Findings

Although no infectious HERV particles have been detected to date, several proviruses of the evolutionary young hominoid group HERV-K(HML-2) harbor open reading frames for viral proteins, as well as promoter competent long terminal repeats (LTRs) [1, 2]. Two subgroups exist (Figure 1A): Type 2 proviruses encode the accessory protein Rec, which mediates nuclear-cytoplasmic translocation of incompletely spliced RNA, thereby allowing translation of viral proteins [3]. Type 1 proviruses harbor a 292-bp deletion within the pol-env boundary, resulting in loss of a functional open reading frame for Rec, but gaining a functional open reading frame for the accessory protein Np9 [4]. In vitro assays have shown that Rec and Np9 interact with cellular proteins like the promyelocytic leukemia zinc finger protein (PLZF), ligand of numb protein X (LNX), testicular zinc-finger protein, androgen receptor, and small glutamine-rich tetratricopeptide repeat protein and might thus support cell transformation [5–9]. Although HERV-K(HML-2) proviruses may be transcribed in somatic cells [10, 11], up-regulated transcription and protein expression, formation of retrovirus particles, and induction of an anti-HERV-K(HML-2) immune response are predominantly associated with germ cell tumors, embryonal carcinoma cell (ECC) lines, melanomas and other cancers [12–17].

Figure 1
figure 1

HERV-K(HML-2) expressed at RNA and protein level in ES/iPS cells. A) Prototypic HERV-K(HML-2) open reading frames B) Type 1 and 2 mRNA species, ∆ denotes the 292 bp deletion in type 1 proviruses. 1) full length type 2, 1∆) full length type 1, 2) env mRNA type 2, 2∆) env mRNA type 1, 3) 1.8 kb mRNA type 2 encoding rec[18], 3∆) 1.5 kb mRNA type 1 encoding Np9[4], 4) 1.5 kb mRNA derived from type 1 or type 2 [18]. PCR primers are indicated by arrows and arrowheads. Primers PrevENV, PrevREC, PrevNP9 overlap the immediate upstream splice site. C) RT-PCR of HERV-K(HML-2) expression in ES cells (H1, H9, BG01), fibroblasts (MRC5, MSC, hFib2) and corresponding iPS cells using the primer pair K20 and K6. OCT4, NANOG served as pluripotency, LMNA as differentiation markers, ACTB as internal control [19, 20]. D) Western blot for HERV-K(HML-2) Gag, OCT4 and ACTB in ES, iPS, fibroblast and EC cells.

Full size image

The phenotype and transcriptional profile of ECCs resemble that of non-malignant pluripotent embryonic stem cells (ESCs) and induced pluripotent stem cells (iPSCs) in that they express key pluripotency factors such as OCT4, NANOG and SOX2 and other markers characteristic of pluripotent stem cells [reviewed in [21]]. In addition, ECCs rely on similar pathways to regulate self-renewal and inhibition of differentiation [21], such as autocrine FGF signalling and activation of downstream signalling cascades, especially the ERK/MEK pathway [22, 23]. These similarities prompted us to test whether HERV-K(HML-2) RNA and protein expression, which is a marker of ECCs, might also be reactivated in non-malignant pluripotent stem cells.

HERV-K(HML-2)RNA and proteins are expressed in ESCs and iPSCs

We analyzed the expression of HERV-K(HML-2) in three ESC lines (H1, H9, BG01), three fibroblast lines (MRC5, MSC, hFib2) and iPSC lines derived from these fibroblasts. All materials and methods are described in detail in Additional file 1 “Materials and Methods”. To augment the probability to detect HERV-K(HML-2) transcripts derived from proviral promoters, we analyzed completely spliced viral RNA by RT-PCR. Full length and spliced transcripts and the location of all primer pairs used are depicted in Figure 1B. We verified expression of the endogenous pluripotency genes OCT4 and NANOG and absence of the differentiation marker gene LMNA in pluripotent stem cells, ACTB specific amplicons served as internal controls [19, 20]. We did not detect any HERV-K(HML-2) specific amplicons in any of the fibroblast cell lines, which readily expressed the differentiation marker LMNA. In contrast, all ESC and iPSC lines tested showed expression of all three types of completely spliced HERV-K(HML-2) transcripts, in addition to the pluripotency markers OCT4 and NANOG (Figure 1C). The intensities of the PCR amplicons varied between individual ESC/iPSC lines, suggesting differential activation of proviruses between lines.

The presence of spliced HERV-K(HML-2) type 2 transcripts in all pluripotent stem cell lines tested prompted us to investigate whether they encode functional Rec protein and thus enable translation of HERV-K(HML-2) Gag and Env proteins. Using a HERV-K(HML-2) Gag specific monoclonal antibody in Western blot analysis, we indeed detected low to moderate levels of this protein in the H9 ESCs and the hFib2-iPS5 iPSC line, but not in the parental fibroblasts hFib2. By contrast, the ECC line GH showed high levels of Gag protein expression (Figure 1D). To ensure the phenotypic similarity between the malignant ECC line and the non-malignant ESC/iPSC lines, we verified by RT-PCR that GH cells expressed the pluripotency markers OCT4, NANOG, SOX2, STELLA as well as the autocrine signaling factor FGF4 (data not shown). The differences in HERV-K(HML-2) Gag protein levels observed in H9 and hFib2-iPS5 correlated with the differences in signal intensities of their rec specific amplicons in RT-PCRs (Figure 1C, compare intensities of amplicons 3), which is consistent with the idea that the levels of functional Rec protein determine the efficiency of Gag protein translation.

Differentiation induces silencing of HERV-K(HML-2)proviruses

To determine whether HERV-K(HML-2) expression in pluripotent stem cells changes during differentiation, we induced embryoid body formation of hFib2-iPS5 cells. We applied semi-qRT-PCR analysis using the primer pair K7/K21 (see Figure 1B). Interestingly, the levels of HERV-K(HML-2) RNA started to decrease from day 6 onwards in hFib2-iPS5 cells, concomitant with a decrease of NANOG and an increase of LMNA expression (Figure 2A). To estimate the degree of the changes we quantified the signal intensities of the RT-PCR bands visualized by ethidium bromide gel analysis (Figure 2A) relative to ACTB (see Additional file 1 “Materials and Methods”). HERV-K(HML-2) expression levels dropped from 100% on day 0 to roughly 50% on day 6, and to 20% on day 10. Similarly, NANOG expression levels decreased from 100% on day 0 to roughly 75% on day 6, and 55% on day 10.

Figure 2
figure 2

HERV-K(HML-2) expression is immediately down-regulated during embryoid body differentiation. A) Semi-quantitative RT-PCR of HERV-K(HML-2) expression in hFib2- iPS5 cells using the primer pair K21 and K7. The faint band at ≈570 bp presumably represents the rec mRNA of provirus ERVK-17 (see Table 1) which contains a small deletion in the amplified region. LMNA was amplified using an alternative primer (see Additional file 1 “Materials and Methods”) B) Quantitative expression analysis of HERV-K(HML-2) env, rec and Np9 transcripts in H9 cells. OCT4, LMNA, and ACTB serve as control [19, 20]. RNA levels are depicted relative to undifferentiated cells on day 0 (n=2, error bars +/-s.e.m).

Full size image

To verify this with a second method and cell line, we performed real-time RT-PCR analysis using differentiating H9 ESCs [19, 20]. This analysis showed an even more dramatic reduction of HERV-K(HML-2) RNA expression during differentiation with significantly reduced transcription already detectable on day 2 of embryoid body differentiation (Figure 2B). Remarkably, the kinetics of HERV-K(HML-2) downregulation is even faster than repression of OCT4 and NANOG.

These slightly divergent results may simply reflect the greater accuracy of real time PCR compared to semi-quantitive PCR or may be rooted in the well-known subtle differences in the expression profiles of pluripotent stem cells especially those of ESC compared to iPSC lines [24].

The recent findings that high RNA levels of the endogenous retrovirus group HERV-H [25] and expression of non-LTR retrotransposons [26] are also associated with human pluripotent stem cells indicate that a pluripotent phenotype might be accompanied by a general relieve of retroelement silencing. We therefore asked whether re-activation of HERV-K(HML-2) proviruses is specific for certain proviral elements or occurs in a stochastic manner.

HERV-K(HML-2)proviruses are activated in ESC/iPSC/ECC lines in varying patterns

We analyzed activation of HERV-K(HML-2) proviruses in malignant and non-malignant pluripotent stem cells using an established protocol [27] to amplify, clone, and sequence a highly conserved part of the gag gene (see Additional file 1 “Materials and Methods”). A proviral sequence was scored as the genomic origin of the respective clone when the two sequences exerted more than 98% identity (Additional file 2 “Alignments”, .msf files of these alignments are available as Additional files 3 “K101(22q11.21)”, 4 “K10(5q33.3)”, 5 “K106(3q13.2)”, 6 “K115(8p23.1)”, 7 “K37(11q23.3)”, 8 “K102(1q22)”, 9 “K108(7p22.1)”, 10 “HKc10-B(10q24.2)”, 11 “K41(12q14.1)”, 12 “K109(6q14.1)” and 13 “KI(3q21.2)”). Out of the 66 full-length or nearly full-length HERV-K(HML-2) proviruses present in the human genome [1], a subset of 11 were found to be expressed in patterns, which varied from cell line to cell line (Table 1). More activated loci may be detected using optimized primers [28], deep sequencing techniques [29] as well as more cell lines. Notably, all transcribed loci belong to the LTR5Hs subgroup of HERV-K(HML-2) proviruses. Interestingly, phylogenetic classifications have shown that all human-specific retroviral integrations fall into this proviral subgroup [1]. Indeed, ten of the expressed proviruses were human-specific, while K37(11q23.3) may also be present in the genomes of great apes. Five of the re-activated proviruses belonged to type 1 and six to type 2 genotypes, respectively, but their number and expression levels varied between individual ESC/iPSC and ECC cell lines. Such variations in HERV-K(HML-2) proviral transcription profiles are commonly observed in somatic malignant and non-malignant tissues and cell lines [10, 28] indicating a somewhat stochastic re-activation mechanism which may reflect varying chromatin dynamics [30]. By contrast, we observed constant and efficient transcription of the type 1 provirus K101(22q11.21) in all pluripotent stem cell lines tested, suggesting a more directed re-activation. K101(22q11.21), K37(11q23.3) and K10(5q33.3) encode gag and protease genes and may account for the protein expression detected, e.g. in the H9 and hFib2-iPS5 lines (Figure 1D). Transcription of one or the other type 2 locus was detected in the ESC/iPSC lines, but only at low level. This is in contrast to the observed constant up-regulation of K108(7p22.1) expression exclusively in the malignant ECC lines. Notably, a recent survey of HERV-K(HML-2) proviruses transcribed in melanomas in comparison to melanocytes also demonstrated K108(7p22.1) derived mRNA only in melanomas but not in the non-malignant precursor cells [28].

Table 1 Relative cloning frequencies of HERV-K(HML-2) loci transcribed in ESC, iPSC and ECC lines
Full size table

Conclusion

The endogenous betaretrovirus group HERV-K(HML-2) is unique in its potential to code for viral and accessory proteins. We observed, in slightly varying patterns, re-activation at the RNA and protein levels of certain human specific, protein encoding HERV-K(HML-2) proviruses with promoter competent LTR sequences in pluripotent ESCs, iPSCs, and ECCs. The surprisingly constant activation of K101(22q11.21) across all cell lines tested may indicate a particular function in or association with pluripotency. We detected K108(7p22.1) only in the malignant ECC lines. This corresponds to up-regulated K108(7p22.1) transcription described in other germ cell tumours, in brain tumours but not in normal brain [10] as well as in malignant melanoma [28] but not in melanocytes. HERV-K(HML-2) proviruses were rapidly silenced upon embryoid body differentiation. Re-activation of this HERV group thus represents another marker for the undifferentiated state of pluripotent stem cells.

Active type 2 proviruses produce Rec protein, which supports cell transformation in vitro and germ line carcinoma in situ in transgenic mice [7, 31]. Type 1 proviruses produce Np9 protein [4]. Both Rec and Np9 interact with transcriptional regulators. For example, Rec and Np9 bind the MYC repressor PLZF resulting in overexpression of MYC [6] and Np9 can bind to LNX [8] which might influence the activity of the NOTCH pathway. MYC and NOTCH are often involved in carcinogenesis but are also important players in the signalling networks controlling self-renewal, pluripotency and differentiation [32]. Since there are substantial differences between the cell cycle regulation of murine and human pluripotent stem cells [21], it is intriguing to speculate that Rec and Np9, which are not encoded by rodent ERVs, might play positive roles in human ESC/iPSCs. The observation that Rec and Np9 encoding HERV-K(HML-2) proviruses were preferentially activated in pluripotent stem cells and that their expression was simultaneously silenced upon differentiation independent of their chromosomal localization is in favour of such an idea.