INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Within the mammalian genome, more than 40 imprinted genes have been identified (for an up-to-date list, see www.mgu.har.mrc.ac.uk/imprinting/). Imprinted genes show parental-allele-specific expression such that only one of the two parental copies present in a diploid cell is expressed. The hallmark of an imprinted gene is the presence of an imprint that ultimately allows the cellular transcription machinery to discriminate between the two parental alleles. Because imprinting can occur in inbred mice with genetically identical parental alleles, imprints cannot be acquired in diploid cells but instead must be set when the parental alleles are in physically separate compartments, during germ cell development or before nuclear fusion in the zygote. In addition, the imprint must be epigenetic, stably inherited through mitosis in embryonic development, and reset upon germ line formation (for a review, see references 7, 28, and 35). Several possibilities exist for an imprint mark, such as chromatin composition, organization, and histone acetylation or methylation state (21, 24). However, DNA methylation is by far the best candidate as it fulfils all the desired criteria; in particular, de novo methylation patterns that are differently acquired by the male and female gamete could function as the imprinting mark. Support for a role of DNA methylation comes from the demonstration that mice lacking a functional maintenance methyltransferase gene (Dnmt1) show a loss of imprinted expression, observed as either biallelic silencing or biallelic expression (13, 16, 25, 26). All tested imprinted genes have been associated with allelic methylation differences in somatic tissue, and some also show gamete-specific methylation (26, 31, 36). Finally, for three imprinted genes, deletion of the sequences carrying the methylation imprint resulted in loss of imprinting, thereby identifying these methylated sequences as imprint control elements (8, 29, 34, 39, 40). Methylation in the mammalian genome is present on CpG dinucleotides, and DNA elements that carry a methylation imprint can be defined as CpG islands, since they contain clustered CpGs and are associated with active transcription. However, methylated CpG island promoters are unusual, as the majority of CpG island promoters are unmethylated regardless of expression state (3).

The reason why imprinted CpG islands attract a methylation imprint during gametogenesis has not yet been resolved, but two models have been proposed. First, specialized sequences are proposed to exist within the imprinted CpG island that either prevent or attract de novo methylation activity in one gamete. An allele discrimination signal (ADS) has been proposed to exist in the insulin-like growth factor type 2 receptor gene (Igf2r) imprinted intron 2 CpG island that attracts gamete-specific shielding factors, thereby protecting a nearby de novo methylation signal (DNS) from methylation. In this model, the primary allele discrimination is not de novo methylation in the oocyte, but the shielding that prevents methylation in spermatozoa. De novo methylation can subsequently occur later in gametogenesis or even after fertilization (9). Second, an alternative sequence-independent model proposed that the presence of direct repeats that have also been found in several imprinted CpG islands attracts methylation in the gamete as a response to unusual secondary structure formation. The ability of methylation to target these repeats in imprinted CpG islands was suggested to have evolved from a host defense response to suppress genomic parasitic DNA elements (5, 20, 38).

In order to identify genes involved in the imprint methylation pathway, the DNA sequences necessary and sufficient to attract and maintain the methylation imprint need to be characterized. Targeted deletions from the endogenous locus can be used to identify essential elements necessary for methylation, but only transgenes can identify the minimal configuration sufficient for the imprint. Such a transgenic approach has been successful in the identification of the minimally imprinted configuration for the maternally expressed H19 and paternally expressed insulin-like growth factor type 2 (Igf2) neighboring loci. Results have shown that large 130-kb transgenes that contain both the H19 and Igf2 genes as well as 140-kb transgenes that contain only the H19 gene are reliably imprinted (1, 15). H19 transgenes as small as 10 kb can be imprinted, but mostly as multicopy transgenes and with less than 100% frequency (12, 14, 23). These results indicate that an imprinted H19 transgene demands multiple elements in addition to the sequences that carry the methylation imprint, the H19 promoter, and the downstream situated enhancers. The G-rich repetitive element located upstream of the H19 promoter is, however, dispensable (12, 14, 23, 30). A transgenic approach has also been applied to the paternally expressed Snrpn gene, and results show that 75- to 85-kb multicopy transgenes containing the entire Snrpn gene and flanking DNA can be imprinted (10, 27). A minimal imprinted configuration has also been delineated on multicopy transgenes as a dual element of 1.2 kb composed of a 200-bp minimal Snrpn promoter coupled to a 1.0-kb human imprint regulatory element located 35 kb upstream of the human SNRPN promoter (27).

The maternally expressed Igf2r gene differs from the above imprinted genes, since it contains, in intron 2, a promoter generating a paternally expressed imprinted antisense RNA (Air) (17) that overlaps the Igf2r promoter (see Fig. 1 for an overview of the imprinted Air-Igf2r locus). This Air RNA is an unusual genomic transcript that is noncoding with no or few introns. An imprinted configuration for Igf2r has been identified on 300-kb transgenes that contain the whole 87-kb Igf2r gene and the 108-kb Air RNA plus additional flanking genes. Igf2r contains a methylation imprint inherited from the oocyte on a 3.7-kb fragment known as region 2 that is located within intron 2 (31). Region 2 contains direct repeats (20), an ADS-DNS motif (9), and the CpG island Air promoter (17). Region 2 is necessary for Igf2r imprinting in a transgene situation, as deletion results in Igf2r expression from both parental alleles (39), and is also necessary for imprinting the endogenous locus (40). In contrast, region 2 is not sufficient for Air imprinting, because short multicopy Air transgenes that contain all or part of region 2 are not imprinted by methylation (39). Imprinting of small Air transgenes might therefore demand multiple elements, similar to H19 and Snrpn transgenes.

View larger version (13K): [in this window] [in a new window]   FIG. 1.   The maternal and paternal endogenous (wt) Air-Igf2r loci are represented by the first two lines. The Air and Igf2r promoters are CpG island promoters (ellipses), and the arrows indicate the transcription starts. The methylated CpG islands are shown as stippled ellipses. Region 2 is presented as a line with the ADS-DNS motif depicted with an asterisk. The Igf2r exons are shown as numbered black boxes. Transgenic constructs REP, LAI, SAI, RAA, and FAA are derived from the endogenous Air-Igf2r locus. Construct REP contains a 9-kb Air fragment (including Igf2r exons 3 through 5) coupled to an EGFP reporter (stippled box), SV40 intron, and polyadenylation cassette (black triangle). The LAI construct spans 44 kb, from the Igf2r promoter up to Igf2r exon 7, and contains an EGFP-tagged (stippled box) Igf2r exon 1. 5' SV40 polyadenylation cassette and 3' rabbit -globin polyadenylation cassettes are shown as black triangles. The small Air-Igf2r SAI construct is similar to LAI, but lacks Igf2r intron 1 and is shorter at the 3' end, lacking Igf2r exons 6 and 7. Constructs RAA and FAA contain a complete mouse Aprt gene with a rabbit -globin polyadenylation cassette inserted at the 5' end (black triangles), the endogenous Aprt polyadenylation signal is used 3'. The Aprt exons are shown as open boxes. In construct RAA, the Air promoter is inserted in an antisense orientation with respect to the Aprt promoter, whereas in construct FAA the Air promoter is inserted in a sense orientation. The probes used for methylation analyses (LA, MC, MS, SB, and SE) and for expression analyses (SVA, GFPAIR, GFPIGF2R, MlMs1, PS, and DE) are shown. For details of the constructs and the probes, see Materials and Methods.

Here, we have tested whether stable Air expression and/or the presence of the Igf2r promoter and Igf2r expression are sufficient for imprinting of small Air transgenes. Our results show that in contrast to multicopy H19 transgenes, multicopy Air-expressing transgenes are not imprinted. However, the Air promoter can, albeit at a low frequency, recapitulate imprinting on a fragment as small as 4.6 kb when embedded in an adenine phosphoribosyltransferase (Aprt) transgene and can, in one instance, transfer imprinting onto the Aprt gene. Interestingly, including the Igf2r CpG island promoter and all flanking sequences up to the Air promoter does not increase the imprinting frequency for Air above that seen in combination with the Aprt CpG island promoter. Thus, this study indicates that a dual-promoter setting may facilitate Air imprinting and that elements for reliable Air imprinting are contained outside a 44-kb region spanning the Air-Igf2r promoters but inside the previously defined 300-kb region.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

DNA constructs. (i) REP. Construct REP (for reporter) consists of a 9.1-kb SpeI fragment (nucleotide bp 120364 to 132177, GenBank accession number AJ249895) containing region 2 isolated from cosmid 940PS (31) ligated into the BglII site of reporter plasmid pEGFP-1 (Clontech). The pEGFP-1 reporter has two modifications. An internal ribosome entry site sequence was introduced into the SmaI site, and a simian virus 40 (SV40) intron obtained as a 179-bp XhoI/NotI fragment from plasmid pCMV (Clontech) was inserted into the NotI site of pEGFP-1. The 10.8-kb construct was excised from the vector by SpeI/SspI digestion.

(ii) LAI. Construct LAI (for large Air-Igf2r construct) was built in a modified pBeloBAC-II vector (Research Genetics) that allowed excision of the final 44-kb LAI construct with RsrII/BsiWI digestion. LAI consists of a 40-kb NotI fragment containing region 2 (bp 98069 to 138490; AJ249895) from cosmid OT1 (17) ligated to a 4.0-kb EcoRI/NotI fragment (bp 94101 to 98069; AJ249895) containing the Igf2r promoter isolated from cosmid 3L (31). A 59-bp AgeI/NotI fragment (bp 98010 to 98069; AJ249895) from Igf2r exon 1 has been replaced in frame by a 735-bp AgeI/NotI fragment of reporter pEGFP-1 (Clontech). In a BglII site (bp 94139; AJ249895) upstream to the Igf2r promoter, a 198-bp BamHI fragment from pCMV (Clontech) containing the SV40 polyadenylation signal was introduced in an antisense orientation to terminate the Air transcript. The 3' end of the construct contains a 1-kb HincII fragment from the rabbit -globin gene, which includes part of intron 2, and complete exon 3 with its polyadenylation signal to terminate the Igf2r transcript.

(iii) SAI. Construct SAI (for small Air-Igf2r construct) was built in the modified pBeloBAC-II vector and excised as a 20-kb RsrII/BsiWI fragment. It consists of the H/X-14 construct (39), which is a HindII/XhoI fragment containing region 2 (bp 133005 to 118712; AJ249895), and is ligated to the same enhanced green fluorescent protein (EGFP)-tagged EcoRI/NotI Igf2r promoter fragment that was used in construct LAI. The 5' and 3' polyadenylation cassettes were identical to that in the LAI construct.

(iv) RAA and FAA. Constructs RAA (for reverse Air-Aprt) and FAA (for forward Air-Aprt) contain a 5-kb NcoI fragment with the entire mouse Aprt gene (a gift from Harry Vrieling, Leiden University). A 1.2-kb EcoRI fragment from rabbit -globin, including part of exon 2, complete intron 2, and exon 3 containing a polyadenylation signal, was inserted in an antisense orientation into the EcoRV site (bp 108; M11310) upstream of the Aprt promoter. A 4.6-kb SpeI/BglII fragment (bp 120364 to 127575; AJ249895) from cosmid 940PS containing region 2 was inserted into the HindIII site (bp 1694; M11310) of Aprt intron 2 in an antisense (in construct RAA) or sense (in construct FAA) orientation with respect to the Aprt promoter. Both constructs were released from the pBluescript II vector (Stratagene) by BssHII/NotI digestion.

Transgenic mice. Transgenic mice were generated by injecting DNA into FVB/N-fertilized oocytes by approved procedures (2). Approximately 300 oocytes injected with the REP construct generated 11 founder transgenes, 550 oocytes injected with the LAI construct generated 3 founder transgenes, 450 oocytes injected with the SAI construct generated 2 founder transgenes, 300 oocytes injected with the RAA construct generated 5 founder transgenes, and 200 oocytes injected with the FAA construct generated 9 founder transgenes. Transgenes were bred hemizygously onto an FVB/N background and identified by Southern blotting or PCR on genomic tail DNA.

DNA and methylation analyses. Genomic DNA was isolated by the sodium dodecyl sulfate-proteinase K procedure (2). The following oligonucleotides were used: for typing construct REP, GFPF (5'-CTGGTGAACCGCATCGAGCTGAA-3') and GFPAR (5'-ACCTCTACAAATGTGGTATGGCTG-3'); for constructs LAI and SAI, EFP1F (5'-GTAAACGGCCACAAGTTCAGC-3') and EFP2R (5'-GGTGTTCTGCTGGTAGTGGTC-3'); for construct FAA, APRTEX2F (5'-TATCTCGCCCCTCTTGAAAGACC-3') and M404F (5'-GTGACTCACTTTTGAGAAC-3'); for construct RAA, APRTEX3R (5'-CCATACTCCAGAGAATAGGAGGC-3') and M404F. Methylation analysis probes (see also Fig. 1) were as follows: for SB, a 723-bp HindIII/BglI fragment (bp 971 to 1694; M11310) from the Aprt gene; for SE, a 700-bp EcoRV/SmaI fragment (bp 108 to 808; M11310) upstream from the Aprt promoter; for LA, a 984-bp SacI/AgeI fragment (bp 97026 to 98010; AJ249895) from the Igf2r promoter; for MS, SfuI/MluI (bp 124992 to126086; AJ249895); for MC, MluI/BglII (bp 126086 to 127575; AJ249895). The digestion of methylation-sensitive enzymes was regularly monitored by hybridization to mitochondrial DNA as previously described (38). Methylation was quantified with a PhosphorImager (Fujix) and expressed as a percentage in multiples of 10. Transgene copy numbers were determined by comparing the transgene to the endogenous signal in a Southern blot with a PhosphorImager (Fujix). Single-copy lines were confirmed by hybridizing with an end fragment from the transgene.

RNA analyses. Total RNA was isolated with Tri Reagent (Molecular Research Center) according to the manufacturer's protocol. For the RNase protection assay (RPA) the RPAIII kit (Ambion) was used, and for the Northern blot assay the formaldehyde method was used (2). Expression analysis probes were as follows: SVA, a 350-bp HpaI fragment from construct REP, protects a 240-bp fragment from the SV40 polyadenylation signal; A3, a 252-bp XhoI/XbaI fragment (bp 2165 to 2417; M11310) protects Aprt exon 3 (134 bp); GFPIGF2R and GFPAIR, templates made by PCR with T3IGGFP (5'-AATTAACCCTCACTAAAGGGAAAACTTGTGGCCGTTTAC-3') and T7IGEX1 (5'-TAATACGACTCACTATAGGGAGTCACGGAGCGCCTCCTC-3') on construct LAI (281 bp); GFPAIR detects endogenous Air (185 bp) and transgenic Air (275 bp). GFPIGF2R detects endogenous (188, 163, 143, and 132 bp) and transgenic (278, 253, 233, and 222 bp) Igf2r RNA; MlMs1 is a MluI/MseI fragment (bp 126086 to 126293; AJ249895) (17) that detects unspliced (207, 171, and 148 bp) or spliced (112, 76, and 53 bp) Air RNA; Gapdh, templates generated by PCR with GAPDHF (5'-CGGGGTACCACAGCCGCATCTTCTTGTGCAG-3') and GAPDHR (5'-CGTCTAGATGGGTGGTCCAGGGTTTCTTAC-3'); PS, a 1.5-kb EcoRI/HindIII fragment (bp 1 to 1694; M11310) containing Aprt exons 1 and 2 (19); DE, a 1.3-kb HindIII fragment (bp 1694 to 3066; M11310) containing Aprt exons 3 to 5 (19).

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Single-promoter Air transgenes are expressed but not imprinted. Previously, it has been shown that short transgenes from 1 to 14 kb containing the intronic CpG island known as region 2 (see Fig. 1) do not carry a methylation imprint (39). These transgenes were not characterized for expression and may have lacked sequences needed for Air promoter activity and termination of the Air transcript. To test whether imprinting of Air requires stable Air transcription, we generated the REP construct that contains 9 kb, including region 2, coupled to an EGFP reporter, as described in Materials and Methods (Fig. 1). Five REP transgenic lines were generated and analyzed for Air imprinting upon hemizygous transmission through the paternal and maternal germ lines.

View larger version (47K): [in this window] [in a new window]   FIG. 2.   Imprinting analyses of transgenic lines REP-65 and REP-30. (A) Methylation analyses of the Air promoter. A Southern blot of tail genomic DNA digested with SacI ( lanes) or SacI and MluI (+ lanes) and hybridized with probe MS detecting endogenous (wt) and transgenic (tg) bands (sizes in kilobases are indicated) is shown. (B) Expression analyses of the transgenic Air promoter. An RPA was carried out of adult cardiac tissue total RNA with the transgene-specific probe (SVA) from the polyadenylation cassette in combination with an Aprt exon 3 probe (A3) as loading control. Lanes: i, input probes; c, tRNA hybridization to the probes; p and m, paternal and maternal transmission of the transgene, respectively; wt, a nontransgenic sample.

Low-frequency imprinting of Air-Igf2r transgenes. Previously, it has been shown that large, 300-kb transgenes that contain both the Air and Igf2r promoters were successfully imprinted in three of three sites when integrated into autosomes (39). During the same experiment, a related transgene with no Igf2r expression was inadvertently derived that failed to be imprinted. This result, combined with the failure of imprinting of small Air transgenes, suggested the possibility that normal Igf2r expression is a requirement for Air imprinting. To test if Igf2r expression is sufficient for imprinting of Air on short transgenes, we generated two differently sized constructs that contain both the Air and Igf2r promoters. The LAI construct is 44 kb and spans from 3.8 kb upstream of the Igf2r promoter up to Igf2r exon 7 and includes an EGFP-tagged Igf2r exon 1 (Fig. 1). The 20-kb SAI construct is similar to LAI but lacks the 20-kb Igf2r intron 1 and only extends up to Igf2r exon 5 (Fig. 1). Both constructs are flanked 5' and 3' by polyadenylation signals to terminate the transcripts from the Air and Igf2r promoters. Three LAI and two SAI transgenic lines were generated.

View larger version (72K): [in this window] [in a new window]   FIG. 3.   Imprinting analyses of transgenic lines LAI-48 and LAI-46. (A) Methylation analyses of the Air promoter by Southern blotting of adult spleen genomic DNA cut with HindIII ( lanes) or HindIII and SfuI (+ lanes) hybridized with probe MC detecting both endogenous (wt) and transgenic (tg) fragments (sizes in kilobases are indicated). Note that HindIII cuts inside the LAI construct and that LAI-46 shows an aberrant fragment due to a deletion of the 3' polyadenylation cassette. (B) Methylation analyses of the Igf2r promoter. A Southern blot of adult tail genomic DNA cut with SspI ( lanes) or with SspI and SalI (+ lanes) hybridized with probe LA detecting endogenous (wt) and transgenic fragments (tg) is shown (sizes in kilobases are indicated). (C ) Expression analyses of endogenous (wt) and transgenic (tg) Air RNA in cardiac tissue by RPA with probe GFPAIR that covers part of the EGFP-tagged Igf2r exon 1. (D) Expression of Air RNA in cardiac tissue with RPA probe MlMs1 that covers multiple transcription starts at the Air promoter, yielding multiple protected bands for both spliced and unspliced Air RNA. The transgenic line REP-27 is used as a control for spliced Air RNA. (E) Expression of Igf2r RNA detected by RPA with probe GFPIGF2R on cardiac (LAI-48) or embryonic 12.5-dpc (LAI-46) total RNA. Note that this probe covers multiple transcription starts of the Igf2r promoter and part of the EGFP tag, resulting in multiple protected fragments for both the endogenous (wt) and transgenic (tg) Igf2r RNA. The part of the LAI-46 line showing transgenic Igf2r RNA (top) has been overexposed compared to the the bottom panel, showing endogenous Igf2r RNA. Lanes: i, input probe; c, tRNA hybridization to the probe; p and m, paternal and maternal transmission of the transgene, respectively; wt, nontransgenic sample.

Low-frequency imprinting of Air-Aprt transgenes. The observation that Air can be imprinted when combined with the Igf2r promoter prompted us to test if nonimprinted CpG island promoters could also facilitate Air imprinting. Two constructs were generated that positioned the Air promoter in opposite orientations into a mouse Aprt transgene. Construct RAA contains the complete Aprt gene, polyadenylation signal, and promoter, plus a 4.6-kb fragment containing region 2 inserted into intron 2 of the Aprt gene so that the Air promoter faces the Aprt promoter and potentially generates an antisense RNA. Upstream of the Aprt promoter, a polyadenylation cassette was introduced to terminate Air RNA (Fig. 1). Five transgenic RAA lines were generated and investigated for imprinting of Air and Aprt.

View larger version (33K): [in this window] [in a new window]   FIG. 4.   Imprinting analyses of transgenes RAA-21 and RAA-14. (A) Methylation analyses of the Air promoter. A Southern blot is shown of genomic spleen DNA digested with EcoRV ( lanes) or EcoRV and MluI (+ lanes) and hybridized with probe MS detecting both endogenous (wt) and transgenic (tg) Air fragments (sizes in kilobases are indicated). (B) Methylation analyses of the Aprt promoter. A Southern blot is shown of spleen DNA cut with NcoI ( lanes) or NcoI and SmaI (+ lanes) and hybridized with probe SE, which detects both endogenous (wt) and transgenic (tg) Aprt fragments (sizes in kilobases are indicated). (C) Expression of the Aprt and Air promoters by Northern blot analyses of adult cardiac tissue total RNA. Probe PS consists of Aprt exon 1 and 2 (top) and detects both endogenous and transgenic Aprt RNA and transgenic Air RNA due to the reverse orientation of the Air promoter (see Fig. 1). The signal from transgenic Aprt is not visible at this exposure with this short probe. Probe DE consists of Aprt exon 3 to 5 (middle) and detects endogenous and transgenic Aprt RNA as indistinguishable bands. Transgenic Aprt expression from line RAA-14 was analyzed in an Aprt/ background. Gapdh was used as a loading control (bottom). Lanes: p and m, paternal and maternal transmission of the transgene, respectively; wt, nontransgenic sample.

/

/

View larger version (32K): [in this window] [in a new window]   FIG. 5.   Imprinting analyses of transgenes FAA-37 and FAA-4. (A) Methylation of the Air promoter was analyzed by Southern blotting with enzymes HincII ( lanes) or HincII and MluI (+ lanes) and hybridized with probe MC, which detects both endogenous (wt) and transgenic (tg) Air fragments (sizes are indicated in kilobases). (B) Methylation analyses of the transgenic (tg) and endogenous (wt) Aprt promoters. A Southern blot of spleen DNA cut with PstI ( lanes) or PstI and SmaI (+ lanes) and hybridized with probe SB is shown. Because PstI cuts outside the transgene at the 5' end, it generates differently sized fragments for each transgenic line and confirms the single-copy integration site. (C) Expression analyses of the Aprt and Air promoters by Northern blotting of total embryo (14.5 dpc) RNA. Note that due to the orientation of the Air promoter, probe PS consisting of Aprt exons 1 and 2 (top) only detects endogenous and transgenic Aprt RNA, whereas probe DE consisting of Aprt exons 3 to 5 (middle) now detects both transgenic and endogenous Aprt RNA and transgenic Air RNA. Line FAA-4 was analyzed in an Aprt/ background. Gapdh was used as a loading control (bottom). Lanes: p and m, paternal and maternal transmission of the transgene, respectively; wt, nontransgenic sample.

/

/

/

+/

/

+/

/

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

Air behaves as a nonimprinted CpG island promoter by default. Exploiting transgenes, we have investigated the minimal setting sufficient to imprint the Air promoter. Although all transgenes carried the ADS, the DNS, and the direct repeat array that have all been postulated to be involved in attracting the methylation imprint (9, 20), in 27 of 30 transgenic lines Air is not imprinted and has no or low-level methylation. The results presented here that examined transgene methylation either in mid-gestation embryos or in adult tissues do not discriminate between a failure to gain a methylation imprint and a failure to maintain the imprint. Despite this, these results may indicate that the default state for the Air CpG island is absence of methylation. If so, this contrasts with other experiments (9) that examined transgenic founders as pre- and postimplantation embryos and showed that a multimerized 113-bp fragment from the core of the Air promoter attracts and maintains a maternal-allele-specific methylation imprint up to blastocyst stages. Results from these early-embryo founder transgenes indicated that methylation was the default state of the maternal allele and that the paternal allele was protected from methylation. Our results that examined adult founders, F₁-F₂ embryos, and adults containing low- to high-copy-number transgenes offer support for the opposite proposalthat absence of methylation is the default state and that the maternal allele actively acquires its methylation imprint.

Air is not imprinted in a single-promoter setting. We have tested whether imprinting of Air requires stable Air expression with the REP reporter construct that contains the Air promoter on a 9-kb fragment in a single-promoter setting. This 9 kb is sufficient for Air to function as a promoter in an integration-independent manner but not enough for Air imprinting, as none of the 11 lines are imprinted. Considering seven additional transgenic lines ranging from 3 to 14 kb that contain the Air promoter but do not recapitulate the methylation imprint (39), we conclude that, as distinct from H19 and Snrpn (10, 12, 27), Air as a single expressed promoter in low- to high-copy-number transgenes is not sufficient to be imprinted.

Air can be imprinted in a dual-promoter setting. The hypothesis that Air imprinting might need an additional CpG island promoter arose from an observation that 300-kb Air-Igf2r transgenes that were inadvertently derived with an inactive Igf2r promoter also lacked imprinting for both Air and Igf2r (39). Considering these transgenes, we tested whether Igf2r promoter activity would be sufficient for Air imprinting with the LAI and SAI lines that contain both the Air and Igf2r CpG island promoters. In addition, we tested whether any nonimprinted CpG island promoter could establish an imprinted configuration for Air with the RAA and FAA lines that contain the Air promoter coupled to the Aprt promoter. We chose the mouse Aprt gene, as the Aprt promoter is a CpG island with a broad expression profile and thus similar to the Igf2r promoter. In addition, transgenic studies have shown that Aprt is unmethylated as a transgene but can be methylated and silent under some circumstances (11, 18, 19). In a dual-promoter setting, Air imprinting was established in 3 of 19 lines (16%), indicating that a second CpG island promoter may facilitate Air imprinting. The frequency of Air imprinting when coupled to the Igf2r promoter (1 in 5 lines) is comparable to the situation when the Air promoter is coupled to the Aprt promoter (2 in 14 lines). This shows that Air on a 4.6-b fragment can be imprinted in the absence of the Igf2r promoter and also indicates that the Aprt promoter is as good at facilitating Air imprinting as an Igf2r promoter. Moreover, the orientation of the Air promoter with respect to the Aprt promoter does not influence the result, since one in five lines in which both promoters face each other (the RAA lines) and one in nine lines with both promoters in the same orientation (the FAA lines) are imprinted for Air. Imprinting of the RAA-14 line is stable with paternal transmission but less robust with maternal transmission, adding further support to the proposal that the maternal methylation is the limiting step in imprinting and that the paternal absence of methylation is the default state. Comparable imprinted behavior has also been observed for a 300-kb Igf2r-Air transgene that integrated on the X chromosome (39) and for some H19 transgenes (12, 14).

Silencing of linked promoters by region 2. Imprinting of the Igf2r promoter depends on the presence of region 2 as deletion of the paternal region 2, including the unmethylated and expressed Air promoter, derepresses the paternal Igf2r promoter on transgenes (39) and in the endogenous locus (40). Because region 2 is present in all our transgenes, we investigated whether it is capable of silencing the linked Igf2r and Aprt promoters in the LAI-SAI and RAA-FAA lines. With line LAI-46, we have shown that imprinted expression of Air correlates with reciprocal expression of the linked Igf2r gene. Note that in contrast to the endogenous locus, the Air RNA in line LAI-46 is spliced. The functional significance of the absence of splicing at the endogenous locus is unclear, since at least for line LAI-46 splicing of Air is compatible with silencing of the linked Igf2r promoter. However, it remains unclear why in the LAI-313, LAI-48, and the SAI-119 lines an expressed unmethylated Air promoter does not silence the cis-linked Igf2r promoter as predicted by the expression-competition models proposed for sense-antisense pairs of imprinted genes (4). In the two cases of Air imprinting in an Aprt transgene, neither showed reciprocal effects on the cis-linked Aprt promoter. Although these numbers are too small for conclusions, they may indicate that not all CpG island promoters can be imprinted by region 2. In one Aprt transgene (line FAA-4), the linked Air and Aprt promoters have the same imprinted methylation and expression, in contrast to the endogenous locus, where Air and Igf2r are reciprocally expressed and methylated. Although the Air promoter in FAA-4 has imprinted Aprt, we argue that it has arisen from a mechanism different from that seen at the Air-Igf2r locus. A possible explanation is that the methylation imprint of the Air promoter has spread onto the Aprt promoter because the two CpG island promoters are closely located within a 2-kb fragment, due to the forward orientation of the region 2 insert in these FAA transgenes. Despite the fact that this transgene may not have recapitulated the mechanism that imprints Igf2r, it has generated an imprinted and paternally expressed Aprt transgene that rescues an Aprt-deficient phenotype in a paternal-allele-specific manner. This result indicates a possibility of imprinting other nonimprinted genes by proximity to the Air CpG island with potential application in creating parent-specific phenotypes.