The PEG10 gene is comprised of two exons separated by a 6753 bp long intron for transcript variant 1 (PEG10-A) or by a 6742 bp long intron for transcript variant 2 (PEG10-B). Transcript variant 2 is the result of alternative splicing where splicing occurs 11 nucleotides after the major splice site of exon 1 (Lux et al., 2010) (see also figure 1). PEG10-A, has a length of 6573 bp and PEG10-B of 6584 bp. The analysis of four different cell lines (HepG2, HEK293, HL60 and SH-SY5Y) suggest that variant 1 is the major transcript. How alternative splicing for PEG10 is regulated and why two alternative splice products exist in parallel is not known.

The major transcription start site (mTSS) was determined to be at position 19.519.958 of reference sequence NT_007933|Hs7_8090 (Lux et al., 2010). This TSS is preceded by a typical TATA-box element in the ideal distance of 24-30 nucleotides. It appears as if there is at least one additional may be cell type dependent but less frequently used TSS further upstream. Several studies attempted to analyse the PEG10 promoter and how PEG10 expression is regulated but the results of these studies do not provide a coherent picture. For example, it was reported that c-MYC upregulates PEG10 expression in pancreatic and hepatic carcinoma cells as well as in a B-lymphocyte cell line (Li et al., 2006). This effect appears to be mediated by c-MYC binding to an E-box sequence in the proximal region of the PEG10 intron, thereby influencing PEG10 promoter activity. In reporter assays, analysing just the promoter sequence upstream of the mTSS, overexpression of c-MYC showed an inhibitory effect (Lux et al., 2010). By bioinformatic analysis of the PEG10 promoter region +1 to -220, binding sites for transcription factors like TBP, Sp1 or E2F were identified (Lux et al., 2010). Binding of E2F members E2F-1 and E2F-4 to this region was experimentally proven and it was demonstrated that both factors positively regulate PEG10 expression (Wang C et al., 2008). A previous report showed that PEG10 expression is also positively regulated by E2F-2 and E2F-3 in the U2OS osteosarcoma cell line (Müller et al., 2001). These data suggest that PEG10 expression can be controled by the E2F/Rb pathway that involves the cyclin D/CDK4 complex, which phosphorylates pocket proteins like the retinoblastoma protein Rb and releases E2Fs. Thus, it can be expected that overexpression of cyclin D and CDK4 does also increase PEG10 expression, which indeed is the case (Wang C et al., 2008). In contrast, the presence of TGF-beta leads to a dephosphorylation of Rb, therefore repressing the expression of E2F target genes like PEG10 (Wang C et al., 2008). The repressive activity of TGF-beta on PEG10 expression is in agreement with our own unpublished results with PEG10 promoter-reporter constructs. Previous results showed that the PEG10-RF1 protein inhibits TGF-beta3 signalling in a TGF-beta-specific luciferase reporter assay (Lux et al., 2005), may be representing a self-protecting mechanism from down-regulation. Furthermore, it was reported that increased PEG10 expression is subjected to hormonal regulation by the male hormon androgen (Jie et al., 2007). In this study, three androgen receptor binding sites (ARE) were identified for PEG10. Two sites were reported for exon 2 and one for the promoter region. Unfortunately, no exact specifications were given about these sites. However, using the in the publication given primer sets for a Blast search against the PEG10 sequence reveals that ARE-1 is not located in the promoter but in exon 1. Therefore, all three PEG10-specific ARE sites are distal to the promoter.

Transcript processing. Northern blot analyses have shown that for humans depending on the tissue PEG10 transcripts of different size exist. One between 6 and 7 kb, corresponding to the major 6.6 kb PEG10 transcript, as well as minor sized transcripts (Ono et al., 2001; Smallwood et al., 2003; Lux et al., 2005). The major 6.6 kb PEG10 transcript is polyadenylated and at the distal end of exon 2 there are two canonical polyadenylation sequences, AATAAA. In a recent study, minor sized alternatively polyadenylated transcripts were isolated but none of these transcripts contained the typical polyadenylation signal motif at their 3-end nor any known alternative polyadenylation signal sequence motifs (Lux et al., 2010). If PEG10 transcripts are truly subjected to alternative polyadenylation then future studies have to address the question whether this influences PEG10 expression, mRNA stability, mRNA localisation or translation and if it might be related to pathological processes. Because PEG10 is most likely derived from a retrotransposon it is interesting to note that non-conserved poly(A) sites are associated with transposable elements to a much greater extent than conserved ones (Lee et al., 2008).

Transcription of PEG10 results in several protein isoforms due to alternative splicing, alternative translation initiation sites, posttranslational proteolytic cleavage and -1 ribosomal frameshift translation. The most prominent feature of PEG10 is its -1 ribosomal frameshift translation mechanism that hints at his retroviral/retrotransposon origin. The existence of PEG10 was reported by three different groups in 2001 (Ono et al., 2001; Shigemoto et al., 2001; Volff et al., 2001). In their search for novelle patternally expressed imprinted genes for an imprinted region on mouse chromosome 6, syntenic to a human imprinting cluster on chromosome 7q21 containing the imprinted SGCE gene, Ono and colleagues performed a database search for EST sequences mapping to this region. Three entries were identified, HB-1 (GenBank Accession No. AF216076), KIAA1051 (GenBank Accession No. AB028974), and 23915 mRNA (GenBank Accession No. AF038197) that were identical and mapped near SGCE. Sequence analysis of these clones predicted two open reading frames with homology to Gag and Pol proteins of some vertebrate retrotransposons, respectively. The deduced gene was named Paternally Expressed Gene 10 (PEG10). Similar results were obtained by Volff and colleagues, when they analysed public sequence databases for long-terminal-repeat (LTR) retrotransposon-like sequences of the Ty3/Gypsy retrotransposon family in mammals. They identified KIAA1051, which showed significant similarities to the Gag structural core protein of some Ty3/Gypsy retrotransposons from the Ty3 family, including Sushi from the pufferfish Fugu rubripes (42.5% similarities), Skippy, Maggy, and Cft1 from different fungi. No significant similarity to other families of Ty3/Gypsy retrotransposons and retroviruses was found and no LTR-like sequences flanking KIAA1051 were identified. It was further reported that the KIAA1051 cDNA contains a partial pol-like sequence (1.5 kb in length), which overlaps the gag-like sequence (1 kb in length) over approximately 250 bp. Its conceptual translation product displayed protease and truncated reverse transcriptase (RT) regions, including well-conserved first and second of the seven RT domains and showed the highest similarity to the Pol protein of again the retrotransposon Sushi (43.7% similarity). Further evidence for the existence of PEG10 came from the work by Shigemoto and colleagues. They analysed the mouse gene Edr, which was identified initially by differential screening of an embryonal carcinoma cDNA library for genes expressed at a reduced level following retinoic acid induced differentiation (Gorman et al., 1985). Their study identified in the Edr gene two open reading frames that overlapped and were set appart by a frameshift of one nucleotide. Subsequent analysis demonstrated that both reading frames are translated by -1 frameshifting (Shigemoto et al., 2001; Lux et al., 2005; Manktelow et al., 2005; Clark et al., 2007).

Translation. In order to perform the -1 frameshift, the reading frame 1 (RF1) - reading frame 2 (RF2) overlap sequence contains a seven nucleotide "slippery" sequence with typical consecutive homopolymeric triplets. The underlined PEG10 "slippery" heptanucleotide sequence G GGA AAC TC follows the general pattern of X XXY YYZ where the A- and P-site tRNAs detach from the zero frame codons XXY YYZ and re-pair after shifting back one nucleotide to XXX YYY and restart translation with the codon after the YYY triplet. Thus, the deduced amino acid sequence of the frameshift site after frameshift translation is GNL. The heptanucleotide "slippery" sequence is completely conserved in all species and the sequence of the downstream pseudoknot is completely conserved in the mammalian species. except for one nucleotide change in the rodent sequence. A detailed analysis of the PEG10 frameshift sequence was done by Manktelow and colleagues (2005).
Due to alternative splicing two transcript variants exist, PEG10-A and PEG10-B, leading to several protein isoforms. These isoforms are first, the result of different translation initiation sites in reading frame 1 (RF1) (Lux et al., 2010). Second, due to the fact, whether the reading frames 1 and 2 (RF2) are translated into an RF1 protein or into an RF1/2 protein by succesfull -1 frameshift translation, and third, in the RF1/2 translation products, shortly after the frameshift site, there is a retroviral typical functional aspartic protease motif usually for Gag-Pol protein processing leading to proteolytic cleavage products (Clark et al., 2007). It was demonstrated that upstream of the originaly predicted ATG translation initiation site (TIS) a second in frame non-ATG exists. For clarification, the non-ATG translation site will be named TIS-1a and the previous ATG start site TIS-2. This non-ATG start codon, a CTG, is 102 nucleotides upstream of the ATG start codon. The alternative splice event that leads to transcript PEG10-B introduces an additional in frame ATG start codon even further upstream of the previous two, which will be named TIS-1b. Their might be an additional TIS further downstream of TIS-2 (Lux et al., 2010). Transcript PEG10-A codes for PEG10-RF1 (using TIS-2), PEG10-RF1a (using TIS-1a), PEG10-RF1/2 and PEG10-RF1a/2. While in theory, transcript PEG10-B can lead to all six isoforms, PEG10-RF1, PEG10-RF1a, PEG10-RF1b (using TIS-1b), PEG10-RF1/2, PEG10-RF1a/2 and PEG10-RF1b/2. Figure 1 shows in a more schematic way the different TIS for transcripts PEG10-A and PEG10-B. The deduced amino acid sequences of the different isoforms are listed in figure 2.
Investigation of PEG10 translation in mouse placenta during gestation and human placenta showed that in vivo both reading frames are translated as an RF1 protein and an RF1/2 fusion protein (Clark et al., 2007). The mouse RF1/2 protein is about 40 kDa larger than the corresponding human protein due to an in frame insertion of approximately 600 nucleotides into the RF2 sequence. Interestingly, the size of RF1 and RF1/2 proteins and the translational frame shift efficiency varies during gestation. From 9.5 dpc when PEG10 expression in mice is first detectable, the 150 kDa frameshift protein is dominant. By using an RF1-specific antibody the frameshift efficiency was estimated and showed an apparent decrease from 68% at 9.5 dpc to 43% by 21.5 dpc. At 15.5 dpc an additional protein of 105 kDa was detected at about equal amounts as the 150 kDa protein. At late gestation, 21.5 dpc, it was present in greater amounts than the 150 kDa PEG10-RF1/2 protein. Mass spectrometry analysis identified the 105 kDa protein as a PEG10 product consisting primarily of PEG10 RF2 but containing peptides from both reading frames. PEG10 protein analysis for amniotic membrane showed a similar profile to that of placenta. Starting with a low expression at 9.5 dpc and then an increased and continued expression throughout gestation. The RF1/2 fusion protein again was the dominant band. Surprisingly, at 10.5 dpc a transient RF1 protein of increased mass of about 50 kDa was detected that disappeared at later time points and only the slightly smaller 47-kDa band identical in size to that in placenta was present again. Furthermore, in this report three RF1 protein populations were detected for HepG2 cells ranging from 47 to 55 kDa. Whether these different PEG10 protein masses are the result of post-translational modifications or due to the use of different TIS or a mixture of both is not clear and awaits further investigations.
In addition, western blot analysis with an RF1-specific antibody of adult mouse heart, spleen and brain tissue extracts showed a weak, single protein band but of different mass, around 50 kDa, for each tissue (Clark et al., 2007). No RF1/2 proteins were detected. The authors concluded based on their further analysis that these proteins do not represent Peg10 proteins.

Protein domains/motifs. By bioinformatic analyses using different programmes like the Simple Modular Architecture Research Tool (SMART), the SUPERFAMILY Sequence Search (SCOP domains) and the Eukaryotic Linear Motif (ELM) resource for functional site prediction, several domains and motifs were predicted. Some are exemplarily shown schematicaly in figure 3 for the 784 amino acid long PEG10-RF1b/2 protein. The Zink-finger domain was consistently identified although the size of the domain varies from amino acid 357-389 or a core region from 370-386 for a ZNF-C2HC (CX2CX4HX4C) consensus sequence, which is highly conserved in Gag proteins in most retroviruses and some retrotransposons. There are two proline rich regions, one at the N-terminus and one at the C-terminus. Proline-rich regions are recognized as presenting binding motifs to, for example, Src homology 2 (SH2) and SH3 domains. The ELM programme predicted the C-terminal proline stretch to be a possible binding site for SH3 domain containing proteins. As already reported, PEG10 contains a retroviral typical aspartyl protease consensus sequence, AMIDSGA. In order to test whether this motif is catalytic active the aspartate was mutated to an alanine (Clark et al., 2007). This change disrupted the protease activity and proved that the aspartyl protease is responsible for the cleavage of the full length PEG10 frameshift protein in to the RF1 and RF2 parts. Taken the protease activity into account the previously estimated PEG10 frameshift efficiency of 15-30% (Shigemoto et al., 2001; Lux et al., 2005) was reestimated to be 60% (Clark et al., 2007).

Interacting proteins. Aside from the protease motif, for none of the other domains it is known whether they are functional nor if they bind to other proteins. The only known binding partners for PEG10 are currently the SIAH1 and SIAH2 proteins (Okabe et al., 2003) and the TGF-beta type I receptor ALK1 (Lux et al., 2005). All three proteins were identified by a yeast two-hybrid screen with the PEG10-RF1 protein and the interactions were confirmed by co-immunoprecipitation experiments. The exact SIAH1/SIAH2 binding region was not determined, but the ELM programme predicted a potential SIAH1 binding site (figure 3, PEG10-RF1b amino acids 329-337). Co-immunoprecipitation experiments by overexpressing PEG10-RF1 and several other type I and II receptors of the TGF-beta superfamily in COS-1 cells showed that PEG10 does also interact with other members of this receptor group (Lux et al., 2005). Nevertheless, when specifically investigated in the two-hybrid assay under stringent conditions none of these receptors reacted with PEG10-RF1 to activate the reporter system. Thus, the most specific interaction appears to be with ALK1.

Based on data obtained from mice, Peg10 is predominantly expressed during embryonic development, whereas later on expression in most tissues ceases or is low except for testis and brain of adult animals (Shigemoto et al., 2001). However, significant induction of Peg10 expression was detected in hepatocellular carcinomas (HCC) and for the regenerating livers of mice after partial hepatectomy (Tsou et al., 2003). Studies with mice by RNA in situ hybridisation showed a high expression during embryonic development especially from day 9.5 to 16.5, specifically in bone and cartilage forming tissues as well as in extra embryonic tissues at all stages between E7.5 and E17.5. For a detailed description see Shigemoto et al. (2001). In humans, expression of PEG10 in adult tissues was seen in brain, kidney, lung, testis and only weak to very weak expression in spleen, liver, colon, small intestine and muscle, but no expression for heart and stomach (Ono et al., 2001). Furthermore, strong expression was also reported for mouse and human placenta. In humans PEG10 expression is low at the early hypoxic phase of placental growth and increases at 11-12 weeks of gestation. This high level of expression is maintained and is significantly increased in term placenta compared with that in early pregnancy (Smallwood et al., 2003). The authors hypothesize that the gene product might be essential for trophoblast differentiation and uterine implantation.
Peg10 expression was observed in relation to adipocyte differentiation in mice. Peg10 was identified as one of the genes expressed early in adipogenesis (Hishida et al., 2007). Expression of Peg10 was elevated after the addition of differentiation inducers in adipocyte differentiable 3T3-L1 cells, but not in the non-adipogenic cell line NIH-3T3. The knockdown of Peg10 by RNA interference inhibited the differentiation of 3T3-L1 cells. Moreover, Peg10 siRNA treatment impaired mitotic clonal expansion (MCE), necessary for adipocyte differentiation, and the crucial expression of C/EBPbeta and C/EBPdelta at the immediate early stage of the differentiation process was inhibited by the knock-down. These results indicate that Peg10 plays an important role at the immediate early stage of adipocyte differentiation.
Aside from bone and cartilage tissue differentiation and adipocyte cell differentiation PEG10 might also be involved in neuronal cell differentiation. In a preliminary experiment with the neuroblastoma cell line SH-SY5Y increasing PEG10 expression was reported over a period of 14 days after treatment with all-trans retinoic acid for differentiation (Lux et al., 2010).

Data regarding PEG10s cellular localisation do only exist for PEG10-RF1. Okabe and colleagues (2003) report for the hepatoma cell lines HepG2, Huh7 and Alexander as well as for hepatocellular carcinoma (HCC) tissues nuclear and cytoplasmic staining of PEG10 with a PEG10-RF1-specific antibody. In experiments overexpressing PEG10-RF1 in HEK293T cells, only cytoplasmic localisation was reported (Tsou et al., 2003), which was also shown by immunofluorescence analysis with a PEG10-RF1-specific antibody for endogenous PEG10 in HepG2 and B-CLL cells (Lux et al., 2005; Kainz et al., 2007).

The knowledge regarding PEG10s protein function is sparse and all direct evidence that exists so far was gained by experiments with the PEG10-RF1 isoform only. No data exists for the recently identified isoform PEG10-RF1a, most likely the major RF1 isoform, or the PEG10-RF1b protein. Functional data regarding the different RF1/2 isoforms are lacking too.
Nevertheless, PEG10-RF1 appears to enhance cell proliferation and blocks apoptosis. Evidence for PEG10s role in cell proliferation were reported by Tsou et al. (2003). Induced PEG10 expression was found during G2/M phase of regenerating mouse liver and elevated expression of PEG10 was found in HCC. The authors state that both, HCC and regenerating mouse livers, represent the two proliferative states of the otherwise quiescent liver tissue. In addition, ectopic expression of PEG10 in 293T cells enhanced cell cycle progression. Complementary data were presented by Okabe and colleagues (2003). The hepatoma cell line, SNU423 that has no endogenous PEG10, was stable transfected with an expression construct for PEG10-RF1 to test the effect of PEG10 on cell growth. The PEG10 stable transfectant cells revealed significant growth promotion compared with the parental or mock cells. Under conditions of serum starvation (0.1% FBS), the mock cells rapidly underwent growth arrest, but stable PEG10-expressing cells continued to proliferate. In the same report, in a yeast two-hybrid screen, the apoptosis inducing protein SIAH1 was identified as a PEG10-RF1 interactor. Overexpression of SIAH1 increased cell death in different hepatoma cell lines, whereas co-expression of PEG10-RF1 in tested SNU423 revealed a partial but significant protection from apoptosis. Anti-apoptotic activity of PEG10 was also reported by Yoshibayashi et al. (2007). The finding that PEG10 enhances cell proliferation was further confirmed by PEG10-specific siRNA knock-down experiments in a series of carcinoma cell lines, i.e. Panc1, HepG2, and Hep3B, which led to a significantly reduced cell proliferation (Li et al., 2006; Yoshibayashi et al., 2007).
Anti-apoptotic activity by PEG10 were not only seen for HCC/hepatoma but also reported for B-cell acute and chronic lymphocytic leukemia, B-ALL and B-CLL. It was observed that B-ALL and B-CLL CD19+CD34+ B cells expressed elevated levels of PEG10, regulated by the chemokines CXCL13 and CCL19 and that these cells were resistant to TNF-alpha induced apoptosis. Treatment of the cells with PEG10 antisense constructs reversed this effect (Hu et al., 2004; Chunsong et al., 2006; Wang et al., 2007). Kainz and colleagues (2007) reported that PEG10 overexpression is associated with high-risk B-CLL. Expression levels in CD19+ B-CLL cells were up to 100-fold higher than in B-cells from healthy donors and expression levels in B-CLL patient samples remained stable over time even after chemotherapy. The intensity of intracellular staining of PEG10 protein corresponded to mRNA levels. Further analysis of PEG10s anti-apoptotic potential showed that short term knock-down (2 days after transfection) of PEG10 in B-CLL cells was not associated with changes in cell survival but long term inhibition (4 days after transfection of PEG10) led to a significant effect on the induction of apoptosis and late apoptosis/necrosis in B-CLL cells.
As described above, during gestation PEG10 is highly expressed in the placenta. Mouse placenta is positive for Peg10 transcripts as well as for proteins from reading frame 1 and reading frame 1/2 (Clark et al., 2007). Peg10 knock-out experiments in mice demonstrated that Peg10 and therefore the Peg10 proteins have an important role during placenta development. Heterozygous knock-outs for the patternal allele died at 10.5 dpc. Their placentas were severely depleted and the labyrinth layer was not developed and the spongiotrophoblast cells were missing (Ono et al., 2006). As discussed by the authors, parthenogenetic embryos die before 9.5 d.p.c. and show early embryonic lethality with poorly developed extraembryonic tissues. Morphological defects of the most developed parthenotes are very similar to those of Peg10-Pat KO embryos; they lack the diploid trophoblast cells of the labyrinth layer and the spongiotrophoblast. However, the majority of parthenotes show more severe phenotypes; suggesting that other genes could also contribute to the parthenogenetic phenotypes. Nevertheless, the result for mouse Peg10 suggests that one of the Peg10 isoforms could be critical for parthenogenetic development in mice and therefore also in man.

PEG10 orthologous sequences can be found in several other eutherian species: Pan troglodytes (chimpanzee), Papio anubis (baboon), Macaca mulatta (Rhesus monkey), Callithrix jacchus (marmoset), Sus scrofa (pig), Canis familiaris (dog), Felis catus (cat), Bos taurus (bovine), Ovis aries (sheep), Mus musculus (mouse), Rattus norvegicus (rat), Rhinolophus ferrumequinum (bat), Sorex araneus (shrew), Monodelphis domestica (opossum) as well as in the metatherian tammar wallaby (Macropus eugenii) (Ono et al., 2001; Brandt et al., 2005a; Brandt et al., 2005b; Suzuki et al., 2007; Clark et al., 2007) and there is a high degree of amino acid conservation. The PEG10 sequence is not conserved in reptiles or birds. Hence, during evolution the PEG10 gene was introduced into the therian mammal genome after the split of birds about 300 million years (myr) ago and after the split of prototherian mammals (monotremes) 166 myr ago, but before the divergence between placental mammals and marsupials about 148 myr ago.