb relative to the transcription initiation site (arrow), by black ovals. The gray box on the top line indicates the location of the G-rich repetitive element (Stadnick et al., 1999). The hatched box designated “DMD” corresponds to the differentially methylated domain: the region of the endogenous H19 gene that displays paternal-specific methylation and mediates the imprinting of the H19 and Igf2 genes (Thorvaldsen et al., 1998; Tremblay et al., 1997). The filled circles beneath the DMD correspond to CTCF sites. The striped boxes represent genomic fragments used for DNA and RNA (S1 nuclease assay) probes (see Materials and Methods). The positions of key restriction endonuclease sites relative to the start of transcription are indicated (in kb) at the top. (B) All transgenic constructs contain a 1-kb deletion (referred to as dBam) spanning the 3⬘ half of exon 1, all of exon 2, and the 5⬘ half of exon 3, as well as the intervening introns. The RRSRdBam and RRSdBam constructs have 3.8 kb of sequence 5⬘ of the start of transcription and were previously described and analyzed (Brunkow and Tilghman, 1991; Elson and Bartolomei, 1997). RRSdBam/AFP has the sequence present in the RRSdBam transgene plus an additional 7.5-kb XbaI fragment (shaded box) harboring the three enhancers from the AFP gene (filled circles) (Godbout et al., 1988). XXRSRdBam contains 5.5 kb of sequence upstream of the transcription start site, a truncated gene body, and 8 kb of 3⬘ sequence flanking the gene body. The 5⬘ end of this transgene is the XbaI site at ⫺5.5 kb. XXRSdBam is identical to XXRSRdBam except that it lacks sequences 3⬘ of SalI and the polyadenylation signal. RRS2e was derived by cloning a 2.5-kb NsiI–BglII fragment with the two H19 endodermal enhancers 3⬘ of the SalI site in RRSdBam. Constructs without enhancers do not express H19 and were assayed for methylation imprinting only. The size of each transgene is indicated in parentheses. The BstXI sites used for the methylation analysis of XXRSRdBam and XXRSdBam and the BamHI sites used for the analysis of RRSdBam/AFP and RRS2e are indicated on each transgene. The 5⬘ BamHI site in these last two constructs is located in polylinker sequence adjacent to the ⫺3.8 kb EcoRI site. The restriction endonuclease sites, including those whose positions are used to designate the various transgene constructs, are BamHI (B), EcoRI (R), SalI (S), XbaI (X), BglII (Bg), NsiI (N), HindIII (H), BstXI (Bx), and PvuII (P). Note that only a subset of these restriction endonuclease sites is indicated on the top line. Not drawn to scale.

102

CRANSTON ET AL.

TABLE 1 Imprinting Status of Mouse H19 Transgenes Construct

Line

Copy No.

Expression a

Methylation a

RRSdBam/AFP

110-6 110-28 110-18 110-40 201-22 201-2 201-18 293-122 293-5 293-29 158-16 220-22 220-22 c 228-3 228-8 228-14 228-30 228-13 228-2 228-5

32 20 5 2 30 5 2 30 2 2 5 7 1 42 40 18 22 5 4 3

M, P M, P M, P M, P M M, P — M, P; M b P M, P M M M — — — — — — —

— — — — P — ND —; P M — P P P Pd Pd M, P P P P P

RRS2e

XXRSRdBam

XXRSdBam

a

M, maternal allele only; P, paternal allele only; M, P, both alleles; —, neither allele; ND, not determined. b Transgene in the N 2 progeny of F 1 animals was expressed and hypomethylated regardless of parental origin. The majority of N 3 progeny imprinted the transgenic DNA. c One transgenic animal in the 220-22 line underwent a germ-line deletion of the transgenic array resulting in a single-copy integrant [confirmed by genomic DNA analysis (data not shown)]. This animal was used to generate additional single-copy transgenic animals for imprinting analyses. d While progeny of transgenic males exhibit hypermethylation of the transgene and progeny of transgenic females exhibit hypomethylated transgenes when the parents are derived from F 1 and N 2 females, the progeny of females derived from a transgenic male do not reverse the male hypermethylation imprint (i.e., progeny from these females exhibit hypermethylated transgenes, Fig. 4).

the 5⬘ end of the three above transgenes (Fig. 1). Two lines from the first construct, XXRSRdBam, were identified and both were imprinted: the maternally derived transgenes were expressed and the paternally derived transgenes were hypermethylated (Table 1). Expression data and methylation data of the 5⬘ sequence for line 220-22 (seven copies of the transgene) are shown in Figs. 2B and 3A, respectively. Interestingly, an animal from line 220-22 underwent a germ-line deletion of the transgenic array that resulted in one full-length copy of the XXRSRdBam transgene remaining at the locus (data not shown). This rare event allowed us the unique opportunity to assess and compare the imprinting of a seven-copy array and a single-copy integrant of the XXRSRdBam transgene at the same locus. The analyses of mice from the derivative line revealed that the single-copy line also exhibited appropriate imprinting patterns (Fig. 3B). Importantly, in addition to displaying imprinted expression, all XXRSRdBam transgenic mice exhibited copy-number-dependent expression (Fig. 2B and data not shown). Taken together these results demonstrate that the 15.7-kb

transgene is the smallest transgene characterized thus far to recapitulate the endogenous pattern of H19 imprinting. This transgene likely harbors all of the elements crucial to imprinting the H19 gene in endodermal tissues. A second transgene, XXRSdBam, also had an extra 1.7 kb of sequence at the 5⬘ end but this transgene lacked the 3⬘ 8-kb fragment including the endodermal enhancers. As with the previously tested RRSdBam transgene, we assayed imprinting by measuring parental-specific methylation (Elson and Bartolomei, 1997). Seven lines harboring between 3 and 42 copies of the transgene were tested for differential methylation. All except one line were hypermethylated on the paternally inherited transgene and hypomethylated on the maternally inherited transgene (Table 1, Figs. 3E and 3F, and data not shown). Line 228-14 was methylated regardless of parental origin of the transgene (data not shown). These results show that additional 5⬘ sequence restored methylation imprinting to the nonimprinted RRSdBam transgene in the absence of H19 expression. Furthermore, unlike previous H19 transgenes that appeared to be required in higher copy arrays for appropriate imprinting (Elson and Bartolomei, 1997; Pfeifer et al., 1996), the XXRSdBam transgene was differentially methylated when present in as few as 3 copies, supporting results with the XXRSRdBam transgene that suggest that transgenes with the entire DMD serve as better models for H19 imprinting. Reversibility of the Imprint A hallmark of imprinting is that a parental allele will reverse its imprint when passed through the germ line of the opposite-sex parent. Thus, an allele with a maternal-specific imprint will acquire a paternal-specific imprint when passed through the germ line of the father and vice versa. Since we routinely carry our transgenes in females (i.e., the N 2 generation is derived from a transgenic F 1 female), we repeatedly demonstrated the reversal of the maternal-specific imprint (acquisition of an inactive transcriptional state and hypermethylation) when propagated by a male. It was, therefore, of interest to determine whether a paternally inherited transgene will become hypomethylated and capable of being expressed when transmitted through the female germ line. We previously demonstrated reversibility for the RRSRdBam transgene (Bartolomei et al., 1993; Elson and Bartolomei, 1997; and data not shown). In the present study, one RRS2e line (201-22) and one XXRSRdBam line (line 220-22) were tested and both exhibited reversal of imprinting (data not shown). Analysis of three XXRSdBam transgenic lines revealed a less consistent pattern. Male N 2 progeny derived from F 1 females were used to generate male and female N 3 transgenic mice. The N 3 mice, which harbored hypermethylated transgenes, were subsequently bred to DBA/2J animals and their progeny were tested

103

H19 TRANSGENE IMPRINTING

FIG. 2. Expression analysis of H19 transgenes. Total liver RNA, isolated from 5-day-old mice, was analyzed by the S1 nuclease assay using the PP probe (Fig. 1A). The transgenic progeny from male and female F 1 or N 2 mice (as indicated above the lanes) from three RRSdBam/AFP (A), one XXRSRdBam (7 copies, B), and two RRS2e (C) lines were analyzed for endogenous and transgenic H19 expression. Endogenous expression (endog) is used as a loading control and migrates at a position corresponding to 343 bp. Probe indicates selfhybridized probe (DNA:DNA hybrids) and migrates at 397 bp. The transgenic (tg) band migrates at 119 bp. Probe DNA was run on lanes marked P. The level of transgene expression for each sample was quantified relative to the endogenous H19 expression and reported as the ratio of the transgene to endogenous gene expression. The ratios are as follows: (A) 2.9 (lane 1), 2.8 (lane 2), 3.0 (lane 3), 5.0 (lane 4), 2.4 (lane 5), 4.9 (lane 6), 0.7 (lane 7), 1.0 (lane 8), 1.4 (lane 9), 1.1 (lane 10), 2.4 (lane 11), 3.0 (lane 12), 1.9 (lane 13), 1.9 (lane 14), 2.4 (lane 15), 1.7 (lane 16); (B) 7.0 (lane 1), 8.0 (lane 2), 4.0 (lane 3), 0.03 (lane 4), 0.01 (lane 5), 0.01 (lane 6), 0.02 (lane 7); (C) 5.7 (lane 1), 6.5 (lane 2), 5.7 (lane 3), 6.2 (lane 4), 1.6 (lane 5), 0.4 (lane 6), 1.0 (lane 7), 2.0 (lane 8), 0.9 (lane 9), 1.4 (lane 10), 1.7 (lane 11), 0.3 (lane 12), 2.2 (lane 13), 1.6 (lane 14), 1.1 (lane 15), 5.8 (lane 17), 4.8 (lane 18), 3.5 (lane 19), 4.8 (lane 20), 0 (lane 21), 0.6 (lane 22), 0.02 (lane 23).

for methylation. Animals from the low-copy line 228-2 exhibited the appropriate reversal of transgene imprinting, with hypomethylated maternally derived transgenic DNA and hypermethylated paternally derived transgenic DNA (Fig. 4A). In contrast, the same breeding protocol for the high-copy line 228-8 resulted in animals that did not reverse the parental imprint. Progeny from both male and female N 3 transgenic parents displayed a significant degree of methylation that was characteristic of a paternally transmitted transgene (Fig. 4B). Similarly, the high-copy line 228-3 did not reverse the paternal-specific hypermethylation when passed through the female germ line (data not shown). Passage of this transgene through two additional, sequential generations of females did not result in hypomethylation of the transgene.

Thus, the XXRSdBam transgene does not likely harbor sequence that enables high-copy arrays to reverse a paternal hypermethylation imprint. Since high-copy arrays of transgenes with 3⬘ sequence do reverse paternal hypermethylation, these results suggest that the 3⬘ sequence is required for such an activity. DISCUSSION

The first H19 transgene (RRSRdBam) exhibiting imprinted gene expression similar to that of the endogenous H19 gene was 14 kb in length and included approximately 4 kb of 5⬘ flanking sequence, the H19 structural gene, and 8 kb of 3⬘ flanking sequence. While high-copy arrays of this and similar transgenes were expressed and hypomethylated after maternal

104

CRANSTON ET AL.

FIG. 3. Methylation analysis of H19 transgenes. Genomic DNA from the livers of transgenic neonates was subjected to Southern analysis for investigation of transgene methylation. The DNA was digested with BstXI (A, B, E, and F) or BamHI (C and D) to liberate the 5⬘ portion of the transgene or endogenous genes (lanes designated by “⫺”). Samples designated by a “⫹” at the top of the lanes were also digested with the methylation-sensitive restriction enzyme HpaII to assess the methylation status of the 5⬘ fragment. Three HpaII sites are assayed in the 2.0-kb transgenic BamHI fragment and 11 HpaII sites were assayed in the 4.4-kb transgenic BstXI fragment. Blots were probed with the RH genomic fragment depicted in Fig. 1. The transgenic line is indicated above each blot and the sex and generation of the transmitting transgenic parent are indicated above each lane. The locations of transgene (tg) and endogenous gene (endog, if shown) fragments that are not digested by HpaII are indicated. For each animal, the intensities of the transgenic BamHI (2.0 kb) or BstXI (4.4 kb) band with and without HpaII digestion were compared to determine the relative methylation level of the transgenic array. The percentage methylation for each HpaII sample is indicated at the bottom of each blot. Similar methylation levels were found for the gene body portion of the H19 transgenes (not shown).

transmission, and repressed and hypermethylated after paternal transmission, low-copy transgenes were not imprinted (Bartolomei et al., 1993; Elson and Bartolomei, 1997; Pfeifer et al., 1996). In contrast, a 130-kb H19 yeast artificial chromosome clone was appropriately imprinted regardless of copy number (Ainscough et al., 1997). Thus, larger H19 transgenes are more likely to mimic the endogenous imprinting pattern. To determine the minimal sequences required to recapit-

ulate faithfully H19 imprinting, we have constructed a new series of transgenes. These experiments reveal that the imprinting of the H19 gene is a complex process involving multiple elements both 5⬘ and 3⬘ of the H19 transcription unit. Transgenes that incorporate these sequences are differentially methylated and expressed, usually reverse their imprints when passed through the opposite-sex germ line, exhibit imprinting in multiple strain backgrounds including DBA/2J and

H19 TRANSGENE IMPRINTING

105

C57BL/6J (data not shown), and are imprinted when present as low-copy arrays. The addition of 1.7 kb of 5⬘ sequence to the 4 kb in RRSRdBam produced transgenes that closely approximated endogenous H19 imprinting. The XXRSRdBam and XXRSdBam transgenes, which include and lack 8 kb of 3⬘ flanking sequence where the endodermal enhancers reside, respectively, harbor all of the differentially methylated sequence at the endogenous locus and were imprinted in eight of nine independent lines. The eight imprinted lines had between 1 and 42 copies of the transgene. Previous transgenes utilizing the EcoRI site in RRSRdBam as the 5⬘ boundary of the transgene lacked the 5⬘ 100 bp of the DMD and required at least 4 copies for imprinted expression, with the lower copy imprinted lines showing a greater number of deviations from imprinted transgene expression (Elson and Bartolomei, 1997; Pfeifer et al., 1996). Furthermore, the 15.7-kb transgene, XXRSRdBam, is the first H19 transgene smaller than 100 kb to display copy-number-dependent expression. One interpretation of these results is that the extra 5⬘ sequence provides regulatory elements crucial to H19 transcription and imprinting. Consistent with this proposal, the active maternal allele of H19 is hypersensitive to nucleases throughout the DMD (Hark and Tilghman, 1998; Kanduri et al., 2000; Khosla et al., 1999; Szabo et al., 2000). While the sequence corresponding to the 5⬘ hypersensitive sites is deleted from the RRSRdBam transgene, the XXRSRdBam and XXRSdBam transgenes include this sequence. It is also possible that the 5⬘ sequence acts as an insulator that blocks negative transcriptional effects from regulatory elements adjacent to the integration site (Pikaart et al., 1998). A boundary or insulator function has been proposed for the H19 DMD at the endogenous locus: on the maternal H19 allele, the DMD blocks access of the H19 endodermal enhancers to the linked and oppositely imprinted Igf2 gene, whereas on the paternal allele the DMD is methylated and cannot function as an insulator, thereby allowing the Igf2 gene exclusive access to the enhancers (Thor-

FIG. 4. Imprint reversibility of two XXRSdBam transgenic lines. The top portion depicts the breeding strategy used to generate animals for the methylation analyses of lines 228-2 (A) and 228-8 (B). Females are represented by circles, males by squares, and mice of unknown sex by diamonds. Striped symbols indicate the hemizygous transgenic animals and the open symbols correspond to DBA/2J mice. The generation is indicated to the left of the pedigree. Only the

animals relevant for this study are shown. The animals that were subject to the methylation analysis in the bottom portion are indicated by an arrow to the left of the N 4. Each of these N 4 animals was derived from an N 3 parent who inherited a hypermethylated transgenic array from a hemizygous transgenic N 2 male. The ability of the N 4 animals to reverse the hypermethylation of the transgenic array was assayed by digesting with HindIII and the methylation-sensitive enzyme HpaII and hybridizing the blot (bottom half) with the RS probe (Fig. 1). This digestion assays the gene body; similar results were found with the upstream sequence (data not shown). Digestion with HindIII alone results in a predominant 3.0-kb fragment for the transgene (tg) and a 4-kb fragment for the endogenous gene (endog). For each animal, the intensities of the transgenic HindIII band with (⫹) and without (⫺) HpaII digestion were compared to indicate the relative methylation level of the transgenic array. These numbers are indicated at the bottom . Note that for line 228-8 the F 1 generation and animals propagated through the female part of the pedigree (not shown) are appropriately imprinted.

106

CRANSTON ET AL.

valdsen et al., 1998; Webber et al., 1998). The hypothesis that the DMD serves to insulate the H19 promoter and enhancer from adjacent regulatory sequences is supported by recent experiments demonstrating that this region does indeed function as an insulator in transient transfection and transgenic mouse assays (Bell and Felsenfeld, 2000; Hark et al., 2000; Kaffer et al., 2000; Kanduri et al., 2000). For example, when a plasmid with the DMD inserted between the human ␥-globin promoter/neomycin-resistance gene and a mouse globin enhancer is transfected into cultured cells, the DMD blocks enhanced transcription of the neomycin-resistance gene (Bell and Felsenfeld, 2000). The insulator activity is mediated through four short conserved elements located within the DMD (Frevel et al., 1999; Stadnick et al., 1999; and Fig. 1A). The ubiquitous transcriptional regulator CTCF binds to these elements in a methylation-sensitive manner (Bell and Felsenfeld, 2000; Hark et al., 2000). Interestingly, the XXRSRdBam and XXRSdBam transgenes have all four CTCF binding sites, whereas the RRSRdBam transgene lacks the 5⬘-most element, again supporting the proposal that the two former transgenes harbor additional information that is critical to H19 imprinting and expression. While the two transgenes incorporating the extra 5⬘ sequences display many of the imprinting properties characteristic of the endogenous H19 gene, XXRSdBam high-copy lines did not reverse the paternal-specific hypermethylation imprint when passed through the maternal germ line. The failure to reverse this hypermethylation is likely due to the inability of the germ cells to demethylate the transgenic DNA since female 13.5-dpc PGCs displayed a significant amount of methylation in the 5⬘ portion of the transgene (data not shown). In contrast, we have shown that the endogenous H19 gene is completely demethylated at this same time in both male and female PGCs (Davis et al., 2000). Although it is unclear why the methylation imprint is not erased in high-copy XXRSdBam transgenic lines, the demethylation failure may be due to the absence of 3⬘ H19 sequences since other high-copy imprinted transgenic lines that do reverse paternal-specific imprints harbor sequences 3⬘ of the polyadenylation signal. Alternatively, when present in high-copy arrays, the transgene with either the four CTCF sites together or the entire DMD may affect the reversal of the imprint. It should be noted that heritability of epigenetic modifications has been observed in other systems such as at the agouti locus in mouse (Morgan et al., 1999). The imprinting observed for lines harboring the RRS2e transgene suggests that sequence 3⬘ of the polyadenylation signal is required for appropriate transgene imprinting when less than the entire DMD is present. There are two primary reasons for this conclusion. First, compared to the nonimprinted RRSdBam transgene, which lacks 3⬘ sequence,

RRS2e, which includes a 2.5-kb 3⬘ fragment containing the two endodermal enhancers, exhibits imprinting in high-copy arrays, indicating that either the enhancers or the sequence surrounding the enhancers promotes imprinting. Second, RRSRdBam, which, unlike RRS2e, contains the 4.5-kb fragment between the polyadenylation signal and the enhancers, exhibits more robust and consistent imprinting behavior in mice. Thus, it is likely that the 4.5 kb of sequence also harbors information that is critical to imprinting. This proposal is supported by a comparison of the sequence between mouse and human. In addition to the similarity of the endodermal enhancers between the two orthologous genes, the mouse and human H19 genes share two blocks of sequence in the 4.5-kb region (Ishihara et al., 2000). In conclusion we have defined sequences both 5⬘ and 3⬘ of the H19 transcription unit that are crucial to H19 imprinting. Inclusion of sequence from the 5⬘ border of the DMD which incorporates all four CTCF binding sites results in imprinting of low-copy transgenic arrays while inclusion of sequence centering around the endoderm enhancers promotes imprinting of high-copy arrays in the absence of the above sequence. Further experiments utilizing both the transgenic system and homologous recombination at the endogenous locus are required to refine the relevant sequences as well as the machinery that targets these elements. ACKNOWLEDGMENTS We thank J. Richa and the University of Pennsylvania Transgenic Core Facility for the production of transgenic mice. We also thank members of the laboratory for critically reading the manuscript. This work was supported by U.S. Public Service Grant GM51279 and the Howard Hughes Medical Institute.

REFERENCES Ainscough, J. F.-X., Koide, T., Tada, M., Barton, S., and Surani, M. A. (1997). Imprinting of Igf2 and H19 from a 130 kb YAC transgene. Development 124: 3621–3632. Allen, N. D., and Mooslehner, K. A. (1992). Imprinting, transgene methylation and genotype-specific modification. Semin. Dev. Biol. 3: 87–98. Auffray, C., and Rougeon, F. (1980). Purification of mouse immunoglobulin heavy-chain messenger RNAs from total myeloma tumor RNA. Eur. J. Biochem. 107: 303–314. Bartolomei, M. S., and Tilghman, S. M. (1997). Genomic imprinting in mammals. Annu. Rev. Genet. 31: 493–525. Bartolomei, M. S., Webber, A. L., Brunkow, M. E., and Tilghman, S. M. (1993). Epigenetic mechanisms underlying the imprinting of the mouse H19 gene. Genes Dev. 7: 1663–1673. Bell, A. C., and Felsenfeld, G. (2000). Methylation of a CTCF-dependent boundary controls imprinted expression of the Igf2 gene. Nature 405: 482– 485. Brenton, J. D., Drewell, R. A., Viville, S., Hilton, K. J., Barton, S. C., Ainscough, J. F.-X., and Surani, M. A. (1999). A silencer element identified in Drosophila is required for imprinting of H19 reporter transgenes in mice. Proc. Natl. Acad. Sci. USA 96: 9242–9247. Brunkow, M. E., and Tilghman, S. M. (1991). Ectopic expression of the H19 gene in mice causes prenatal lethality. Genes Dev. 5: 1092–1101.

H19 TRANSGENE IMPRINTING Chaillet, J. R. (1994). Genomic imprinting: Lessons from mouse transgenes. Mutat. Res. 307: 441– 449. Constaˆncia, M., Pickard, B., Kelsey, G., and Reik, W. (1998). Imprinting mechanisms. Genome Res. 8: 881–900. Davis, T. L., Yang, G. J., McCarrey, J. R., and Bartolomei, M. S. (2000). The H19 methylation imprint is erased and reestablished differentially on the parental alleles during male germ cell development. Hum. Mol. Genet. 9: 2885–2894. Elson, D. A., and Bartolomei, M. S. (1997). A 5⬘ differentially methylated sequence and the 3⬘ flanking region are necessary for H19 transgene imprinting. Mol. Cell. Biol. 17: 309 –317. Frevel, M. A. E., Hornberg, J. J., and Reeve, A. E. (1999). A potential imprint control element: Identification of a conserved 42 bp sequence upstream of H19. Trends Genet. 15: 216 –218. Godbout, R., Ingram, R. S., and Tilghman, S. M. (1988). Fine-structure mapping of the three mouse ␣-fetoprotein gene enhancers. Mol. Cell. Biol. 8: 1169 –1178. Hark, A. T., Schoenherr, C. J., Katz, D. J., Ingram, R. S., Levorse, J. M., and Tilghman, S. M. (2000). CTCF mediates methylationsensitive enhancer-blocking activity at the H19/Igf2 locus. Nature 405: 486 – 489. Hark, A. T., and Tilghman, S. M. (1998). Chromatin conformation of the H19 epigenetic mark. Hum. Mol. Genet. 7: 1979 –1985. Ishihara, K., Hatano, N., Furuumi, H., Kato, R., Iwaki, T., Miura, K., Jinno, Y., and Sasaki, H. (2000). Comparative genomic sequencing identifies novel tissue-specific enhancers and sequence elements for methylation-sensitive factors implicated in Igf2/H19 imprinting. Genome Res. 10: 664 – 671. Kaffer, C. R., Srivastava, M., Park, K.-Y., Ives, E., Hsieh, S., Batle, J., Grinberg, A., Huang, S.-P., and Pfeifer, K. (2000). A transcriptional insulator at the imprinted H19/Igf2 locus. Genes Dev. 14: 1908 –1919. Kanduri, C., Holmgren, C., Pilartz, M., Franklin, G., Kanduri, M., Liu, L., Ginjala, V., Ulleras, E., Mattsson, R., and Ohlsson, R. (2000). The 5⬘ flank of mouse H19 in an unusual chromatin conformation unidirectionally blocks enhancer-promoter communication. Curr. Biol. 10: 449 – 457. Khosla, S., Aitchison, A., Gregory, R., Allen, N. D., and Feil, R. (1999). Parental allele-specific chromatin configuration in a boundary/imprinting-control element upstream of the mouse H19 gene. Mol. Cell. Biol. 19: 2556 –2566. Malim, M. H. (1995). Cellular assays for Rev. In “HIV,” Vol. II, “A Practical Approach” (J. Karn, Ed.), pp. 187–209, Oxford Univ. Press, Oxford.

107

Morgan, H. D., Sutherland, H. G. E., Martin, D. I. K., and Whitelaw, E. (1999). Epigenetic inheritance at the agouti locus in the mouse. Nat. Genet. 23: 314 –318. Pfeifer, K., Leighton, P., and Tilghman, S. M. (1996). The structural gene of H19 is required for transgene imprinting. Proc. Natl. Acad. Sci. USA 93: 13876 –13883. Pikaart, M. J., Recillas-Targa, F., and Felsenfeld, G. (1998). Loss of transcriptional activity of a transgene is accompanied by DNA methylation and histone deacetylation and is prevented by insulators. Genes Dev. 12: 2852–2862. Southern, E. M. (1975). Detection of specific sequences among DNA fragments separated by gel electrophoresis. J. Mol. Biol. 98: 503– 517. Stadnick, M. P., Pieracci, F. M., Cranston, M. J., Taksel, E., Thorvaldsen, J. L., and Bartolomei, M. S. (1999). Role of a 461 bp G-rich repetitive element in H19 transgene imprinting. Dev. Genes Evol. 209: 239 –248. Szabo, P. E., Tang, S.-H., Rentsendorj, A., Pfeifer, G. P., and Mann, J. R. (2000). Maternal-specific footprints at putative CTCF sites in the H19 imprinting control region give evidence for insulator function. Curr. Biol. 10: 607– 610. Thorvaldsen, J. L., Duran, K. L., and Bartolomei, M. S. (1998). Deletion of the H19 differentially methylated domain results in loss of imprinted expression of H19 and Igf2. Genes Dev. 12: 3693–3702. Tremblay, K. D., Duran, K. L., and Bartolomei, M. S. (1997). A 5⬘ 2-kilobase-pair region of the imprinted mouse H19 gene exhibits exclusive paternal methylation throughout development. Mol. Cell. Biol. 17: 4322– 4329. Wahl, G. M., Stern, M., and Stark, G. R. (1979). Efficient transfer of large DNA fragments from agarose gels to diazobenzyloxymethylpaper and rapid hybridization by using dextran sulfate. Proc. Natl. Acad. Sci. USA 76: 3683–3687. Webber, A. L., Ingram, R. S., Levorse, J. M., and Tilghman, S. M. (1998). Location of enhancers is essential for the imprinting of H19 and Igf2 genes. Nature 391: 711–715. Wutz, A., Smrzka, O. W., Schweifer, N., Schellander, K., Wagner, E. F., and Barlow, D. P. (1997). Imprinted expression of the Igf2r gene depends on an intronic CpG island. Nature 389: 745–749. Yoo-Warren, H., Pachnis, V., Ingram, R. S., and Tilghman, S. M. (1988). Two regulatory domains flank the mouse H19 gene. Mol. Cell. Biol. 8: 4707– 4715.