An Endogenous Retrovirus Derived from Human Melanoma Cells

Human endogenous retroviruses (HERVs) represent a cellular reservoir of potentially pathogenic retroviral genes. The human genome harbors 1–2% of endogenous retroviral sequences. Prototype HERV-K is the only known human endogenous retrovirus with open reading frames for the structural and enzymatic proteins gag, prt, pol, and env (1, 2). In addition, HERV-K encodes a nuclear RNA export factor, termed rec (formerly corf), a functional homologue of the HIV-1 Rev protein (3, 4). In general, the expression of ERVs is repressed. Exogenous factors such as UV radiation, chemicals, related exogenous retroviruses, as well as endogenous factors such as hormones and cytokines can activate repressed ERVs. Although full-length mRNA expression is detectable in many tissues, expression of all proteins and enzymes as well as particle production has been demonstrated only in cell lines established from human teratocarcinomas (1, 5, 6, 7). However, although endogenous retroviruses of other species than humans were shown to be infectious (8), no infectious HERV-K has been detected thus far. One reason for the absence of infection was suggested to be because of the lack of Env-processing, a prerequisite for viral entry into the host cell (9).

Melanoma is the most malignant type of skin cancer in man with an alarmingly increasing incidence and death toll worldwide (10). It arises from melanocytes (pigment cells) in the epidermis or melanocytic nevi (moles), and UV radiation is suspected to play a role in its development (11). C57BL/6 mouse melanomas contain MelARV, a retrovirus capable of infecting cultured murine melanocytes. In some instances, MelARV can also induce malignant transformation (12). MelARV probably originated from the endogenous ecotropic provirus Emv-2 that exists in all cells of C57BL/6 mice. This provirus is defective and unable to generate replication-competent retrovirus. It has been postulated that MelARV emerged as a result of a recombination between ecotropic Emv-2 and nonecotropic sequences during malignant transformation or tumor progression. MelARV does not contain an oncogene, but it was found to be inserted into the c-maf proto-oncogene in cells transformed after infection (13). As for human melanomas, studies by Balda et al. (14) performed in the 1970s showed particles that package RNA and RT and possess the density characteristics of RNA tumor viruses. In addition, particles with a morphology similar to C-type virions were detected in primary melanoma and metastatic melanoma cells (15, 16). A recent study suggesting that retroviruses are associated with melanoma showed that HERV-K antigens are targeted by cytolytic T lymphocytes in melanoma patients (17).

We characterized particle preparations derived from human melanoma cells. The particles contain retroviral Pol, Gag, and Env proteins and package sequences homologous to HERV-K. In addition, we find the nonstructural protein Rec to be expressed in these cells.

HERV-K-specific antibodies, rabbit anti-HERV-K Gag (5), rabbit anti-HERV-K Rec (18), and goat anti-HERV-K Env, recognizing the transmembrane domain (TM), were used.

Primary melanomas, cutaneous melanoma metastases, lymph node metastases of melanoma, and benign melanocytic lesions (junctional, compound, dermal, congenital, and dysplastic nevi) were obtained from patients, verified by histopathology, and characterized by in situ hybridization, RT assays and immunohistochemistry.

The melanoma cell lines SK-Mel-28 and SKMel-1 were obtained from the American Type Culture Collection, and melanoma cell line 518A2 (19) was provided by Peter Schrier from the University of Leiden (Leiden, the Netherlands). Cell line Mel-Juso was provided by Jürgen Lehman from the Institute of Immunology, University of Munich (Munich, Germany). Human neonatal melanocytes (NHEMs) neo 5935, neo 4528, and neo 6083 were obtained from Szabo Scandic (Vienna, Austria). Madin-Darby bovine kidney (MDBK) cells were also obtained from American Type Culture Collection.

For electron microscopy, sequencing and immunoblotting supernatants were filtered through a 0.22-μm low-protein membrane and purified on iodixanol density gradients or cushions. Iodixanol is an iodinated, nonionic density gradient medium (Nycomed Pharma, Oslo, Norway). It has a low viscosity and provides isoosmotic conditions up to densities of 1.32g/ml (20). The supernatants were overlaid on a cushion of 5 ml of 50% iodixanol. The tubes were centrifuged in a SW28 rotor at 45,000 × g for 2 h at 4°C, and the supernatant was removed from the tubes by suction, leaving a volume of 4 ml of the medium in proximity of the cushion. This fraction was harvested, pelleted in a SW41 rotor at 150,000 × g for 90 min at 4°C, resuspended in PBS, and analyzed. Alternatively, for additional purification, the cushion-derived fraction was loaded on iodixanol density gradients, and the fraction corresponding to a density of ∼1.16 g/ml was harvested. The harvested fractions were diluted in PBS and pelleted in a SW41 rotor at 150,000 × g for 90 min at 4°C and resuspended in PBS.

Particle preparations for infection studies were obtained by loading supernatants filtered through a 0.22-μm low-protein membrane on 20% sucrose cushions and centrifugation for 2 h at 150,000 × g. The resulting pellets were resuspended in PBS.

Supernatants from melanoma cell lines and normal human melanocytes were centrifuged at 3000 × g at 4°C and sterile-filtered through a 0.22-μm low-protein membrane (Nunc) to remove cells and cellular debris. The clarified supernatants were centrifuged for 20 min at 250,000 × g at 4°C in a Beckman SW50.1 rotor. Pellets were rinsed with PBS and centrifuged for 15 min at 250,000 × g and resuspended. The resulting particle suspensions were used as an enzyme source for RT assays.

Pelleted particles derived from 4-ml cell-free supernatants of ∼10⁶ cells were analyzed. RT activity was determined by fluorescent probe-based product-enhanced reverse transcriptase assay as previously described, with modifications (21). First, pellets were suspended in lysis buffer (Roche). Then RT was performed, using the suspended pellet as source of enzyme, the primer 3′-A10 (5′-CACAGGTCAAACCGCCTAGGAATG-3′) and 0.3 μg of MS2-RNA as a template (Roche no. 165948). To limit unspecific RT, 0.5 μg of calf-thymus DNA (Sigma no. D4522) were added. After incubation at 42°C for 1 h, a 5-μl aliquot of this reaction was amplified by real-time PCR, by adding 25 μl of TaqMan Universal PCR Master Mix (Applied Biosystems no. 4304437), 1 μl of the primers 3′-A10 and 5′-A11 (5′-TCCTGCTCAACTTCCTGTCGAG-3′) at a concentration of 10 μm each, 1 μl of fluorescent probe-based product-enhanced reverse transcriptase probe (genXpress, 10 μm; 5′(FAM)-TCTTTAGCGAGACGCTACCATGGCTA-(TAMRA)3′), and 17 μl of H₂O. The resulting mixture was then amplified by incubating 10 min at 95°C, followed by 40 cycles at 94°C for 20 s and 64°C for 1 min in a SD 7700 (Perkin-Elmer). The calibration curve was generated by plotting the RT activity of a serial dilution of Moloney murine leukemia virus reverse transcriptase (Superscript II Life Technologies, Inc., no. 18064-014).

Iodixanol cushion-purified particles from SK-Mel28 supernatants were pelleted, resuspended in PBS, and analyzed. The amount of soluble proteins was quantified by a modified Bradford analysis (Bio-Rad, Richmond CA). Five μg of total protein were applied per lane and separated by SDS-PAGE (10%). Proteins were transferred to polyvinylidene difluoride membranes (Millipore, Bedford, MA) by Western blotting, and generated blots were incubated with the Gag- and Env-specific antisera and the corresponding preimmune sera. The membranes were washed twice with blocking solution, and HERV-K-specific proteins were detected with an alkaline-phosphatase-conjugated second step antibody.

Aliquots of iodixanol density gradient-purified particle suspensions from 518A2 supernatants were loaded on Formvar-coated grids and left there for 15min. Excess fluid was removed from the edges of the grids by filter paper. The grid with the remaining sample was air dried for 1 h. The sample was either directly exposed to 1% uranyl acetate for negative staining or, for immunoelectron microscopy, fixed in paraformaldehyde-lysine-periodate (22) for 15 min, rinsed in destilled water, and quenched in PBS/1% BSA. After fixation in 2.5% glutaraldehyde, washing and negative staining with 1% uranyl acetate, the grid was left to air dry. All grids were examined with a JEOL 1010 electron microscope.

Touch preparations of nevi, primary melanomas, and melanoma metastases surgically removed from patients were made by dipping freshly excised tissue on coated slides (Dako, Biotek Solutions). Slides were fixed in 4% paraformaldehyde for 20 min, washed in PBS, dehydrated through graded alcohols to absolute ethanol, and were air dried.

The hybridization mixtures consisted of Hybrisol VI (Oncor, Gaithersburg, MD) and digoxigenin-labeled cDNA probes (final concentration of 2 ng/μl). Slides were covered with glass coverslips and sealed with Gelbond (ICN). Probe and cellular material were denatured by heating to 80°C for 5 min. Hybridization was carried out at 37°C overnight in a humid chamber. After removal of the coverslips, the slides were washed at 46°C three times with 50% formamide/2× SSC, once with 2× SSC for 10 min, followed by a single wash step in 2× SSC containing 0.1% NP40 for 10 min. Signal detection was performed after a blocking step in 1% blocking reagent (Boehringer Mannheim) 30 min at 37°C by incubation with anti-digoxigenin antibody, conjugated to rhodamine at a dilution of 1:10 in 1% blocking reagent for 30 min at 37°C in a humidified box. After extensive washings in PBS, slides were counterstained with 10 μg/ml 4′,6-diamidino-2-phenylindole for 20 min and visualized with a Zeiss fluorescence microscope using a triple bandpass filter and software from PSI.

Cells grown on chamber slides, as well as touch preparations from clinical samples of touch preparations from patients with melanomas or benign melanocytic lesions on coated slides (Dako), were fixed in 4% paraformaldehyde for 20 min, washed in PBS, and dehydrated through graded alcohols to absolute ethanol and air dried. Immunofluorescence staining was performed by incubating chamber slides, with HERV-K-specific antibodies at a dilution of 1:100 for 1 h at 37°C in a humidified box, followed by washing three times with PBS and subsequent incubation with Alexafluor-488-conjugated antibodies at a dilution of 1:200 for 1 h at 37°C. Counterstaining was performed by mounting in Vectashield containing 4′,6-diamidino-2-phenylindole (Vector). Preparations were analyzed by using a Zeiss fluorescence microscope with appropriate filters.

For detecting particles released by the cells and subsequent sequencing of viral RNA packaged by the particles, supernatants from ∼10⁷ 518A2 cells were filtered and pelleted 2 h at 28 K. For additional purification, the pellet was resuspended, loaded on iodixanol density gradients, and the fraction corresponding to a density of ∼1.16 g/ml was harvested. This fraction was resuspended in PBS, pelleted again 2 h at 28 K, and resuspended in 500 μl of Trizol. To this suspension, 100 μl of chloroform were added. After vortexing and centrifugation the, upper phase was precipitated with isopropanol. The precipitate was pelleted by centrifugation and washed with 70% ethanol. The pellet was resuspended, treated with DNase in the presence of 25 mm MgCl₂ for 45 min at 37°C, and the DNase was inactivated for 10 min at 65°C. After precipitation with 96% ethanol, the pellet was washed with 70% ethanol, resuspended in RT reaction buffer, and RT was performed with the random primer p(dN)6. One of 10 of the RT reaction was used as template for PCR with the oligonucleotide primers 5′-108 propol 3925 5′-CCACTGTAGAGCCTCCTAAACCC-3′ and 3′-108 pol 4315 5′-GCTGGTATAGTAAAGGCAAATTTTTC-3′ (Codon Genetic Systems, Vienna, Austria), which corresponds to conserved regions within the pol gene. Ten μl of each amplification product were analyzed by electrophoresis in a 1.5% agarose gel and visualized by ethidium bromide staining. An aliquot of the amplification product was cloned and sequenced in both directions. For PCR, a polymerase with proofreading activity was used (Pwo; Roche).

HERV-specific DNA present in MDBK cells was detected by PCR of genomic DNA from 10⁷ MDBK cells using the primers 5′-108propol3925 and 3′-108pol4315. The control-reaction was performed with the β-actin-specific primers 5′-TCACCCACACTGTGCCCATCTACG-3′ and 5′-CGTCATACTCCTGCTTGCTGATCC-3′.

Cell-free supernatants of ∼10⁸ 518A2 melanoma cells were loaded on 20% sucrose cushions and pelleted at 28 K in a Beckmann SW28 rotor. The resulting pellets were washed and resuspended in 500 μl of PBS. Bovine MDBK cells were grown to ∼30% confluency and exposed overnight to 100-μl aliquots of the particle preparations in the presence of 0.8% Polybrene. As controls, cells were treated with particle preparations that were heat inactivated. Twenty-four h after infection, the inoculum was removed, and the cells were washed twice with PBS and incubated with normal media. Starting at 7 days after infection, the cells were continuously passaged at a ratio of 1:10. Presence of HERV-DNA in MDBK cells exposed to melanoma-derived particles was analyzed by PCR of genomic DNA with pol-specific primers. The amplification products were cloned and sequenced.

To detect extracellular virions, we first analyzed supernatants from the melanoma cell lines SKMel-28, SKMel-1, 518A2, MelJuso, as well as primary melanoma cells (HV-Mel7 derived from a primary melanoma and HS-Mel2 and JH-Mel6 derived from lymph node melanoma metastases) for pelletable RT. Supernatants from cultured NHEMs served as controls. Cell-free supernatants from ∼10⁶ cells were concentrated by centrifugation. The reverse transcription activity of the resulting pellets was analyzed by fluorescent probe-based product-enhanced reverse transcriptase assay. Supernatants of all melanoma cell lines tested were found to contain a fluctuating yet continuously detectable RT activity, corresponding to the activity of up to 10,000 units of Moloney murine leukemia virus reverse transcriptase (Superscript II; Life Technologies, Inc.)/ml supernatant. On the other hand, supernatants from cultured melanocytes NHEM neo 5935, NHEM neo 4528, and NHEM neo 6083 did not contain detectable RT activity (detection limit 10⁻¹ microunits; data not shown). The fact that supernatants derived from melanoma cells but not from melanocytes contain pelletable RT activity suggests that melanoma cells contain proviral sequences with sufficient genetic information to form particles containing a functional RT. Because no RT activity was detected in the supernatants of melanocytes, production of particles containing RT appears to be a property of melanoma cells.

Particles derived from SK-Mel28 melanoma cells were purified on iodixanol cushions and analyzed for the presence of HERV-K-specific Env and Gag proteins in Western blots. The envelope (env) gene of retroviruses displays an open reading frame for the surface protein (SU) and a membrane-spanning protein (TM). The Env precursor is usually cleaved into the SU and TM subunits before translocation to the cell surface and incorporation into virus particles. To determine whether the Env protein is present on the particles, we performed immunoblotting with an antiserum recognizing the TM domain. As shown in Fig. 1,A, two bands are visible in the Western blot (Fig. 1 A, Lane 1). The upper band corresponds to the precursor migrating at M_r ∼80,000–90,000. In addition, a lower band migrating at M_r ∼37,000 is visible, suggesting cleavage of the precursor into subunits. However, because the anti-Env antiserum recognizes the TM subunit, we do not know whether the SU domain is present after cleavage from the TM domain.

Immunoblotting with a HERV-K anti-Gag antiserum revealed a double band at M_r ∼76,000, corresponding to Gag precursors, as well as processed intermediate Gag proteins at M_r ∼61,000, M_r 30,000, and one band migrating with the front (less than M_r 19,000; Fig. 1 A, Lane 3). The M_r 30,000 protein corresponds to the putative major core protein of HERV-K. The presence of processed Gag proteins in the Western blots indicates a functional protease. However, the particles still contain a substantial proportion of uncleaved or partially cleaved intermediates.

To confirm the presence of physical particles, we performed electron microscopy. Pellets derived from supernatants of the melanoma cell line 518A2 were purified on iodixanol density gradients (Fig. 1,B). Electron microscopy revealed the presence of retrovirus-like particles characterized by membrane bound spherical structures with diameters ranging between 80 and 120 nm (Fig. 1 B).

The integration sites of human endogenous retroviral elements have been found to be distributed over the whole human genome. For example, the HERV-K family was reported to be present in ∼30 copies/human haploid genome (23). On the basis of the observation that melanoma cells (but not melanocytes) produce retrovirus-like particles, we hypothesized that the formation of virus-like particles might be because of the activation of retroviral genes that are usually repressed. We therefore used the cloned pol sequence derived from melanoma cell-derived particle preparations as a probe and looked for expression of this sequence in the cytoplasm of melanoma cells by in situ hybridization.

To determine the specificity of the pol sequence for melanoma, we analyzed touch preparations of primary melanomas, lymph node metastases, and cutaneous metastases that had been surgically removed from melanoma patients. As shown in Fig. 2, high copy numbers of the pol sequence were found in tumor cells of all melanoma preparations tested. In comparison, lymph node and benign nevus cells of healthy individuals were negative. A probe specific for the nucleoprotein gene of influenza virus tested on touch preparations at the same time was always negative. As a positive control, we used a probe recognizing the melanoma-inhibiting activity gene (24). The rate of melanoma cells expressing the pol gene was in the range of 60–90%, a percentage similar to the one obtained with the melanoma-inhibiting activity-specific probe (Fig. 2). The observation that the sequences are not found in all tumor cells might be because of the sensitivity of the assay and variations in the expression levels. Alternatively, this result might suggest that the pol gene is not expressed in all malignant cells. It could be argued that it was absent from the neoplastic clone at the time of malignant transformation, although selective loss of expression remains another possibility.

To determine whether HERV-specific proteins were detectable in melanoma cells, we performed immunofluorescence analysis with antisera recognizing the Gag, Rec, and Env proteins of HERV-K. Fig. 3,A shows that the melanoma cell lines Mel-Juso and SK-Mel28 express the Gag, Rec, and Env proteins. The percentage of Gag-expressing cells was in the range of 1–10%, Rec was found to be present in ∼20%, whereas expression of the Env protein was detected in ∼10% of the analyzed cells. Expression of Gag and Env was found in the cytoplasm, whereas Rec was mainly found in the nucleus. In contrast, cultured human melanocytes (Nhem) and Vero cells did not react with any HERV-K-specific antisera (Fig. 3 A). None of the corresponding preimmune sera was reactive with any of the cells tested.

To assess whether HERV-K protein expression could be associated with malignancy, we performed an immunofluorescence study on clinical samples of touch preparations from patients with melanomas or benign melanocytic lesions. Primary melanomas, lymph node metastases, cutaneous metastases, benign lymph nodes, and naevi were diagnosed by H&E staining, coded and characterized by immunofluorescence analysis.

A total of 50 samples was tested. Nine samples were primary melanomas, 3 were cutaneous metastases, and 9 were lymph node metastases. In addition, 4 benign sentinel lymph nodes and 25 benign melanocytic lesions were analyzed. Our results show that expression of Gag, Rec, and Env proteins was highly specific for tumor cells, and all patients with primary melanomas, lymph node, or cutaneous metastases were positive for HERV-K protein expression. In contrast, only one of the benign lesions showed HERV-K protein expression (Table 1). A representative experiment is shown in Fig. 3 B.

In an attempt to analyze the particle preparations on a molecular level, we amplified a sequence corresponding to a sequence of to the pol gene. Iodixanol gradient-derived particle preparations of the melanoma cell line 518A2 were used for RT-PCR analysis. Particles were treated with DNase to remove contaminating traces of genomic DNA. RT-PCR of these preparations revealed the anticipated amplification products of 390 nucleotides (Fig. 4,A, Lane 1). Because no amplification products were obtained when RT was omitted, possible DNA contaminations can be excluded (Fig. 4,A, Lane 2). In addition, RT-PCR of the glyceraldehyde-3-phosphate dehydrogenase gene was negative, reflecting the lack of contamination with cellular mRNA (data not shown). The amplification product was cloned into an expression vector, sequenced, and compared with the corresponding sequence of HERV-K 108 and HERV-K113. Sequence analysis of 9 clones revealed that the particles package heterologous sequences with high homology to the corresponding region of the HERV-K clone 108 (25). The homology to HERV-K108 was in the range from 95.5 to 97% (Fig. 5). The sequences showed also high homology (93–95%) to the corresponding region of HERV-K113 (26). These data suggest that the melanoma-associated retrovirus exists as a quasispecies, defined as a group of closely related but genetically distinct sequences.

We then attempted to obtain the whole proviral sequence of the pol sequence of the melanoma 518A2-derived particles by sequencing genomic DNA of the bovine MDBK cells that were exposed to 518A2 particles. Analysis of genomic DNA from 518A2 melanoma cells with pol-specific primers by PCR reveals the anticipated amplification product of 390 bp (Fig. 4,B, Lane 4). In contrast, nonexposed MDBK cells did not reveal a signal, indicating that bovine MDBK cells are free of sequences that are homologous to HERV-K (Fig. 4,B, Lane 3). We therefore exposed the bovine cell line MDBK to 518A2-derived particles and subsequently PCR-amplified genomic MDBK DNA with primers covering the whole pol gene. Sequence analysis of the pol open reading frame of genomic DNA from the particle-exposed MDBK cells indicates that the melanoma-derived sequence (MERV; accession no. AY186778) is closely related but not identical to the HERV-K clone 108 (accession no. AF164614), which showed the highest homology. A total of 13 nucleotide differences was found. Five of these differences result in amino acid changes (Table 2). The sequence is also highly homologous (20 nucleotides difference) to the pol gene of a full-length, intact HERV-K element (HERV-K113), which is polymorphic in humans (26). It should be noted that there is no exact match to this sequence anywhere in the human genome, suggesting that this MERV element is not fixed in the population. This result indicates that this sequence is not part of the genome analyzed in the genome project. One possible explanation would be that because the sequence appears to be transcribed and translated in melanoma cells and is packaged into particles, it may have evolved during retrotransposition or infection.

We show that retroviral sequences and particles are expressed in melanoma cells. Although this does not establish pathogenicity, the mere presence of retroviral sequences and particles in melanoma raises the question whether they are activated as a by-product of malignant transformation or play an active role in melanoma formation or progression.

Partial sequence analysis, immunoblotting, and immunoelectron microscopy studies suggest that the particles belong to the HERV-K family. In contrast to other known HERV-K-like viruses, which lack infectivity, the Env precursor appears to be cleaved in the melanoma-derived particles, resulting in approximately equimolar amounts of the precursor and the putative TM. It has been shown that murine leukemia virus Env mutants containing a similar ratio of the precursor and TM are almost as infectious as the wild-type virus containing a completely cleaved Env. Mutants containing uncleaved precursors only were noninfectious (27).

The observation that melanoma-derived HERV-specific sequences were detected in particle-exposed MDBK cells suggest that the particles produced by the melanoma cells are infectious for MDBK cells, although further evidence is required. However, up to now we have failed to infect cultured melanocytes. This could be because simple retroviruses only infect dividing cells, and cultured melanocytes show very limited proliferative capacity. We are currently investigating whether melanocytes transformed by UV radiation are susceptible for infection. However, even in the case that melanocytes are nonpermissive for the melanoma-derived viruses, they may have retained important features of replicating retroviruses. De novo insertion by retrotransposition has the same pathogenic impact on cellular genes as insertion by infection. Both can activate expression of cellular genes, may disrupt open reading frames, and have the ability to be retrotransposed either by a retroviral RT or by the RT encoded by a nonlong terminal repeat retrotransposon family. De novo insertion into exons usually interrupts open reading frames and may result in the loss of gene function (28, 29). Destruction of tumor suppressor genes by insertional mutagenesis may also contribute to the multistep process required for carcinogenesis. For example, a retrotransposed L1 element was detected in a c-myc allele in a breast adenocarcinoma (30). Disruption of the APC gene, which is considered to be a tumor suppressor gene, is caused by somatic insertion of a LINE-1 sequence (31). Because long terminal repeats have transcriptional regulatory signals, reintegration of long terminal repeat-containing retroviral elements has great potential to affect adjacent genes. For example, high activity of the long terminal repeat promoter might be the cause for transcriptional fusion between two adjacent genes in teratocarcinoma cells (32).

However, although reintegration of LINE-1 elements has been reported to be involved in transformation to malignant cells, we are not aware of any such event being demonstrated for HERVs. It will be interesting to determine whether HERV-mediated reintegration events in melanoma cells may have led to inactivation of tumor suppressors or to activation of proto-oncogenes.

We thank Heidrun Karlic for suggestions, Sabine Brandt, Oliver Scheiber, Edward Fiedler, and Sabine Kallenda for technical assistance, Markus Dawid for providing clinical material, and Joachim Denner (Robert Koch Institut, Berlin, Germany) for providing HERV-K Env-specific antisera.