A well-studied megabase-scale region of DNA on distal mouse chromosome 7 (Chr7) contains multiple genes subject to parental imprinting. Nearly all genes in this region show conserved synteny with genes on human Chr11p15.5. The extended imprinted region appears to have a bipartite structure in that it contains at least two separate imprinted sub-domains. Each of these subdomains is regulated by a distinct imprinting control element. These correspond to short differentially methylated DNA sequences (DMRs) – one immediately upstream of the H19 gene and another within an intron of the Kcnq1 gene. These two elements control, respectively, the allele-specific expression of the H19/Igf2/Ins2 gene cluster, and the second gene cluster containing Kcnq1, the antisense transcriptKcnq1ot1, p57^Kip2/Cdkn1c, Slc22a1l, Ipl/Tssc3 and possibly additional genes 12. There is good evidence assigning one border of the overall imprinted region to the DNA immediately downstream of H19 3. However, the other border, which must lie upstream of Ipl is less well-defined 4. Here we show evidence for weak imprinting of a gene, Tnfrh1, which is located upstream of Ipl in mice, and which encodes a putative decoy receptor in the TNF receptor family. We also map the tissue distribution of Tnfrh1 mRNA and discuss the significance of our findings in terms of possible functions of this gene in placentation.

Maps of the chromosomal region containing Tnfrh1 and Tnfrh2 and the exons of these genes are shown in Figure 1a,1b. These two genes have a similar intron/exon organization, suggesting that they arose by a local gene duplication 4. They are also 95 percent identical in their cDNA sequences, with 84 percent overall amino acid identity (Fig.2). The proteins encoded by Tnfrh1 and Tnfrh2 share a conserved arrangement of multiple cysteine residues, characteristic of the extracellular ligand-binding domains of other TNF receptor family members (Fig.2). Tnfrh1 and Tnfrh2 lie ~350 kb distant from the KvDMR1 'imprinting center'. Two other genes, Cars and Nap1l4, lie between the strongly imprinted gene cluster regulated by this DMR and the Tnfrh1/Tnfrh2 gene pair. The Cars gene is not functionally imprinted, and the Nap1l4 gene shows a weak expression bias in some mouse tissues, but not in several human tissues examined to date 45. The Obph1 gene, which shows a strong allelic bias in mRNA expression in the placenta, but not in other tissues, is located upstream of Tnfrh1. Since assessment of allelic bias in the placenta is complicated by the presence of maternal cells, it is not known with certainty whether Obph1 is imprinted. Searches of the genomic and expressed sequence databases using the Tnfrh1 nucleotide and protein sequences did not uncover any strongly similar genes in humans, and low stringency hybridization of Northern blots of human tissue RNAs with the Tnfrh1 cDNA probe did not yield specific bands. Moreover, in the interval of the human genome between CARS and the human orthologue of Obph1 (the OSBPL5 gene), there were no sequences with detectable similarity to Tnfrh1.

Previously, Engemann et al. described Tnfrh1 as having ubiquitous expression 4. Consistent with this, northern blotting using Tnfrh1-specific probes matching the relatively divergent 5' end of this gene showed easily detectable Tnfrh1 mRNA transcripts in nearly all adult tissues, as well as in late fetal organs and structures (Fig.3). However, by far the strongest signals were seen in the pregnant uterus and whole placenta (Fig.3). Measurements by PhosphorImaging showed that Tnfrh1 mRNA in whole placenta was 5-fold more abundant than in the non-pregnant uterus, and 10-fold more abundant than in the whole fetus. Two classes of transcripts were present: the smaller transcript, which migrated at 1.7 kb based on other blots with size standards, matches the size predicted from the cDNA sequence. The larger transcript, which migrates at 3.7 kb, must contain additional sequences, but these have not yet been mapped. To evaluate the relative expression of Tnfrh1 and Tnfrh2 in various organs and tissues, we performed reverse transcription polymerase chain reaction (RT-PCR) in multiplex reactions using a shared downstream primer and Tnfrh1- and Tnfrh2-specific upstream primers. This showed that both genes are expressed in most fetal and adult organs and structures, but that the expression in several structures including placenta and muscle is highly skewed in favor of Tnfrh1 (Fig.4).

The whole placenta, after dissection from the uterus in mid-gestation, contains both a fetal component (the placenta proper, consisting of the labyrinth, spongiotrophoblast and giant cell layers) and a maternal component (the decidua basalis, as well as maternal blood vessels). We were therefore interested to determine the precise localization of the Tnfrh1-positive cells in this organ. To this end we first dissected each of two frozen placentas, obtained at 12.5 dpc, into a superficial half, enriched in the superficial fetal components, and a deeper half, enriched in deep fetal and superficial maternal components, and extracted RNA from each half. This procedure crudely separates the labyrinthine trophoblast from all of the deeper tissues. Hybridizing the northern blot with a probe for Ipl mRNA (Fig.4) validated this procedure. Ipl is specifically expressed in the placental labyrinth 67, and this probe gave a strong signal only in RNA from the superficial halves. The same blot re-hybridized with the Tnfrh1 cDNA probe gave the opposite pattern, clearly showing that Tnfrh1 mRNA is restricted to layers of the placenta deep to the labyrinth (Fig.5).

To determine whether Tnfrh1 is expressed in the deep fetal component or, alternatively, in the maternal component of the placenta, we next carried out in situ hybridization with a digoxigenin-labeled Tnfrh1 cDNA probe. As shown in Figure 6a,6b,6c, this gave a detectable signal only in a narrow band of cells, situated immediately deep to the trophoblast giant cell layer. This restricted distribution of Tnfrh1-positive cells was observed both at 10.5 dpc, when the definitive placental layers have recently formed, and at 12.5 dpc, after these layers have begun to mature (Fig.6a,6b,6c). At high magnification, the Tnfrh1-positive cells were seen closely juxtaposed to trophoblast giant cells (Fig.6d). The morphology of these cells, and their location deep to the giant cell layer, suggested decidual parenchyma. These cells did not coincide with PAS-positive granulated uterine natural killer lymphocytes, or with CD3-positive infiltrating T-lymphocytes (data not shown). However, additional characterization will be necessary for a definitive assignment of cell type.

The expression of Tnfrh1 only at the fetal-maternal boundary obviously suggests that this gene may play a role in modulating either immune or trophic interactions between the invasive placental trophoblast and the uterine host tissue. Since Tnfrh1 encodes a receptor in the TNFR family lacking a cytoplasmic domain (i.e., a decoy receptor), this gene might function to block the action of TNF-related ligands. A similar scenario has been proposed by Phillips et al. 89 for another TNF ligand-receptor system, namely TRAIL and TRAIL receptors, which are expressed in human placentas. However, in their studies the major decoy receptor for TRAIL, DcR1, was found expressed in trophoblast, not in decidua, while another decoy receptor, DcR2, was expressed by placental macrophages. Fas ligand is also highly expressed by human and murine trophoblast, although its functional role in the placenta is unknown 1011.

To investigate potential imprinting of Tnfrh1 and Tnfrh2, we next searched for genetic polymorphisms in these genes that would allow us to distinguish maternal from paternal alleles in interspecific mouse crosses. We failed to identify polymorphisms in Tnfrh2, but we found multiple single-nucleotide polymorphisms (SNPs) in the Tnfrh1 gene (Table 2). Using cDNAs from fetal and adult tissues derived from interspecific crosses, we amplified a region of Tnfrh1 containing several of these SNPs. The PCR strategies generated products that crossed either 4 or 5 exon-exon boundaries, thereby eliminating the possibility of genomic contamination (Fig.1b). Since one of the SNPs that distinguished the standard C57BL/6 laboratory strain from the divergent strain M. m. castaneus (CAST) created an MboII restriction site in the CAST sequence, we used RFLP analysis to assess allelic representation in cDNAs from F1 progeny of BL/6 × CAST and CAST × BL/6 reciprocal crosses. This indicated an obvious but partial maternal bias in the fetal liver, and a nearly complete bias towards the maternal allele in placenta (Fig.7). This was confirmed by direct sequencing of the cDNAs, and the parent-of-origin dependence of this effect was emphasized by the opposite patterns observed in the reciprocal crosses (Fig.7). The bias in fetal liver must reflect parental imprinting of Tnfrh1, but based on the information from Northern blotting and ISH, the pattern seen in the placental cDNAs in these experiments reflects expression from maternal cells, not imprinting.

Other fetal and adult tissues were also assessed in these crosses, and showed a maternal expression bias, but weaker than that observed in fetal liver (data not shown). The parent-of-origin dependent allelic bias in Tnfrh1 mRNA was superimposed on a 'baseline' bias towards hyper-expression of the BL/6 allele (Fig.7). This most likely reflects the effects of polymorphisms in regulatory promoter and/or enhancer sequences. Indeed, sequencing of the BL/6 and CAST Tnfrh1 promoter sequences revealed multiple SNPs in the region -200 to +100 relative to the transcriptional start site (data not shown). The alternative trivial explanation for the baseline bias, polymorphisms within primer binding sites, was excluded by direct sequencing of genomic DNAs.

We next repeated the imprinting analysis using cDNAs from reciprocal crosses between BL/6 and another divergent strain, M. m. mollosinus (MOLD). Since the polymorphisms present in MOLD do not affect restriction sites, this experiment was done using SSCP to distinguish the alleles. The results again showed an incomplete bias towards expression of the maternal allele in several fetal and adult tissues, reflecting parental imprinting, as well as a stronger bias in the placenta, reflecting expression from maternal cells (Fig.8). Direct sequencing of cDNAs confirmed the SSCP data (Fig. 8). As was true of the BL/6 × CAST crosses, the parent-of-origin dependence of the allele-specific expression was superimposed on a baseline bias towards hyper-expression of the BL/6 allele. As was true of CAST, the MOLD Tnfrh1 promoter region differed from BL/6 at several SNPs, thereby providing a possible explanation for this constitutive bias (data not shown).

As a control for these assays, we assessed the allele-specific expression of the Obph1 gene, using the same cDNA preparations. Consistent with the previous report from Engemann et al. 4, this showed equal allelic representation in cDNAs from various fetal and adult tissues, and a strong maternal bias in the placenta (Fig.9). As mentioned above, additional work will be needed to determine whether the results in the placenta reflect imprinting of Obph1, or alternatively indicate expression from maternal cells.

The findings described here include 'leaky' but consistent imprinting of Tnfrh1 in several organs, and high-level expression of this gene in a distinctive population of cells restricted to the interface between the placental trophoblast and the uterine lining. The absence of identifiable orthologues of Tnfrh1 and Tnfrh2 in the human genome suggests that both of these genes arose subsequent to the divergence of placental mammals. This fact, together with our observation of preferential expression of Tnfrh1 mRNA in cells at the fetal-maternal boundary, highlights Tnfrh1 as a potential example of the rapid evolution of genes with functions specific to the placenta, a process postulated to be driven by conflict between fetal and maternal alleles. Anecdotal examples, including pregnancy-associated glycoproteins, trophoblast interferons, the Pem and Psx homeobox genes, and the placental lactogen genes support the notion that 'placental genes' evolve rapidly 121314151617, but counter-examples can also be adduced. Two imprinted genes with placenta-specific expression and function, namely Ipl and Mash2 671819 are highly conserved. This contrasts with the lack of conservation of Tnfrh1, and the species-specific imprinting of at least one placental lactogen gene 20. Of more immediate interest is the biological function of Tnfrh1, which will need to be determined by knockout experiments. Since the sequence of this gene predicts that it encodes a TNF decoy receptor, a likely possibility is that it acts to dampen immune responses to the fetal semi-allograft.

Imprinting of genes on mouse distal Chr7 is controlled by two DMRs, which act as 'imprinting centers'. Of these two control elements, the closest to Tnfrh1 is KvDMR1, located in an intron of the Kcnq1 gene and giving rise to the Kcnq1ot1 non-translated RNA. This element, which is ~350 kb distant from Tnfrh1, is essential in cis for the imprinting of at least 4 genes, Kcnq1, Cdkn1c, Slc22a1l and Ipl/Tssc3 (212223 and M. Higgins, personal communication). All of these genes are relatively repressed on the paternal allele and active on the maternal allele. The simplest explanation for our finding of weak but consistent functional imprinting of Tnfrh1, with relative hyper-expression from the maternal allele, is that the KvDMR1 control element exerts distance-dependent effects. However, studies using KvDMR1-mutant mice will be necessary to confirm this.