Comparative Genome Analysis of the Mouse Imprinted Gene Impact and Its Nonimprinted Human Homolog IMPACT: Toward the Structural Basis for Species-Specific Imprinting

RESULTS AND DISCUSSION

Identification of cDNA for Human IMPACT

Combinatorial use of human EST clones showing homology to mouseImpact, cDNA library screening, and rapid amplification of cDNA ends (RACE) cloning allowed us to deduce a contiguous sequence 3683 bp long for the putative human homolog of Impact. To verify this reconstructed structure, we designed oligonucleotide primers from this sequence and used them for RT-PCR. We readily obtained PCR products of expected sizes, which were subjected to direct cycle sequencing to eliminate the effect of potential misincorporation during PCR. The confirmed sequence was deposited in the database under the accession number AB026264. We assume that it is a full- or nearly full-length structure, because it is fairly coincident with the length of the cognate transcript estimated by Northern blot hybridization (see below) and because clones extended further in the 5′ direction were obtained by neither library screening nor RACE.

The cDNA bears an open reading frame of 960 bp, and can encode a protein composed of 320 aa (amino acids) (Fig.1A). The predicted product shows significant homology with that of Impact (82.2% identity, 96.6% similarity) as well as its Xenopus homologXimpact (Yamada et al. 1999) (60.0% identity, 87.5% similarity). The predicted product also has high homology with members of the YCR059c/yigZ hypothetical protein family, or the UPF0029 family, composed of proteins of unknown functions (Fig. 1B). As shown in Figure2B, the region termed B is found in hypothetical proteins from various bacteria and Arabidopsis, and is strikingly conserved among the members. In contrast, the eukaryote-specific regions A and D are less conserved, although the region A contains a GI domain, which has recently been found to mediate a specific protein–protein interaction (Kubota et al. 2000). The cDNA is homologous to that of Impact not only in the ORF but also in the 3′ UTR (total: 60.4 %). In particular, the 180-bp region derived from the 3′ extremity is 94.4% homologous to that of mouseImpact. Such a high homology suggests a role for this segment in, for instance, post-transcriptional processing or regulation. Judging from these features, we designated the gene from which this cDNA is derived as IMPACT, for the human homolog ofImpact.

Chromosomal Localization of Human IMPACT

To determine the chromosomal locus of IMPACT, the cDNA clone was used as a probe for fluorescence in situ hybridization (FISH). As shown in Figure 2, clear doublet signals were unequivocally detected on the proximal region of chromosome 18 (Fig. 2A), but not on any other chromosome. Judging from the banding pattern of the same metaphase spread (Fig. 2B), the locus of IMPACT was determined to be chromosome 18q11.2–12.1 (Fig. 2C). This region is syntenic to mouse chromosome 18 A2–B1, the locus to which mouse Impactwas mapped (Hagiwara et al. 1997). These results also support the hypothesis that IMPACT is the ortholog of Impact.

Tissue Distribution of Human IMPACT mRNA

To examine the tissue distribution of IMPACT mRNA, we performed Northern blot hybridization using a probe derived from its ORF. The probe detected two messages: One is ∼3.9 kb long, showing good coincidence with the cDNA, and the other is ∼2.1 kb (Fig.3A). Both are detected in all the tissues examined and display an identical tissue preference pattern (Fig. 3A). We found that a 3′-UTR probe detected only the longer RNA (Fig. 3B) and thus assume that the shorter RNA is generated through differential polyadenylation.

Although a modest tissue preference was observed in the distribution ofIMPACT mRNA (Fig. 3), its expression is basically ubiquitous. This is in marked contrast with mouse Impact, which is preferentially expressed in adult brain (Hagiwara et al. 1997). These results may raise the possibility that the clone we obtained is derived from a paralog rather than the ortholog of Impact. However, the striking structural conservation observed not only in the ORF but also in the 3′ UTR, the syntenic localization, and the well-conserved exon–intron organization (see below) support our idea that IMPACT is the ortholog of Impact. Also, every effort to find other homologs of Impact has been, so far, unsuccessful. We thus assume that the IMPACT reported here is the human ortholog of mouse Impact.

Allelic Expression of Human IMPACT

To determine the allelic expression status of IMPACT, nucleotide polymorphisms must be found in the transcribed region. We thus sequenced the 3′ UTR of IMPACT amplified from genomic DNAs from seven Caucasian, seven African-American, and six Asian subjects. Consequently, we found four single nucleotide polymorphisms (SNPs) in this region: 1155 C/T, 2070 T/G, 2180 A/G, and 3104 G/T nucleotide variants.

The C/T heterozygotes at position 1155 (1155 C/T) were not found in the 48 Japanese examined by direct sequencing (data not shown). Similarly, screening by single-nucleotide allele-specific primer extension (SNAS) assay (Vu and Hoffman 1997) revealed only one C/T heterozygote (fetus no. 13527) from a total of 44 fetal subjects (not shown). Because this SNP was originally identified in three out of the seven African-Americans used in the initial screening, it may be enriched in this population. We thus screened peripheral blood DNAs from 56 adult African-Americans and identified five heterozygotes, but, owing to the low expression in peripheral blood leukocytes, our RT-PCR detected theIMPACT transcript in only two of the five informative cases. To examine the allelic expression of IMPACT in these three cases (i.e., one fetus and two adult blood samples), the SNAS assay was applied to the cDNA synthesized from each specimen. As shown in Figure4A, both C and T alleles were detected in fetal brain and adrenal gland (lanes 3, 5: brain; lanes 8, 10: liver; and lane 13: adrenal tissues). No detectable PCR products were observed in the negative controls (Fig. 4A, lanes 4, 6, 8, 11, 14). Thus, humanIMPACT was expressed equally from both parental alleles in these fetal tissues. Similarly, both parental alleles were expressed in one adult blood sample (Fig. 4A, lanes 16, 19) with no detectable bands in negative controls (Fig. 4A, lanes 17, 20). However, in the other blood specimen derived from another adult, the C alleles may appear to be preferentially expressed (Fig. 4A, lane 19), taking the faint T band that appeared even in the control reaction into account (Fig. 4A, lane 21). Although the allelic expression at lower expression level has to be carefully evaluated, this might indicate leaky imprinting ofIMPACT in this particular individual. Another possibility causing apparent preferential allele expression would be cell type-specific imprinting.

The 2070 T/G polymorphism was identified in one of the seven Caucasian samples by DNA sequencing. We performed the SNAS assay to genotype 48 DNA samples (including 44 fetuses) and found that all 48 were T/T homozygotes (data not shown). This SNP, therefore, appears to be quite rare. The 2180 A/G polymorphism that creates a polymorphicMfeI restriction site (CAATTG) was identified in two of the seven African-Americans by DNA sequencing. We used PCR-RFLP to genotype 44 fetuses and found one heterozygote (fetus no. 13527) and 43 A/A homozygotes (data not shown). This fetus also had the 1155 C/T polymorphism that had already been analyzed by the SNAS assay described above.

The 3104 G/T polymorphism that creates a polymorphic ApoI restriction site (AAATTT) was also identified in two of the seven African-American DNA samples. By PCR-RFLP analysis, we identified two heterozygotes (fetuses nos. 13466 and 13713) and 42 G/G homozygotes. The fetus no. 13466, whose tissues were available, was analyzed for the allelic expression. As shown in Figure 4B (lanes 1–7), all the tissues examined, including brain, lung, heart, testis, limb, intestine, and adrenal gland, demonstrated biallelic expression of humanIMPACT. In addition, the maternal tissue (endometrium, lane 8) also showed biallelic expression. No detectable PCR products were observed in the mock negative controls (−RT, data not shown).

It should be noted that variations in the relative levels of the two alleles were observed in these tissues (Fig. 4B, lanes 1–7). The control heterozygous genomic DNA also showed a predominance of undigested G allele (Fig. 4B, lane 9). Although the restriction digestion was complete (data not shown), heteroduplexes refractory to the digestion were often observed in PCR-RFLP, which can explain the predominant G allele in the G/T DNA control (Fig. 4B, lane 9) and in various tissues (Fig. 4B, lanes 1–3, 5, 7–8). However, two tissues (testis and intestine: Fig. 4B, lanes 4, 6) showed a predominant expression of the digested T allele. Unfortunately, the maternal subject was also a 3104 G/T heterozygote (Fig. 4B, lane 8), and hence we could not determine the parental origin of the digested T allele.

It is conceivable that the imprinting of IMPACT is leaky and polymorphic, at least, in some tissues or cell types. In this context, it is interesting to note that we identified two rare variant forms ofIMPACT cDNA, each of which bears a unique 5′ end but is expressed much less abundantly. Notably, the polymorphic sites used in the allelic expression studies are shared by both variants, and hence all the isoforms are assayed collectively. Accordingly, the allelic expression status of these minor variants are masked by that of the major transcript. It is thus interesting to examine the allelic expression status of each isoform. Unfortunately, despite our exhaustive screening, we have failed to find any isoform-specific SNP, and hence such studies are currently hampered.

Genome Organization of Mouse Impact and Human IMPACT

The data described above show that the human IMPACT gene is basically expressed in a biallelic manner, although an allelic bias may be observed occasionally in some tissues. This is in good contrast with the mouse Impact gene, which is expressed exclusively from the paternal allele in all the tissues examined (Hagiwara et al. 1997). Because the two genes encode highly conserved proteins, they may well share a common genome organization, and hence the comparison between the two may readily pinpoint structural elements crucial for genomic imprinting. We thus determined complete nucleotide sequences for these genes by means of bacterial artificial chromosome (BAC) cloning (Shizuya et al. 1992) and our unique nested deletion strategy (Hattori et al. 1997). Consequently, we elucidated 37,954 bp of contiguous sequence for mouse Impact (accession no. AF232228) and 29,644 bp for human IMPACT (accession no. AF232229).

The genome structures of mouse Impact and humanIMPACT are depicted in Figure 5with the minimum contigs of the subclones. Alignments of the genome sequences with those of cDNAs revealed that both genes have 11 exons. The average size of the exons is about 100 bp except for the last one, which contains the termination codon and is longer than 2 kb. All of the splice junctions follow the GT–AG rule (Table1) and split the open reading frames at the identical positions between the two species. Thus, the overall genome organizations of these genes are well conserved, thereby providing further evidence for their orthology.

A remarkable difference between the two genes was found in their upstream promoter regions. The promoter region and the first exon of human IMPACT constitute the sole CpG island of this gene. In contrast, the corresponding region of the mouse gene is rather AT rich (43% GC). Although the ratio of observed versus expected CpG dinucleotides of this region is 0.35, which is significantly higher than the average for the ∼38-kb region (0.25), it does not meet the criteria for a CpG island.

Instead, mouse Impact has a CpG island in its first intron. The intronic island has many TCGGC sequences and a characteristic tandemly reiterated structure (Fig. 6). It is known that such tandem repeats often associate with imprinted genes (Constancia et al. 1998; Feil and Khosla 1999). Notably, we failed to find any such tandemly repeated structures either in the CpG island or elsewhere in the nonimprinted human IMPACT.

Parent-of-Origin-Specific Methylation of the Mouse CpG Island

Because tandem repeats often occur in imprinted genes and are implicated in the establishment of genomic imprinting (Constancia et al. 1998; Feil and Khosla 1999), we examined the methylation status ofIMPACT. For this, we prepared two parental mouse strains,Mus musculus domesticus C57BL/6 (B6) and M. musculus molossinus JF1 (JF), and reciprocal F₁ hybrids between the two as described previously (Hagiwara et al. 1997). Then we amplified and sequenced the CpG islands from B6 and JF to search for polymorphisms between the two mouse strains. Fortunately, the island of B6 is 181 bp longer than that of JF, owing to the difference in repeat organization. This allows us to discriminate between the B6 and JF alleles simply by their lengths.

To examine the methylation status, we developed a methylation-specific PCR assay. In this assay, genomic DNAs are first digested with methylation-sensitive restriction endonucleases such as HhaI or HpaII, and then used for PCR to amplify the locus of interest. Although unmethylated targets are cut by the enzymes and will not be amplified, a methylated target survives the digestion to serve as the template for subsequent PCR. In other words, the methylated allele is amplified. Because both B6 and JF alleles for the CpG island of Impact share the same five methylation-sensitive restriction sites, namely, two HhaI sites and threeHpaII sites, we can apply the methylation-specific PCR assay to this island.

We digested the genomic DNAs from B6, JF, (B6 × JF) F₁, and (JF × B6) F₁ with HhaI, HpaII, orMspI and used them as the templates for PCR spanning the intronic island (Fig. 7A). When native undigested genomic DNAs of the F₁ hybrid mice were used as the templates, we readily obtained two bands derived from B6 and JF alleles that can be clearly separated by gel electrophoresis. When the DNAs treated with HhaI or HpaII were used for the PCR, only one of the two bands was obtained. The B6 allele was amplified from (B6 × JF) F₁, whereas only the JF allele was detected from (JF × B6) F₁ (Fig. 7A). When we usedMspI, a methylation-insensitive isoschizomer ofHpaII, as a control, we could not amplify any bands at all. These results clearly demonstrated that the island is methylated in a parent-of-origin-dependent manner—the silenced maternal allele is hypermethylated, and the active paternal one is undermethylated. Thus, the island serves as a differentially methylated region (DMR) for this gene (Fig. 7C).

We next examined the methylation status of the promoter region, which bears two HpaII sites at positions −55 and −14. We readily amplified the expected DNA fragments from theHpaII-digested genomic DNAs derived from the parental strains and reciprocal F₁ hybrids (data not shown). This indicated that the region is at least partially methylated. We thus sequenced the amplified fragments or the methylated allele to know their parental origins by SNP at the position −355 (not shown). When the amplified fragments from undigested genomic DNAs of F₁ hybrid mice were used as the templates, we detected both T and C at this position, representing B6 and JF alleles, respectively (not shown). In contrast, we can detect only T from the HpaII-digested DNA from (B6 × JF) F₁ and only C from (JF × B6) F₁(not shown). These results indicate that the imprinted maternal allele is methylated, and the expressed paternal one is not methylated, like the CpG island (Fig. 7C). We also applied a similar assay to the regions flanking this gene and found that these sites are methylated on both alleles (not shown; see Fig. 7C). Whether the promoter region and the intronic island comprise a single DMR remains to be elucidated through methylation study of this region in early embryonic stages, which is currently underway.

Finally, we applied the methylation-specific PCR assay to the human CpG island, which spans the promoter and the first exon. In contrast to the mouse intronic island and promoter, not only MspI digestion but also HhaI or HpaII treatment completely abolished amplification of the human island, thereby suggesting its undermethylation (Fig. 7B). Furthermore, Southern blot hybridization analyses with the CpG island probe revealed that HpaII digests the island to tiny fragments as efficiently as MspI does (data not shown). These results indicate that the whole island is unmethylated on both chromosomes as are conventional CpG islands (Gardiner-Garden and Frommer 1987).

Taking all these data together, it is clear that the CpG island and the promoter region are subjected to parent-of-origin-dependent methylation and constitute DMRs in mouse, whereas those of the human counterpart are not. The intronic DMR is of particular interest, because it has a characteristic tandemly repeated structure, which is found in many other imprinted genes and is missing in the nonimprinted human counterpart. Although DMR has been implicated in genomic imprinting, its precise role is still controversial; some DMRs have been shown to be necessary for imprinted gene expression, but others seem not to be (Constancia et al. 1998; Feil and Khosla 1999). To elucidate the role of this DMR in genomic imprinting of Impact, we have to examine the temporal coincidence in the establishment of its differential methylation and the monoallelic expression ofImpact. We would be able to undertake much more straightforward tests using transgenic and gene targeting techniques, because the basis for such experiments has already been laid by this study through the elucidation of the genome structure.