Expression of type C-related endogenous retroviral sequences in human colon tumors and colon cancer cell lines

HERV-E-mediated modulation of PLA2G4A transcription in urothelial carcinoma

Human endogenous retroviruses (HERV) and related elements account for more than 8% of the human genome and significantly contribute to the human transcriptome by long terminal repeat (LTR) promoter activity. In this context, HERVs are thought to intervene in the expression of adjacent genes by providing regulatory sequences (cis-effect) or via noncoding RNA including natural antisense transcripts. To address the potential impact of HERV activity in urothelial carcinoma, we comparatively analyzed the HERV transcription profiles in paired samples of non-malignant urothelium and urothelial carcinoma derived from 13 patients with bladder cancer by means of a retrovirus-specific microarray (RetroArray). We established a characteristic HERV signature consisting of six ubiquitously active HERV subgroups (E4-1, HERV-Rb, ERV9, HERV-K-T47D, NMWV3, HERV-KC4). The transcription pattern is largely identical in human urothelial carcinoma, non-malignant urothelial tissue, four tumor-derived cell lines and in a non-malignant urothelial cell line (UROtsa). Quantitative reverse transcriptase PCR (qRT-PCR) of HERV-E4-1, HERV-K(HML-6) and HERV-T(S71-TK1) revealed a bias to lower HERV activity in carcinoma samples compared to non-malignant tissue. Determination of active HERV-E4-1 loci by cloning and sequencing revealed six HERV-E4-1 proviral loci that are differentially regulated in urothelial carcinoma cells and normal tissue. Two full-length HERV-E4-1 proviruses, HERV-Ec1 and HERV-Ec6, are located in antisense orientation in introns of the genes PLA2G4A and RNGTT, respectively. PLA2G4A encodes a cytosolic phospholipase A2 (cPLA2) that is dysregulated in many human tumors. PLA2G4A and HERV-Ec1 displayed reciprocal transcript levels in 7 of 11 urothelial carcinoma patients. Moreover, reciprocal shifts were observed after treatment of UROtsa cells with HERV-Ec1 and PLA2G4A-directed siRNAs or 5-aza-2′-deoxycytidine (aza-dC) pointing to an antagonistic regulation of PLA2G4A and HERV-Ec1 transcription in human urothelial cells. We suggest that transcription of HERV-Ec1 contributes to fine tuning of cPLA2 expression, thereby facilitating tumorigenesis.

Expression of HNF4alpha variants in pancreatic islets and Ins-1 beta cells

BACKGROUND

Hepatocyte nuclear factor (HNF4alpha) is a nuclear receptor essential for endodermal differentiation and cell functions in the adult pancreas, liver, and other tissues. Mutations in the HNF4A gene cause MODY1. Up to nine protein variants arise from two developmentally regulated promoters. Because some variants lack the N-terminal activation function 1 (AF-1) and/or C-terminal inhibitory F domain, defining their tissue-specific regulation and function is important for understanding pancreatic beta cell behaviour.

METHODS

Expression of HNF4alpha variants in islets, rat Ins-1 insulinoma cells, and human Hep3B hepatocellular carcinoma cells was assessed using a long-range reverse transcription-polymerase chain reaction (RT-PCR) strategy capable of recognizing each combination of mRNA termini. Protein expression was verified by immuno-blotting with terminus-specific antibodies and DNA-binding assays.

RESULTS

Mouse islets and both cell lines express HNF4alpha9, which lacks both AF-1 and the F domain. Islets also expressed the HNF4alpha P1 promoter variants HNF4alpha1/alpha2, and Hep3B cells expressed HNF4alpha3. When ectopically expressed in COS-7 cells, HNF4alpha1, alpha3, alpha7, and alpha9 each stimulated an HNF4alpha-dependent promoter. Variants containing exon 1B (HNF4alpha4 - alpha6) were not detected. Lack of canonical splicing signals and species conservation argues against exon 1B usage.

CONCLUSIONS

This is the first report of HNF4alpha9 expression in any tissue. Our findings extend our understanding of HNF4alpha gene transcription and function. This knowledge may be useful in efforts to recover or establish regulated insulin secretion.

Expression and functions of human endogenous retroviruses in the placenta: an update

The placenta is unique amongst normal tissues in transcribing many different human endogenous retrovirus (HERV) families at high levels and this has led to the suggestion that HERVs may fulfil important functions in reproduction. This review discusses our current knowledge of the placental expression of HERVs, in particular the envelope proteins of ERV3 and HERV-W which may have critical roles in placental function.

Selective deletion of the Hnf1beta (MODY5) gene in beta-cells leads to altered gene expression and defective insulin release

Hepatocyte nuclear factor 1alpha (HNF1alpha) and HNF1beta (or vHNF1) are closely related transcription factors expressed in liver, kidney, gut, and pancreatic beta-cells. Many HNF1 target genes are involved in carbohydrate metabolism. Human mutations in HNF1alpha or HNF1beta lead to maturity-onset diabetes of the young (MODY3 and MODY5, respectively), and patients present with impaired glucose-stimulated insulin secretion. The underlying defect in MODY5 is not known. Analysis of HNF1beta deficiency in mice has not been possible because HNF1beta null mice die in utero. To examine the role of HNF1beta in glucose homeostasis, viable mice deleted for HNF1beta selectively in beta-cells (beta/H1beta-KO mice) were generated using a Cre-LoxP strategy. beta/H1beta-KO mice had normal growth, fertility, fed or fasted plasma glucose and insulin levels, pancreatic insulin content, and insulin sensitivity. However, beta/H1beta-KO mice exhibited impaired glucose tolerance with reduced insulin secretion compared with wild-type mice but preserved a normal insulin secretory response to arginine. Moreover, beta/H1beta-KO islets had increased HNF1alpha and Pdx-1, decreased HNF4 mRNA levels, and reduced glucose-stimulated insulin release. These results indicate that HNF1beta is involved in regulating the beta-cell transcription factor network and is necessary for glucose sensing or glycolytic signaling.

Differential gene expression in a murine model of cancer cachexia

Murine adenocarcinoma 16 (MAC16) tumors and cell lines induce cachexia in NMRI nude mice, whereas histologically similar MAC13 tumors do not. After confirming these findings in BALB/c nude mice, we demonstrated that this tissue wasting was not related to decreased food intake or increased total body oxidative metabolism. Previous studies have suggested that MAC16's cachexigenic properties may involve the production of tumor-specific factors. We therefore screened for genes having increased expression in the MAC16 compared with the MAC13 cell line by performing hybridization to a murine cDNA expression array, by generation and comparison of cDNA libraries from each cell line, and by PCR-based subtractive hybridization. Northern blot hybridization was performed to confirm differences in transcript expression. Transcripts encoding insulin-like growth factor binding protein-4, cathepsin B, ferritin light and heavy chain, endogenous long-terminal repeat sequences, and a viral envelope glycoprotein demonstrated increased expression in the MAC16 cell line. The roles of a number of these genes in known metabolic pathways identify them as potential participants in the induction of cachexia.

Influence of the human endogenous retrovirus-like element HERV-E.PTN on the expression of growth factor pleiotrophin: a critical role of a retroviral Sp1-binding site

Germ line insertion of a human endogenous retrovirus-like element (HERV-E.PTN) into the growth factor pleiotrophin (PTN) gene generated a phylogenetically new promoter driving the expression of functional HERV-PTN fusion transcripts. Here we show by in situ hybridization, that HERV-PTN fusion transcripts are expressed in malignant trophoblasts (i.e. choriocarcinoma) and in the proliferative and in the invasive trophoblasts of gestational trophoblastic tissue. Additionally, a 1.9 kb fragment of the HERV-derived PTN promoter was analysed which has strong activity when transiently transfected into choriocarcinoma JEG-3 cells in contrast to HeLa cells. Deletion of the retrovirally-derived promoter portion abolished its activity and an enhancer (+443 to +486) was identified in this region. Electrophoretic mobility shift and supershift experiments identified a Sp1 binding site in this enhancer and site specific mutation of this site abolished its activity in choriocarcinoma cells. Sp1 overexpression in Drosophila SL2 cells showed that the enhancer activity is mediated via Sp1 binding in vivo. Furthermore, mutation of the Sp1 binding site reduced the activity of a promoter test fragment in choriocarcinoma cells by 80%. Our result shows that a retroviral Sp1 binding site in the PTN promoter is important for the expression of growth factor pleiotrophin in human choriocarcinoma cells. Oncogene (2000) 19, 3988 - 3998.

Structure and phylogenetic analysis of an endogenous retrovirus inserted into the human growth factor gene pleiotrophin

A human endogenous retrovirus-like element (HERV), flanked by long terminal repeats of 502 and 495 nucleotides is inserted into the human pleiotrophin (PTN) gene upstream of the open reading frame. Based on its Glu-tRNA primer binding site specificity and the location within the PTN gene, we named this element HERV-E.PTN. HERV-E.PTN appears to be a recombined viral element based on its high homology (70 to 86%) in distinct areas to members of two distantly related HERV type C families, HERV-E and retrovirus-like element I (RTVL-I). Furthermore, its pseudogene region is organized from 5' to 3' into gag-, pol-, env-, pol-, env-similar sequences. Interestingly, full-length and partial HERV-E.PTN-homologous sequences were found in the human X chromosome, the human hereditary haemochromatosis region, and the BRCA1 pseudogene. Finally, Southern analyses indicate that the HERV-E.PTN element is present in the PTN gene of humans, chimpanzees, and gorillas but not of rhesus monkeys, suggesting that genomic insertion occurred after the separation of monkeys and apes about 25 million years ago.

Retroviruses and systemic lupus erythematosus

In some animal models of autoimmune diseases the roles of exogenous and endogenous retroviruses are clearly defined. In ungulates caprine arthritis encephalitis virus, equine infectious anemia virus or Maedi-Visna virus infections cause a well-defined autoimmune disease and the appearance of seropositivity of the animals is of diagnostic value. Likewise, in MRL lpr/lpr mice insertion of a retrotransposon into the fas gene could clearly be shown to cause survival of autoreactive lymphocytes. Despite intensive research in this field over a long period of time, molecular data on retroviral involvement in either etiology or pathology of human SLE and other autoimmune rheumatic diseases remain rather scarce. However, the analysis of retroviral antibodies and antigens in human autoimmune disease is undoubtedly important with regard to the search for retroviruses as disease-causing agents.

The potential roles of endogenous retroviruses in autoimmunity

Endogenous retroviruses (ERVs) are estimated to comprise up to 1% of human DNA. While the genome of many ERVs is interrupted by termination codons, deletions or frame shift mutations, some ERVs are transcriptionally active and recent studies reveal protein expression or particle formation by human ERVs. ERVs have been implicated as aetiological agents of autoimmune disease, because of their structural and sequence similarities to exogenous retroviruses associated with immune dysregulation and their tissue-specific or differentiation-dependent expression. In fact, retrovirus-like particles distinct from those of known exogenous retroviruses and immune responses to ERV proteins have been observed in autoimmune disease. Quantitatively or structurally aberrant expression of normally cryptic ERVs, induced by environmental or endogenous factors, could initiate autoimmunity through direct or indirect mechanisms. ERVs may lead to immune dysregulation as insertional mutagens or cis-regulatory elements of cellular genes involved in immune function. ERVs may also encode elements like tax in human T-lymphotrophic virus type I (HTLV-I) or tat in human immunodeficiency virus-I (HIV-I) that are capable of transactivating cellular genes. More directly, human ERV gene products themselves may be immunologically active, by analogy with the superantigen activity in the long terminal repeat (LTR) of mouse mammary tumour viruses (MMTV) and the non-specific immunosuppressive activity in mammalian type C retrovirus env protein. Alternatively, increased expression of an ERV protein, or expression of a novel ERV protein not expressed in the thymus during acquisition of immune tolerance, may lead to its perception as a neoantigen. Paraneoplastic syndromes raise the possibility that novel ERV-encoded epitopes expressed by a tumour elicit immunity to cross-reactive epitopes in normal tissues. Recombination events between different but related ERVs, to whose products the host is immunologically tolerant, may also generate new antigenic determinants. Frequently reported humoral immunity to exogenous retrovirus proteins in autoimmune disease could be elicited by cross-reactive ERV proteins. A review of the evidence implicating ERVs in immune dysfunction leads to the conclusion that direct molecular studies are likely to establish a pathogenic role for ERVs in autoimmune disease.

Identification of human endogenous retroviruses with complex mRNA expression and particle formation

Retroviruses comprise strains with considerable disease potential in animals and humans. In addition to exogenous strains transmitted horizontally, endogenous proviruses are transmitted through the germ line. Some of these endogenous retroviruses can be pathogenic in mice and possibly in other animal species. They may also be considered as mobile genetic elements with the potential to produce mutations. In humans, genomic DNA contains numerous endogenous retroviral sequences detected by their partial relatedness to animal retroviruses. However, all proviruses sequenced so far have been found to be defective. In this communication, we describe the expression of a family of human endogenous retrovirus sequences (HERV-K) in GH cells, a teratocarcinoma cell line producing the human teratocarcinoma-derived retrovirus (HTDV) particles previously described by us. Four viral mRNA species could be identified, including a full-length mRNA. The other three subgenomic mRNAs are generated by single or double splicing events. This expression pattern is reminiscent of the more complex control of virus gene regulation observed, for example, with lenti- or spumavirus strains, although HERV-K shows no sequence homology to human T-lymphotropic virus or human immunodeficiency virus. Sequence analysis of expressed HERV-K genomes revealed non-defective gag genes, a prerequisite for particle formation. Open reading frames were also observed in pol and env. Antisera raised against recombinant gag proteins of HERV-K stained HTDV particles in immunoelectron microscopy, linking them to the HERV-K family.

Reactivities of HIV-1 gag-derived peptides with antibodies of HIV-1-infected and uninfected humans

A group of 41 peptides, each 24 amino acids long and overlapping with each other by 12 residues spanning the total gag open reading frame (orf) of HIV-1 (HTLV-IIIBH 10 isolate) were synthesized using Fmoc chemistry. The purified compounds were used in ELISA assays and tested for antibody reactivities in sera of human HIV-1-infected and noninfected individuals. Sera of HIV- humans showed reactivity against four defined regions, two in p17, one in p24, and one in p15. The values of these reactivities were elevated especially in serum samples of HIV- individuals showing cross-reaction with gag proteins on Western blot. Amino acid sequence comparison of HIV-1 gag proteins with those of human endogenous retroviruses (ERV K10, ERV 3) revealed significant similarities predominantly in the domains showing elevated antibody cross-reactions. The majority of sera from HIV-1+ individuals showed strong reactivities to the cross-reactive regions and to various other peptide sequences, a sequential epitope recognized by all HIV-1+ sera could, however, not be identified. The results suggest that human individuals may have immune reactions to endogenous retroviral protein sequences, which are enhanced by infections with HIV-1. Specific antibodies to HIV-1 gag proteins are probably mainly directed to tertiary structure defined epitopes formed by particle formation of the p24 monomers to the nucleocapsid.

Safety issues related to retroviral-mediated gene transfer in humans

The first three approved human clinical trials utilizing retroviral-mediated gene transfer are now underway. While this technology holds great promise for the study and treatment of human disease, it also poses a number of safety concerns. In evaluating clinical protocols, potential complications and the likelihood of their occurrence are estimated by review committees so that a risk/benefit assessment can be made. Current knowledge, reviewed in this article, suggests that no acute complications secondary to retroviral-mediated gene transfer are likely, but the possibility of long-term or unforeseen sequelae in patients suggests the need for post-treatment monitoring.

Autonomous expression of RTVL-H endogenous retroviruslike elements in human cells

Northern (RNA) blot analysis of RNA from various human cell lines and tissues has demonstrated that elements belonging to the RTVL-H family of human endogenous retroviruslike sequences are expressed in several cell types. The highest levels of RTVL-H-related RNAs were observed in teratocarcinoma cell line NTera2D1, HeLa cells, two bladder carcinoma cell lines, and normal amniotic tissue. Expression was also observed in normal chorion and in some other cell lines. The RTVL-H transcription pattern varied among the different cell types, but several expressed a unit-length 5.6-kilobase transcript. Characterization of cDNA clones corresponding to transcripts present in NTera2D1 cells indicates that the complex transcription pattern observed in these cells is generated by the following: (i) transcription of both full-length and deleted genomic elements, which is initiated within the 5' long terminal repeat (LTR) and, in all but one case, polyadenylated in the 3' LTR; (ii) the splicing of both unit-length transcripts and transcripts from a deleted element; (iii) transcription involving solo LTR sequences; and (iv) transcription which, in one case, reads through the 3' LTR into flanking cellular sequences. Sequence data obtained from 25 cDNA clones revealed that at least 13 RTVL-H elements are expressed in NTera2D1 cells. The positions of several termination codons within the pol region are the same among nine different elements, indicating that an ancestral RTVL-H element bearing these mutations dispersed within the genome. We also found that RTVL-H expression varied among samples of amnion and chorion tissue isolated from different individuals. These findings demonstrate that regulated autonomous expression of RTVL-H sequences occurs in human cells.

Detection of multiple, novel reverse transcriptase coding sequences in human nucleic acids: relation to primate retroviruses

A variety of chemically synthesized oligonucleotides designed on the basis of amino acid and/or nucleotide sequence data were used to detect a large number of novel reverse transcriptase coding sequences in human and mouse DNAs. Procedures involving Southern blotting, library screening, and the polymerase chain reaction were all used to detect such sequences; the polymerase chain reaction was the most rapid and productive approach. In the polymerase chain reaction, oligonucleotide mixtures based on consensus sequence homologies to reverse transcriptase coding sequences and unique oligonucleotides containing perfect homology to the coding sequences of human T-cell leukemia virus types I and II were both effective in amplifying reverse transcriptase-related DNA. It is shown that human DNA contains a wide spectrum of retrovirus-related reverse transcriptase coding sequences, including some that are clearly related to human T-cell leukemia virus types I and II, some that are related to the L-1 family of long interspersed nucleotide sequences, and others that are related to previously described human endogenous proviral DNAs. In addition, human T-cell leukemia virus type I-related sequences appear to be transcribed in both normal human T cells and in a cell line derived from a human teratocarcinoma.

Evidence for expression of the facilitated glucose transporter in rat hepatocytes

The eukaryotic facilitated glucose transporter (GT) is expressed by many cell types, with the notable exception of hepatocytes; however, GT is expressed by several hepatoma cell lines, including the well-differentiated lines Fao, Hep3B, and HepG2. We report on studies carried out to determine the aspect(s) of the transformed phenotype that might be responsible for activating GT expression. Using RNA blot analysis with probes derived from rat GT cDNA, we found that GT was expressed by rat hepatocytes under two conditions (i) in vitro, when isolated hepatocytes were placed in cell culture, and (ii) in vivo, when rats were subjected to starvation for greater than or equal to 2 days. However, GT expression was not an obligatory feature of hepatomas, since two primary hepatocellular carcinomas did not express any GT mRNA. GT expression in hepatocytes was reduced by addition of dimethyl sulfoxide or sodium butyrate to the culture medium. Since these reagents are known to promote differentiation in some cell culture systems, their effect on hepatocytes may be to maintain the GT repression normally observed in vivo. Inclusion or exclusion in the culture medium of several other agents that enhance hepatocyte viability (serum, insulin, corticosteroids, epidermal growth factor, or triiodothyronine) did not affect GT expression. It is unclear whether the two conditions that led to GT expression in hepatocytes are related by a common signaling mechanism. Possibly, both cases involve a "stress" response: in vivo, a normal physiological response to starvation; in vitro, a response to a major alteration in the cellular environment.

Sodium butyrate increases glucose transporter expression in LLC-PK1 cells

The effect of sodium butyrate on the expression of the facilitated glucose transporter (GT) was investigated in the pig kidney cell line LLC-PK1. When cells were treated with butyrate, GT mRNA expression was remarkably enhanced with a maximal effect at 5 mM. Levels of GT mRNA were increased at 1 day after butyrate treatment and continued to increase for at least 4 days; however, acetate and propionate did not affect GT mRNA levels significantly. The induction of GT mRNA by butyrate was accompanied by an increase in GT function. The expression of GT mRNA decreased in HepG2, HT-29, and COS cells by treatment with butyrate for 1 day. Interestingly, glucose deprivation of LLC-PK1 cells reduced the induction of GT mRNA by butyrate, although starvation itself slightly enhanced steady-state GT mRNA levels. Therefore, expression of GT in LLC-PK1 cells is strongly induced by butyrate by a pathway that apparently depends on the presence of glucose in culture medium.

Partial suppression of anchorage-independent growth and tumorigenicity in immunodeficient mice by transfection of the H-2 class I gene H-2Ld into a human colon cancer cell line (HCT)

Many human tumors, particularly those of epithelial origin, appear to express greatly reduced levels of major histocompatibility complex class I antigens on their surface. It has been previously reported that the class I gene H-2Ld, introduced into adenovirus type 12-transformed mouse cells, induces reversal of oncogenesis in immunocompetent BALB/c mice. We have tested the hypothesis that the H-2Ld gene, when transfected into HCT colon cancer cells, may alter their transformed phenotype. Two H-2Ld transfectants, HCT-Ii and HCT-If, were found to exhibit a markedly reduced-to-virtually suppressed ability to form colonies in soft agar in comparison to a transfectant (HCTh) carrying only the neomycin-resistance gene. We also compared the tumorigenicity of HCTh vs. HCT-If cells in two different strains of immunodeficient mice: nude (T-) and triple-deficient mutants (T-, NK-, B-). At 28 days postinjection of 10(7) and 10(6) cells, the size and growth rate of HCT-If tumors were greatly reduced compared to HCTh cells. Therefore, as assayed in immunodeficient animals, expression of the class I H-2Ld gene in HCT cells appears to correlate with partial suppression of the tumorigenic phenotype, suggesting that the expression of a transfected class I gene may by itself alter the phenotype of the recipient cell and that such phenotypic changes may be independent of the immune system.

Expression of human sequences related to those of mouse mammary tumor virus

Sequences related to those of the mouse mammary tumor virus (MuMTV) genome have been cloned from human DNA by screening a library prepared from the DNA of a human breast cancer cell line with MuMTV gag-pol DNA. Nine distinct groups of (MuMTV-related) sequences were identified among 100 lambda recombinants by cross-hybridization experiments with subcloned fragments containing gag-pol-related DNA. The largest group, of 64 recombinants, contains the MuMTV-related sequences cloned by others. The other eight groups contain MuMTV-related sequences that have not been described previously. The gag-pol regions of one recombinant from each of the nine groups were hybridized to RNA prepared from five human breast cancer cell lines, from placenta, and from two cell lines derived from other malignancies. RNAs were detected by probes for seven of the groups. The RNAs ranged in size from 1.2 to 12 kilobases. Probes for six of the groups detected large RNAs that could represent transcripts of full-length proviral DNA. Two of the probes detected RNA in one breast cancer cell line only. Most of the RNAs were detected in more than one cell line.

Differentiation-associated decrease in the proportion of fucosylated polylactosaminoglycans of CaCo-2 human colonic adenocarcinoma cells

CaCo-2 cells are human colonic adenocarcinoma cells which can differentiate spontaneously into enterocytes when maintained confluent for extended periods of time. Cells kept in culture for 4 days (rapidly growing), 7-9 days (early confluence) and 19-22 days (late confluence) were incubated for 24 h with L-[5,6-3H]fucose or D-[6-3H]glucosamine in order to examine the changes in glycoprotein carbohydrate structure that occur during this differentiation. Labelled glycopeptides obtained by exhaustive Pronase digestion of the cell-surface and cell-pellet fractions were fractionated on Bio-Gel P-6. A high-Mr glycopeptide fraction which was excluded from Bio-Gel P-6 was present in all cases. These glycopeptides were then fractionated by affinity chromatography on Datura stramonium agglutinin-agarose. The glycopeptides which were specifically bound to the lectin column were largely degraded by endo-beta-galactosidase, thereby indicating that they consisted of fucosylated polylactosaminoglycans. The proportion of labelled polylactosaminoglycans decreased with increasing time in culture, whereas sucrase activity, which is characteristic of differentiated enterocytes, increased. These results demonstrate that a relatively large decrease in the proportion of fucosylated polylactosaminoglycans occurs with differentiation of CaCo-2 cells.

Tissue-specific expression of human provirus ERV3 mRNA in human placenta: two of the three ERV3 mRNAs contain human cellular sequences

Three polyadenylated RNAs, 9, 7.3, and 3.5 kilobases long, of a human endogenous retrovirus, ERV3, are abundant in human placental chorion, representing about 0.03 to 0.05% of the total mRNA. We characterized the structure of these mRNAs by Northern blot and S1 nuclease mapping analyses. We found that all three RNAs were spliced mRNAs that lacked 5.9 kilobases of proviral sequence, including the gag gene and most of the pol gene. In contrast to the transcription pattern usual for other retroviruses, the transcription pattern of the ERV3 provirus did not include a genome-length mRNA. All three of the ERV3 mRNAs initiated transcription at the same point in the 5' long terminal repeat (LTR) and contained identical splice junctions in the provirus. The 3.5-kilobase RNA was a typical subgenomic proviral mRNA, with its polyadenylation site in the 3' LTR. The two larger ERV3 mRNAs, however, extended through the polyadenylation site in the 3' LTR and were spliced at a second position approximately 370 nucleotides downstream from the 3' LTR. This finding suggests that when the ERV3 retrovirus integrated at this genomic locus in an ancestor of humans, it integrated within or adjacent to a cellular gene.