Cell type-specific expression of LINE-1 open reading frames 1 and 2 in fetal and adult human tissues.

The LINE-1 (L1) family of non-long terminal repeat retrotransposons is a major force shaping mammalian genomes, and its members can alter the genome in many ways. Mutational analyses have shown that coexpression of functional proteins encoded by the two L1-specific open reading frames, ORF1 and ORF2, is an essential prerequisite for the propagation of L1 elements in the genome. However, all efforts to identify ORF2-encoded proteins have failed so far. Here, applying a novel antibody we report the presence of proteins encoded by ORF2 in a subset of cellular components of human male gonads. Immunohistochemical analyses revealed coexpression of ORF1 and ORF2 in prespermatogonia of fetal testis, in germ cells of adult testis, and in distinct somatic cell types, such as Leydig, Sertoli, and vascular endothelial cells. Coexpression of both proteins in male germ cells is necessary for the observed genomic expansion of the number of L1 elements. Peptide mass fingerprinting analysis of a approximately 130-kDa polypeptide isolated from cultured human dermal microvascular endothelial cells led to the identification of ORF2-encoded peptides. An isolated approximately 45-kDa polypeptide was shown to derive from nonfunctional copies of ORF2 coding regions. The presence of both ORF1- and ORF2-encoded proteins in vascular endothelial cells and its apparent association with certain stages of differentiation and maturation of blood vessels may have functional relevance for vasculogenesis and/or angiogenesis.

A retrotransposition-competent, functional L1 element (RC-L1) covers ϳ6.1 kb and contains a 5Ј-untranslated region with an internal, CpG-rich promoter, a 1-kb ORF1 encoding a protein (p40) of approximately 40 kDa with RNA binding capability, followed by a 3.8-kb ORF2 coding for a protein (p150) with a predicted molecular mass of about 150 kDa with endonuclease (EN) and reverse transcription (RT) activities and a cysteine-histidine-rich domain. The 3Ј-end of L1 is terminated by a short 3Ј-untranslated region and a poly(A) tail (10) (Fig. 1A). L1 mRNAs are atypical of mammalian RNAs because they are bicistronic, and the mechanism of translation of L1 is not understood. The two ORFs are in-frame and separated by a 63-bp noncoding spacer region. Mutational analyses demonstrated that both ORF1-and ORF2-encoded functional proteins are required for retrotransposition (4,11). At least three functions of the ORF2-encoded protein were shown to be essential for retrotransposition, RT activity, EN activity, and a function associated with the cysteine-histidine-rich motif. Insertion of a new L1 copy into the loose genomic target sequence 5Ј-TTTT/ A-3Ј (8,(11)(12)(13) is initiated by a process termed target-primed reverse transcription (14,15). The structure of the target site duplications flanking de novo L1 integrants suggests a model for second strand synthesis of L1 termed "microhomologydriven single strand annealing" (9,16).
To date the only L1-specific translational product identified in human cells is encoded by ORF1. Human ORF1 protein (ORF1p) could be detected in the cytoplasm of a number of testicular germ cell tumors, in breast carcinoma, medulloblastoma, and a variety of transformed cell lines (17)(18)(19)(20). The only nonmalignant human tissue ORF1p has been detected in so far is the epithelium of normal mammary gland (17). ORF1p encoded by L1s in mice was verified in embryonal carcinoma cell lines, male and female germ cells, Leydig cells of embryonic testis, theca cells of adult ovary, and a large variety of transformed cell lines (21)(22)(23). However, all efforts to detect ORF2 proteins (ORF2p) in murine or human tissues using antibodies that detect baculovirus-produced or bacterially expressed ORF2p failed so far (10), and no direct evidence for in vivo translation of ORF2 has been provided hitherto.
Although it is clear that L1 retrotransposition events have occurred in somatic and germ cells (for review, see Ref. 10), only those events that take place in cells destined for the next generation are evolutionarily successful. In mammals these cells include primordial germ cells, germ cells, and early embryos. Thus, if L1 retrotransposition takes place in these cell types, functional proteins encoded by ORF1 and ORF2 are expected to be present because of their absolute requirement for L1 retrotransposition (4).
During human fetal development, primordial germ cells and supporting somatic cells, the precursor Sertoli cells, are enclosed in seminiferous cords, where primordial germ cells differentiate into prespermatogonia. These multiply by mitosis, arrest in the G 0/1 phase of the cell cycle, and subsequently enter another mitotic phase. Between embryonic weeks 10 and 22, primordial germ cells and all three types of prespermatogonia (mitotic, quiescent transitional, and mitotic transitional cells) appear side by side. After developmental week 26, only quiescent and mitotic prespermatogonia remain in fetal testis (24). In the seminiferous tubules of adult testis (Fig. 2), all intermediate cell types on the pathway to male gametes, like spermatogonia, spermatocytes, and spermatids, are present continuously throughout adulthood and can thus be viewed in single sections. Spermatogonia line the basal membrane of the seminiferous tubules. As germ cells undergo meiotic divisions and differentiation, they move toward the lumen, where mature spermatids are released. In the human testis, testosteroneproducing Leydig cells and seminiferous tubules are connected by the microvasculature, which plays an essential role in the distribution of hormones and nutritional factors.
To explore the potential of functional human L1s to expand during germ cell development, we examined gonadal and other tissues at different developmental stages for the presence of proteins encoded by ORF1 and ORF2. To this aim we raised a polyclonal antibody against polypeptides encoded by the 5Ј-end of ORF2 and employed a previously generated antibody directed against the p40 protein (25). Coexpression of ORF1-and ORF2-encoded proteins was detected in fetal prespermatogonia, spermatocytes, immature spermatids, as well as in somatic cell types, such as Leydig, Sertoli, and vascular endothelial cells and placental syncytiotrophoblasts. Identified ORF2-en-Gal4 DNA-binding domain and EN (Gal4-DBD-EN). As a negative control extracts from yeast cells transfected with the empty vector pGBKT7 expressing solely the Gal4-DBD were loaded in lane 3. D, immunofluorescence detection of transiently expressed ORF2p in the cytoplasm of HeLa cells using the anti-EN antibody. Cells were transfected with either RC-L1 expressing pJM101/L1 RP (a, c, e, g, and h) or with the parental empty vector pCEP4 (b, d, and f). Staining and immunofluorescence imaging were performed 48 h after transfection (a, b, c, and d) and after 38 days of hygromycin selection (e, f, g, and h). Magnifications of the details in the stippled boxes are shown in c and d. As negative controls, the immunofluorescence procedure was performed on pJM101/L1 RP -transfected cells with peptide-depleted primary anti-EN antibody (h) and without adding primary antibody (g). Similar transfection frequencies of pCEP4 and pJM101/L1 RP were controlled for by cotransfection with an EGFP reporter plasmid. Staining with anti-EN antibodies was visualized with Cy3-labeled secondary antibodies; nuclei were stained with DAPI.
FIG. 1. Raising polyclonal anti-EN antibodies to detect L1 ORF2 proteins. A, structure of a functional human L1 element. Abbreviations: C, cysteine-histidine-rich domain; TSD, target site duplication; UTR, untranslated region; AAA n , poly(A) tail. Red color represents the EN domain. B, chickens were immunized with peptides 1420 and 1419 representing amino acids 48 -63 and 152-166 of the EN domain, respectively. Peptide 1419 contains an additional N-terminal cysteine that is not part of the EN amino acid sequence. C, immunoblot analysis examining the specificity of the anti-EN antibodies raised. Lanes 1 and 2 were loaded with bacterially expressed L1-EN and whole cell extracts from chicken embryo fibroblasts (CEF), respectively. Extracts from yeast cells transfected with expression plasmid pCBW12 were separated in lane 4. pCBW12 encodes a fusion protein between the coded proteins were characterized by peptide mass fingerprinting analysis. ORF1-and ORF2-encoded proteins that were detected in fetal early meiotic cell types destined for the next generation may represent the enzymatic machinery for L1 retrotransposition and genomic expansion. The existence of both proteins in human vascular endothelial cells and the apparent association with differentiation and maturation of these cells may also have functional relevance for vasculogenesis and/or angiogenesis.

EXPERIMENTAL PROCEDURES
Tissue Samples-Human adult testicular tissues were obtained from 40 -70-year-old patients orchiectomized because of prostate cancer. Paraffin-embedded fetal tissues were provided by the Department of Pathology and Gynecopathology of the University Clinic Hamburg-Eppendorf. Research on adult and fetal human tissues was approved by the ethics commission of the institution.
Yeast Expression-To construct plasmid pCBW12 expressing a fusion of the L1-EN domain to the C terminus of the Gal4 DNA-binding domain (Gal4-DBD-EN), the DNA encoding L1-EN was inserted inframe between NdeI and BamHI restriction sites of yeast expression vector pGBKT7 (Clontech). The empty parental vector pGBKT7 or pCBW12 was transfected into yeast AH109 cells to express Gal4-DBD proteins and Gal4-DBD-EN fusion proteins. Growth and manipulation of the yeast strain and preparation of cell extracts were performed according to standard procedures (26).
Primary Antibodies-The well characterized antibody AH40.1 directed against L1 ORF1-encoded p40 protein was a gift from Thomas G. Fanning, Armed Forces Institute of Pathology, Washington, D. C. (19,25). To raise antibodies detecting the EN domain (amino acids 1-239) of ORF2p, peptides 1420 and 1419, covering amino acids 48 -63 and 152-166 of ORF2p, respectively (see Fig. 1, A and B), were chosen, based on algorithms that predict antigenic sites (Lasergene Software, DNASTAR). Numbering corresponds to the amino acid positions in the active L1.2 element (27). Two chickens were immunized commercially (Eurogentec) by three boost injections of each of the two peptides. IgY antibodies were prepared from egg yolks (28) and purified by affinity chromatography. For that purpose peptides 1420 and 1419 (2.5 mg each) were covalently coupled to 1 g of CNBr-activated Sepharose 4B (Amersham Biosciences). The gel was packed into a column and washed using standard methods. After equilibration with storage buffer (0.1 M phophate buffer, pH 7.6, 0.1 M NaCl, 0.1% NaN 3 ), the IgY fraction was passed over the column twice. Subsequently, the column was washed with storage buffer, and bound antibodies were eluted with 10 bed volumes of 100 mM glycine, pH 2.5, and mixed with 1 bed volume of 1 M Tris-HCl, pH 8.0. The affinity-purified antibody, referred to as anti-EN antibody and directed against peptides 1419 and 1420, was dialyzed against storage buffer and stored at a concentration of 12.5 mg/ml at Ϫ80°C. All experiments described were performed using the affinitypurified anti-EN antibody.
Immunoblot Analysis-Purified L1-EN and cell extracts were separated on an SDS-12% polyacrylamide gel and electroblotted onto a Hybond-P membrane (Amersham Biosciences). After transfer, membranes were blocked in phosphate-buffered saline (PBS) containing 4% bovine serum albumin, incubated with the affinity-purified polyclonal anti-EN antibody at a dilution of 1:1,000 for 45 min, washed with PBS, and incubated with rabbit anti-chicken IgG horseradish-peroxidaseconjugated secondary antibody (Sigma) at a dilution of 1:10,000. The membrane was washed three times with Tris-buffered saline (TBS), 0.5% Tween 20, once with TBS, 3% Tween 20, and twice with TBS, and immune complexes were visualized using the Amersham ECL enhanced chemiluminescence detection kit.
Immunofluorescence Imaging-HeLa cells were grown at 37°C in an atmosphere containing 7% carbon dioxide and 100% humidity in high glucose Dulbecco's modified Eagle's medium lacking pyruvate. Dulbecco's modified Eagle's medium was supplemented with 10% fetal bovine calf serum and 0.4 mM glutamine. HeLa cells were transfected with 2 g of pJM101/L1 RP (5) or pCEP4 (Invitrogen) using FuGENE 6 transfection reagent (Roche Applied Science). 48 h after transfection, one set of transfected cells was fixed in 4% formaldehyde and preabsorbed with blocking buffer (10% donkey serum, 5% horse serum, 1% bovine serum albumin in PGS (0.5% glycine, 0.05% saponin in PBS)). Subsequently, coverslips were incubated with the affinity-purified anti-EN antibodies (1:50 dilution in PGS buffer) for 1 h, washed in PGS, and reacted with a 1:500 dilution of Cy3-labeled donkey anti-chicken IgG antibodies (Dianova). After washing and nuclear staining with DAPI, mounted coverslips were photographed using a Zeiss Axiovert 100 microscope equipped with a digital camera. Three days after transfection, another set of transfected cells was selected for the presence of pJM101/L1 RP or pCEP4 by growth in Dulbecco's modified Eagle's medium containing 200 g/ml hygromycin. After 38 days of selection, hygromycin-resistant cells were fixed in formaldehyde, preabsorbed with blocking buffer, and stained as described before.
Immunohistochemistry and Immunocytochemistry-Immunohistochemistry was carried out as described previously in detail (29). Briefly, tissues were fixed in Bouin's solution or 8% formalin, embedded in paraffin, and sectioned serially at 5-7 m. Sections were deparaffinized, rehydrated, and incubated with the anti-EN or AH40.1 antibodies at a final concentration of 12.5 ng/ml or at a 1:100 dilution, respectively, in PBS buffer containing 0.2% bovine serum albumin, 0.1% NaN 3 for 24 -48 h at 4°C. After washing, sections were incubated with biotinylated rabbit anti-chicken IgG (Abcam) or biotinylated swine anti-rabbit IgG secondary antibody (Dako), respectively. The immunoreaction was visualized using a combination of the peroxidase-antiperoxidase and avidin-biotin complex methods enhanced by a modified nickel-enhanced glucose oxidase technique (29). To test for the specificity of the anti-EN antibody, peptides 1419 and 1420 were immobilized separately on CNBr-activated Sepharose 4B and preincubated overnight with the antibody at a ratio of 1:3 (12.5 ng of anti-EN antibody and 37.5 ng of peptides/ml). The supernatant of this mixture containing the peptide-depleted antibody was used for immunohistochemistry in control experiments. The following controls were performed in addition: (i) primary or secondary antibody was replaced by PBS, (ii) peroxidase reaction alone was visualized without antibody pretreatment. Preabsorption experiments reexamining the specificity of the AH40.1 antibody were performed as described previously (19). Immunostained sections were counterstained with Calcium Red and photographed with a Leica microscope equipped with a digital camera.
Human dermal microvascular endothelial (HDME) cells, derived from dermis (PromoCell, Heidelberg, Germany) were cultured to confluence on gelatin-coated dishes in Endothelial Cell Growth Medium MV (PromoCell) and fixed in 4% formaldehyde for 30 min. After washing with PBS, cells were incubated overnight at room temperature with anti-EN antibodies. The peroxidase reaction was carried out as described above.
Immunoprecipitation and Matrix-assisted Laser Desorption/Ionization Time-of-flight (MALDI-TOF) Mass Spectrometry (MS)-HDME cell lysates were incubated with affinity-purified anti-EN antibodies immobilized on CNBr-activated Sepharose 4B beads overnight at 4°C. Beads were pelleted by centrifugation for 1 min at 600 ϫ g and washed three times with 1 ml each of TBS and 0.5% Tween 20; TBS and 350 mM NaCl and 0.5% Tween 20; TBS and 350 mM NaCl and 3% Tween 20, and 3 ϫ PBS. Finally, the bead fraction was separated on an SDS-12% polyacrylamide gel and visualized by silver staining. Two immunoprecipitated proteins with estimated molecular masses of ϳ45 and ϳ130 kDa (see Fig. 4B) were excised from the gel, digested with trypsin, and extracted sequentially with 40% acetonitrile and 0.1% trifluoroacetic acid. After evaporation of the solvent, the peptides were resuspended in 0.1% trifluoroacetic acid, mixed with an equal volume of 1% ␣-cyano-4hydroxycinnamic acid in 50% acetonitrile and 0.1% trifluoroacetic acid, and spotted onto a stainless steel plate for analysis by MALDI-TOF MS (Axima CRF, Shimadzu). Identified peptides were compared with peptide sequences of annotated coding regions by applying the MASCOT search program.

Anti-EN Antibodies Directed against L1
ORF2p-To investigate human germ line tissues for expression of L1 ORF2p we raised antibodies against peptides within the N-terminal EN domain of ORF2p (Fig. 1, A and B). These anti-EN antibodies recognized the bacterially expressed EN domain (amino acids 1-239) with an apparent molecular mass of 28 kDa (13) (Fig.  1C, lane 1) and Baculovirus-produced 150-kDa ORF2p. 2 As a negative control we used whole cell extracts from chicken embryo fibroblasts because the chicken genome does not contain LINE-1 sequences. The chicken-specific retrotransposon CR1 is only distantly related to mammalian L1 elements and differs greatly in the relevant ORF2 sequences (30,31). As expected, the anti-EN antibody failed to detect any protein in these extracts (Fig. 1C, lane 2). Next, we tested the antibody using a yeast plasmid expressing a fusion protein between the Gal4-DBD and the L1-EN domain (Gal4-DBD-EN). The fusion protein with an apparent molecular mass of 48 kDa was recognized in extracts from transfected yeast cells but not in extracts from control cells expressing the Gal4-DBD alone (Fig. 1C,  lanes 3 and 4).
To test the anti-EN antibody for its ability to detect native ORF2p in situ, we transfected the episomal retrotransposition reporter plasmid pJM101/L1 RP carrying an RC-L1 element under the control of a heterologous promoter (5) into HeLa cells and performed immunofluorescence experiments (Fig. 1D).
Although HeLa cells that were transfected with the empty vector pCEP4 (Fig. 1D, b, d, and f) as well as untransfected HeLa cells (data not shown) displayed only weak background immunoreactivity that may be explained by endogenous ORF2 expression, transfection with pJM101/L1 RP resulted after only 48 h in much more intense immunostaining of the cytoplasm (Fig. 1D, a and c) of a subset of cells. 38 days of hygromycin selection of pJM101/L1 RP -transfected cells resulted in strong fluorescence of the cytoplasm of each cell in the culture (Fig.  1D, e). In contrast, pCEP4-transfected hygromycin-resistant cells exhibited only weak background activity (f) that is also observed in untransfected HeLa cells. Because pJM101/L1 RP differs from pCEP4 exclusively by an active RC-L1, the data strongly suggest that the difference in immunoreactivity is caused by the interaction of the affinity-purified anti-EN antibody with overexpressed L1 RP ORF2p. There was no immunoreactivity detectable, when pJM101/L1 RP -transfected hygromycin-resistant cells were immunostained with peptidedepleted primary anti-EN antibody, which was generated by preabsorption with peptide 1419 and 1420 immobilized on Sepharose beads (Fig. 1D, h) or when the primary antibody was omitted (g).
To evaluate the possibility that cellular toxicity caused by overexpressed functional ORF2 proteins or by retrotransposition events, and not the detection of the ORF2 protein itself, might be responsible for the significant fluorescence in pJM101/L1 RP -transfected cells, an inactive L1-reporter was transfected into HeLa cells, and immunostaining with the anti-EN antibody was performed (data not shown). Transfection and 38 days of hygromycin selection of the inactive L1.2Areporter construct harboring a missense mutation in an endonuclease active site (H230A) (11) resulted in strong staining of each transfected cell (data not shown) similar to the staining observed after transfection of the active L1 RP-reporter construct (e). Additionally, the generation time of the L1.2A mutant-expressing cells did not differ from pJM101/L1 RP -transfected cells. Taken together, these results indicate that the detected fluorescence signals are a consequence of the specific interaction between anti-EN antibody and ORF2p and are not caused by cells dying of any toxic effect of L1-encoded proteins or deleterious L1 retrotransposition events.
Our finding that pCEP4-transfected and untransfected cells are stained weakly by the anti-ORF2 antibody (Fig. 1, b, d, and f) is consistent with previous results from Skowronsky and Singer (32), who detected heterogeneously sized L1 transcripts in various human cell lines including HeLa (32). Translation of at least a subset of those L1-specific transcripts in HeLa could be demonstrated by Leibold and co-workers (25), detecting low amounts of reactive L1-ORF1p in the same cell line.
Detection of ORF1p and ORF2p in Adult Testicular Tissue-To explain the evolutionary genetics of L1, L1 retrotransposition is expected to occur during germ cell development. Therefore, we first looked for ORF2 expression in the adult testis ( Figs. 2 and 3). Within the seminiferous tubules, extensive immunostaining with the anti-EN antibody was observed in the cytoplasm of Sertoli cells, in a subset of secondary spermatocytes and immature spermatids, as well as in some residual bodies (Fig. 3, A and D, and Table I). There was no immunoreactivity detectable in spermatogonia at the basis of the seminiferous tubules, primary spermatocytes, peritubular myofibroblasts, vascular smooth muscle cells, and interstitial tissue. Unexpectedly, strong immunostaining was also observed in some somatic cell types: in endothelial cells of capillaries and large blood vessels, in interstitial cells covering the seminiferous tubules and blood vessels (covering cells (33)), and in the cytoplasm of Leydig cells (Fig. 3, A, C, and D). There was no specific immunostaining when the anti-EN antibody was incubated with peptide 1420 immobilized on Sepharose beads prior to its use for immunohistochemistry (Fig. 3B). In contrast, specific immunostaining was not affected by either preabsorption of the anti-EN antibody with immobilized peptide 1419 (Fig. 3C) or preincubation with Sepharose beads alone (Fig. 3D). These results indicate that the anti-EN antibodies are directed predominantly against the epitope represented in peptide 1420.
Staining with antibodies to ORF1p also revealed immunoreactivity in the cytoplasm of Sertoli cells, in secondary spermatocytes, immature spermatids, residual bodies, and in Ley- LINE-1 ORF2-encoded Proteins in Human Germ Line Tissue dig cells but not in spermatogonia (Fig. 3E). Furthermore, anti-ORF1 antibodies stained endothelial cells of blood vessels, although the fraction of immunoreactive blood vessels was reduced compared with staining with anti-EN antibodies. Se-rial control sections exposed to the secondary antibody alone (Fig. 3F) and to preabsorbed anti-ORF1 antibody (data not shown) did not exhibit any specific immunostaining. Taken together, these results indicate that the pattern of cell types in adult human testis immunoreactive to the EN domain of ORF2p is nearly identical to that stained with anti-ORF1 antibodies (Table I).
Identification of ORF2p by MALDI-TOF MS-Vascular endothelial cells from all human tissues examined so far exhibited ORF2p immunoreactivity (Figs. 3 and 6). 3 To analyze ORF2 products recognized by our antibody in this cell type in more detail, we used primary cultured cells of the same cell type, HDME cells, because we expected that cells of the same cell type would express the same genomic L1 copies. ORF2 immunoreactivity in HDME cells was mainly cytoplasmic, but a considerable fraction of cells stained weakly or not at all (Fig.  4A). After immunoprecipitation of HDME cell lysates with anti-EN antibodies, we visualized in silver-stained polyacrylamide gels that the antibody specifically precipitated two polypeptides with estimated molecular masses of ϳ130 and ϳ45 kDa (Fig. 4B). For identification, both polypeptides were excised from the gel, digested with trypsin, and analyzed by MALDI-TOF MS (Fig. 4C and 5A). In the case of the ϳ130-kDa protein we found 10 mass peaks for tryptic peptides that are identical in sequence to tryptic peptides of ORF2 proteins from two RC-L1 elements, L1.3 and L1 RP (Fig. 4D) (5). L1 RP ORF2 predicts an approximately 150-kDa protein encompassing 1275 amino acids (Fig. 4D); the 10 tryptic peptides identified match 13% of L1 RP ORF2p. Because of the high molecular mass of the immunoprecipitated ϳ130-kDa polypeptide it is not possible to make any precise comment on the molecular mass of this peptide based on its mobility in the polyacrylamide gel. Additionally, there is the possibility that the immunoprecipitated complete ORF2p migrates slightly faster as a result of unusual protein compositions. Therefore, it could be that the peptide that is comigrating with the 133 kDa band of the molecular mass marker (Fig. 4B) is in fact the full-length ORF2-encoded protein with a predicted molecular mass of 150 kDa. Peptide mass fingerprinting analysis of the ϳ45-kDa polypeptide (Fig.  4B) identified five tryptic peptides (Fig. 5A), which matched those encoded by the 5Ј-truncated ORF2 of a genomic L1PA2 copy (GenBank accession no. S23650) described previously (34) (Fig. 5B). This L1 copy contains two nonsense mutations within the central region of ORF2 and several one-nucleotide deletions. Two of the identified tryptic peptides (Fig. 5; peaks 4 and 5) were identical in sequence to those from L1 RP ORF2p (Fig.  5), whereas the remaining three peptides (peaks 1-3) contained several point mutations as compared with L1 RP ORF2. By searching the Ensembl human genome data base (proteinview, contigview (35)) these peptides identified two closely related 5Ј-truncated L1PA2 copies located on chromosomes 1p32 and 1p35. Both L1 copies have coding capacity for the identified 45-kDa polypeptide, and their sequences are consistent with the L1PA2 sequence described by Hohjoh and co-workers (34).
Because primary HDME cells represent the same cell type as adult testicular vascular endothelial cells which also stained positive for ORF2p (Fig. 3), we concluded that ORF2-encoded proteins identified by MALDI-TOF MS are also responsible for the ORF2 immunoreactivity observed in vascular endothelial cells in the tissue sections. However, given the high copy number of L1s in the genome we do not know whether L1-encoded proteins detected in other somatic cell types and in germ cells are the result of the products of the same expressed L1 elements.
To sum up it can be said that all data obtained in the described experiments argue for the specific interaction of our anti-EN antibody with ORF2-encoded proteins and against any cross-reaction with unspecific cellular proteins.
Detection of ORF1p and ORF2p in Fetal Testicular and Placental Tissues-To obtain insight into the temporal and spatial regulation of L1 expression in the germ line we examined testicular tissue of an 18-and a 28-week fetus (Fig. 6). Immunostaining with the anti-EN antibody was observed mainly in the cytoplasm of prespermatogonia as well as in Leydig cells (Fig. 6, A-C and Table I). L1 expression in prespermatogonia was also confirmed by the presence of ORF1p (Fig. 6D). Furthermore, endothelial cells of more mature blood vessels exhibited immunostaining for both ORF1p and ORF2p (Fig. 6A), whereas those of small and immature capillaries were negative. Other fetal testicular cell types like vascular smooth muscle cells, peritubular myoid cells, and cells of the interstitial connective tissue did not exhibit any immunoreactivity with the applied antibodies (Table I). No immunoreactivity was detected when the peptide-depleted anti-EN antibody was used (Fig. 6E). All somatic cell types of the testicular tissue that were immunoreactive in our study are characterized by the presence of receptors for the steroid hormones, androgen and/or progesterone (36). This finding led us to examine other fetal tissues that contain cell types that are major targets of those steroids, such as epididymis and placenta. We observed ORF1 and ORF2 immunoreactivity not only in the columnar epithelium of the epididymis (Fig. 6, F and G), but also in syncytiotrophoblasts and endothelial cells of placental blood vessels (Fig.  6, I and K). All three cell types display receptors for androgen or progesterone. In contrast, neither of the two L1-specific antibodies revealed any immunoreactivity in interstitial tissue, vascular smooth muscle cells, epididymal peritubular myoid 3 S. Ergü n and G. G. Schumann, unpublished results.  (Table I). Serial control sections incubated with the peptide-depleted anti-EN antibody (Fig. 6, H and L) or with preabsorbed anti-ORF1 antibody (data not shown) did not lead to any immunoreactivity.

DISCUSSION
Although an intact ORF2 was demonstrated to be essential for L1 retrotransposition (4,11), all efforts to detect ORF2 proteins have failed so far (10). Here we have generated an antibody to polypeptides of the EN domain of human L1 ORF2p to discern ORF2p immunoreactivity in human adult and fetal FIG. 4. Immunoprecipitation and peptide mass fingerprinting analysis of ORF2 proteins. A, immunolocalization of ORF2p in HDME cells using anti-EN antibodies (ϩanti-EN). As a control, immunoperoxidase reaction was performed without primary anti-EN antibodies (Ϫanti-EN). Magnification, ϫ650. B, immunoprecipitation of ORF2p from HDME cell lysates. The immunoprecipitate with anti-EN antibodies obtained from HDME cell lysates was resolved by SDS-PAGE and visualized by silver staining (lane 2). As a control, immunoprecipitation was performed without adding cell lysate (lane 1). Two bands with apparent masses of ϳ130 and ϳ45 kDa appear specifically FIG. 5. Peptide mass fingerprinting analysis of the immunoprecipitated 45-kDa polypeptide. A, MALDI-TOF MS revealed five mass peaks (peaks 1-5) of tryptic peptides corresponding to those encoded by L1PA2 copies on chromosomes 1p32 and 1p35 (GenBank accession no. S23650). The table beneath lists the deviations (Delta) between observed (mass a ) and calculated (assigned b ) masses with corresponding amino acid sequences. Amino acid residues differing from the peptide sequence of the functional L1 RP ORF2p (Fig. 4D) are shaded. Peaks without numbering result from contaminating keratinspecific tryptic peptides; T, peaks resulting from trypsin autodigestion; m/z, mass-to-charge ratio. B, graphical representation of the data base hits. Matched tryptic peptides within L1PA2 ORF2p (GenBank accession no. S23650) are shown. Amino acid position 2 of the presented peptide sequence corresponds to amino acid position 565 of the functional L1 RP ORF2p shown in Fig. 4D. in lane 2. The predominant 68-kDa band is caused by anti-EN antibodies (anti EN). C, peptide mass fingerprinting analysis of the immunoprecipitated ϳ130-kDa polypeptide. MALDI-TOF MS of the trypsindigested ϳ130-kDa polypeptide revealed 10 mass peaks (peaks 1-10) of tryptic peptides identical in sequence to tryptic peptides from L1.3 and L1 RP . Mass peaks resulting from trypsin autodigestion are marked (T). The table beneath lists the deviations (Delta) between observed (mass a ) and calculated (assigned b ) masses with corresponding identified amino acid sequences (Sequence). Peaks without numbering result from contaminating keratin-specific tryptic peptides; m/z, mass-to-charge ratio. D, graphical representation of the data base hits. Matched tryptic peptides within L1 RP ORF2p (GenBank accession no. S65824) are shown in red. LINE-1 ORF2-encoded Proteins in Human Germ Line Tissue 27760 tissues. We report the presence of both ORF1 and ORF2 translation products at different stages of the human male germ cell development. Coexpression of both ORFs was observed in basally as well as centrally localized prespermatogonia of seminiferous cords during the early fetal period (week 18) and was found to be restricted to the basal tubular cell layers at a later stage of the fetal development (week 28). It remains unclear whether L1 expression starts during the migration of prespermatogonia or begins after the cells have settled in seminiferous cords. Immunostaining indicates that p40 (18) as well as ORF2 proteins are localized in the cytoplasm of L1 expressing cells (Fig. 6, A-D).
Immunostaining of adult testicular tissue uncovered translation products from both ORFs in spermatocytes and immature spermatids, but not in spermatogonia, the stem cells of the germinal epithelium (Figs. 2 and 3 and Table I). Thus, L1 expression seems to be activated after quiescent spermatogonia differentiated into primary spermatocytes that are entering meiosis I. These results are consistent with the previous finding that both full-length L1 RNA and ORF1 protein are present in spermatocytes but absent in spermatogonia of prepuberal mouse testis (21). Taken together, our data imply that L1 proteins are only expressed in differentiating and proliferating cells of the germinal epithelium. Provided that these proteins are functional, their presence would fulfill the requirement for transposition in cells that spawn the next generation. This is the first observation of proteins encoded by both L1-specific ORFs in nonmalignant human germ line tissue.
Factors that determine germ line specificity of L1 transcription have not been identified, but it has been suggested that the SOX family of transcription factors and the ubiquitous nuclear transcription factor YY1 may be involved (37,38). The interpretation of DNA methylation patterns by methyl-CpG-binding proteins was shown to be a further potent regulator of L1 expression (39). Interestingly, MeCP2, a major repressor of transcription of methylated L1s (39), is widely expressed in mammalian tissues, but undetectable in human 4 as well as in murine germ cells (40).
This study also revealed ORF1 and ORF2 expression in a variety of somatic cells, specifically, Sertoli, Leydig, and testicular covering cells, epididymal columnar epithelial cells, placental syncytiotrophoblasts, and vascular endothelial cells. Notably, our results are in line with previous studies, demonstrating mouse L1-ORF1 expression in the cytoplasms of mouse embryonal spermatogonia, Leydig cells, placental syncytiotrophoblasts, and in spermatocytes and spermatids (21,23). Because ORF1 immunoreactivity was detected in Leydig cells of the mouse testis and in theca cells of the mouse ovary, Trelogan and Martin (23) noted a possible correlation between L1 expression and the production of steroid hormones. Besides Leydig cells and placental syncytiotrophoblasts, we observed human L1 expression in several cell types not producing steroids. Thus, there is no direct hint for an interrelationship between L1 expression and steroid production in human tissues. Nevertheless, it is worth mentioning that all immunostained somatic cell types identified in our study exhibit receptors targeted by either androgen or progesterone (36). It remains to be studied whether there is a connection between the presence of these steroid receptors and the up-regulation of L1 expression in certain somatic cell types.
Although it has been known for several years that high expression of human endogenous retroviruses is one of the characteristics of syncytiotrophoblasts (41), it is quite intriguing that significant protein expression of L1s in human and mouse is demonstrated in the same specific cell type. Although evidence has recently accumulated suggesting that endogenous retroviral gene expression may be involved in mediating the cell fusion observed in the placenta (42), it is completely unclear whether the expressed L1-encoded proteins have any function for the host cell.
In contrast to previous findings by Bratthauer and Fanning (19), we observed immunostaining of different cell types in nontransformed adult testicular tissue with the AH40.1 antibody. This can be explained by differences in the fixation and staining methods between the two laboratories. Although the previous study used formalin for fixation resulting in a reduced staining (19), our tissues were fixed in Bouin's solution as part of a more sensitive method to retain antigenic sites (29).
In the adult testis ORF2p was observed in all vascular endothelial cells, whereas ORF1p staining was restricted to endothelial cells of the microvasculature. This could be explained either by a lower expression level of ORF1p relative to ORF2p or by a lower affinity of the AH40.1 antibody to its antigen compared with the anti-EN antibody. Our findings indicating high expression levels of ORF2p in endothelial cells of mature but not of immature blood vessels in developing fetal testicular tissue and in human placenta suggest increasing expression during maturation of these cells. If the identified L1 proteins were functional and retrotransposition could occur in these cells, this finding would be surprising because endothelial cells are considered as genetically stable. It is also unclear what the purpose of retrotransposition would be in these cells because somatic retrotransposition is an evolutionary dead end. A new insertion in a somatic cell cannot be passed onto the next generation, and, if the insertion harms the host, the precursor of the insertion might not be passed on. The high amount of L1-encoded proteins in quiescent vascular endothelial cells and their obvious accumulation accompanying endothelial maturation implicate a functional relevance of L1 in vasculogenesis and/or angiogenesis. Because targeting the endothelial cells is a major focus of anti-angiogenic tumor therapy (43) it is of great interest to find out if and how L1-encoded proteins influence the behavior of endothelial cells, particularly of those in tumors.
Another explanation for the detection of ORF2p alone without ORF1p is the expression of mutated or 5Ј-truncated L1 copies that express ORF2p but lack coding capacity for a functional ORF1p. Although elements harboring a disrupted ORF1 gene, but a functional ORF2 gene, are not able to promote either their own transposition (4) or the generation of processed pseudogenes (44), they should still be competent for the mobilization of Alu sequences (45). Thus, the adventitious synthesis of ORF2p from even truncated L1 elements may be a potential source of RT which could catalyze both somatic or germ line retroposition of SINE elements. Therefore, not only RC-L1s may constitute a genetic load to their host, but also L1 copies devoid of a functional ORF1. This hypothesis is supported by the recent observation that less than full-length L1 elements, at times, were under negative selection (46). It was suggested that some long (Ͼ500 bp) non-full-length L1 elements could produce L1 products, such as L1 RNA or an active RT, which could be deleterious and make them subjects to purifying selection (46).
Peptide mass fingerprinting analysis that followed protein immunoprecipitation with our anti-EN antibody uncovered two polypeptides in HDME cell lysates with estimated molecular masses of 45 and 130 kDa featuring sequence identity in a number of tryptic peptides to ORF2 proteins. Although all data obtained on the amino acid sequence of the ϳ130-kDa peptide are so far consistent with a functional ORF2p, it is still possible that the peptide is encoded by a nonfunctional L1 copy because our analysis by mass spectrometry covers only 13% of the predicted ORF2p sequence. Also, none of the tryptic peptides identified by mass spectrometry is unique to the so-called "hot" RC-L1 elements (3) because they correspond to sequences that are highly conserved in L1 ORF2p and are also present in very much less active members of the Ta (L1PA1) family. They even differ only by a single amino acid substitution from the corresponding coding regions of the older now extinct L1PA2 and L1PA3 families (47). Therefore, instead of representing functional L1 activity the ϳ130-kDa polypeptide may represent the adventitious translational product of nonfunctional readthrough transcripts. Peptide mass fingerprinting analysis of the ϳ45-kDa protein indicates that it derived from the mutated ORF2 of nonfunctional L1PA2 elements on chromosome 1p32 and 1p35 (34). Because zinc finger domains have recently been shown to mediate protein-protein interaction (48,49), interaction of the ϳ45-kDa protein with the ϳ130-kDa protein through the cysteine-histidine-rich, putative zinc-knuckle domains at their C termini may account for the fact that our anti-EN antibody, directed against the N-terminal end of ORF2p, pulled down a ϳ45-kDa protein representing the Cterminal portion of the protein. If this is the case and the ϳ45-kDa protein does not interact with the anti-EN antibody directly, as suggested by our MS data, immunohistochemical staining of HDME cells (Fig. 4A) should be derived from the ϳ130-kDa ORF2 polypeptide. At this point we cannot exclude the possibility that other EN domain including ORF2-specific polypeptides are at least in part responsible for the immunohistochemical staining observed in testis, epididymis, and placenta.
The first indirect evidence for in vivo translation of ORF2p was provided by Maxine Singer's laboratory using an L1 expression vector with the L1-ORF2p coding region replaced by a lacZ reporter cassette in the teratocarcinoma cell line NTera2D1 (50). After transfection of the L1 reporter construct, ␤-galactosidase-expressing cells were detected, indicating that ORF2 was translated. Although the translation of ORF2 was shown to be enhanced when the translation of ORF1 is decreased, the absence of ORF1 translation led to a decrease in the expression of ORF2. The authors presented several pieces of evidence indicating that the translation of L1Hs ORF2 is most likely to be initiated internally. It was suggested that translation of ORF2 may occur by some form of internal ribosomal entry such as by the use of an internal ribosomal entry site or by ribosomal shunting (10,50). The fact that we identified exclusively ORF2-and not ORF1-specific peptides in our peptide mass fingerprinting analyses is consistent with the assumption that ORF1 and ORF2 are translated separately (50). Also, Ilves et al. (51) provided evidence that ORF2 of the rodent L1 is translated by reinitiation or internal initiation rather than frameshifting, even though, in contrast to L1Hs, the two ORFs overlap and are out of frame.
Our data also show that immunoprecipitation combined with peptide mass fingerprinting analysis is a valuable tool to investigate the expression status of replication-competent and mutated, inactive L1 elements in individual cell types. Further investigation will fruitfully complement the genetic studies on L1 composition of the human genome.