Findings

Long interspersed elements type one (LINE-1; L1), the only active autonomous transposable element in the human genome, have played a major role in human genome evolution and are also responsible for an increasing number of sporadic human genetic diseases[1–3]. To make new copies, a source L1 element must successfully navigate through every stage of the retrotransposition process (that is, transcription, translation, and target-primed reverse transcription). An essential intermediate step is the formation of an L1 ribonucleoprotein (RNP) complex between L1 mRNA and proteins[4–6]. L1 RNP must enter the nucleus before a new copy is made via target-primed reverse transcription[7]. Therefore, the nuclear import of L1 RNP presents an opportunity for cells to regulate L1 retrotransposition post-translationally. As nuclear import can occur passively when nuclear envelope breaks down during cell division, the efficiency of retrotransposition is predicted to be higher in actively dividing cells. Indeed, the effect of cell division has been investigated by two previous studies, which compared the level of L1 retrotransposition in cell-cycle arrested primary cell strains and cancer cell lines[8, 9]. Although both observed varied degrees of inhibition in retrotransposition when cells were experimentally arrested in different stages of the cell cycle, one study concluded that cell division was required for retrotransposition and the other determined that L1 retrotransposition could occur in non-dividing cells (Table 1). Hence, the role of cell division on retrotransposition remains highly debated to date.

Table 1 A comparison of the methods and findings from three studies
Full size table

Development of a stable HeLa Tet-ORFeus cell line

To investigate the effect of cell-cycle arrest on L1 retrotransposition, we wished to establish an assay system that meets the following criteria: (1) It must be a stable cell line with an integrated L1 reporter. Having an integrated L1 reporter eliminates variation in transfection efficiency that is inherent in transient assays. However, this requirement necessitates the use of an inducible promoter because, otherwise, L1 insertions will accumulate while the cell line is being established. (2) The promoter activity (that is, transcription) can be conveniently monitored in parallel to L1 retrotransposition. (3) Both the promoter activity and L1 retrotransposition can be measured with high sensitivity and in a wide dynamic range. Accordingly, we designed an inducible dual-luciferase L1 assay vector, pYX056 (Figure 1A; detailed in Additional file1). The design combined a gene regulation and a gene delivery system. First, the Tet-Off Advanced Inducible Gene Expression System allows stringent control of L1 expression. The bi-directional PTight inducible promoter drives expression of Renilla luciferase (Rluc) and a hyperactive synthetic mouse L1, ORFeus[10]. The latter is tagged with a firefly luciferase/antisense intron (FlucAI) reporter cassette[11]. The bi-directional PTight promoter consists of a modified tetracycline-responsive element flanked by two minimal CMV promoters. In the presence of doxycycline, the tetracycline-controlled transactivator advanced (tTA) is complexed with doxycycline and is unable to activate PTight. Upon doxycycline withdrawal, free tTA will bind to PTight and activate both L1 and Rluc transcription. Second, the non-viral two-component Sleeping Beauty (SB) system enables stable gene delivery. The L1/Rluc bi-directional expression cassette is flanked by a pair of inverted terminal repeats from SB, and can be ‘cut and pasted’ into the genome by a hyperactive SB transposase (SB100X)[12] (Figure 1B). Single cell clones were acquired by limiting dilution method and screened for the lack of Rluc expression in doxycycline-supplemented medium but high levels of Rluc expression upon doxycycline withdrawal (Figure 1C; detailed in Additional file1).

Figure 1
figure 1

L1 retrotransposition in a HeLa Tet-ORFeus stable cell line. (A) A schematic of the bi-directional inducible L1 construct. The bi-directional tet-responsive promoter PTight drives the expression of an upstream Rluc cassette and a downstream L1 cassette. The L1 cassette features coding sequences (that is, ORF1 and ORF2) from the synthetic L1 ORFeus and an antisense-stranded FlucAI reporter cassette[11]. In the presence of doxycycline, PTight is inactive. FlucAI can be transcribed from its own SV40 promoter. However, no Fluc activity is expected because Fluc coding sequence is interrupted by an antisense intron (sense relative to the L1 cassette). (B) Incorporation of the L1 construct into HeLa-tTA cells. The L1 construct is terminally flanked by ITRs of the Sleeping Beauty DNA transposon (see panel A). To make a stable cell line, the L1 construct was co-transfected with SB100X into HeLa-tTA cells. Single cell clones were established through limiting dilution in the presence of doxycycline. Rluc and Fluc were measured after doxycycline withdrawal. (C) The rationale of L1 retrotransposition assay with Tet-ORFeus cells. In the absence of doxycycline, PTight is bound by tTA and activates the transcription of a Rluc mRNA and an L1 pre-mRNA. The intron is removed from L1 pre-mRNA through splicing. The mature L1 mRNA is reverse transcribed and integrated into the genome (shown as a 5′ truncated insertion), forming a functional Fluc cassette.

Full size image

Control of L1 retrotransposition in HeLa Tet-ORFeus cells by doxycycline

To characterize L1 retrotransposition in HeLa Tet-ORFeus cells, we first tested the dose response by seeding cells in different concentrations of doxycycline. Both Fluc and Rluc signals were doxycycline dose-dependent (Additional file2). Significant Fluc signals were first observed after 30 h incubation in doxycycline-free medium and subsequently increased exponentially to 460-fold above background after 48 h incubation (P <0.01; Figure 2A). The rapid induction of the PTight promoter via doxycycline withdrawal was demonstrated by continued increase of Rluc signals from three-fold (at 6 h) to 280-fold (at 48 h) above background (P <0.01; Figure 2A). To directly measure L1 expression, we quantified L1 ORF1 protein (ORF1p) by western blot (Figure 2B). L1 ORF1p signals were first observed at 9 h, peaked at 24 h, and subsequently maintained for the duration of the experiment (Figure 2B). Induction of Rluc or ORF1p was not observed in HeLa Tet-ORFeus cells cultured in 100 ng/mL of doxycycline, indicating PTight was completely suppressed. Indeed, cells maintained under 100 ng/mL of doxycycline showed no accumulation of Fluc-positive cells over 10 passages but could be robustly induced upon doxycycline withdrawal (Additional file3). To confirm that Fluc signals were due to retrotransposition, we monitored intron removal by genomic DNA PCR as previously described[11]. Consistent with Fluc measurement, the intronless amplicon became most prominent at 30–48 h although weak amplicons could be observed in earlier time points. As a control, no intronless band was seen in HeLa Tet-ORFeus cells under 100 ng/mL of doxycycline (Figure 2C). Similar to the transient dual-luciferase assays[11], retrotransposition was inhibited by a nucleoside reverse transcriptase inhibitor in a dose-dependent manner (Additional file4). The number of L1 insertions was further quantified by a quantitative PCR (qPCR) method as previously described[13]. Similar to transient transfection experiments[11], statistically significant signals were first detected at 24 h (normalized activity = 5.6%, P <0.01) (Figure 2D). It should be noted that, as opposed to antibiotic or fluorescent protein reporters, which can be used to track individual retrotransposition events, the HeLa Tet-ORFeus system measures retrotransposition from a population of cells.

Figure 2
figure 2

The time course of L1 retrotransposition in HeLa Tet-ORFeus cells. (A) Fluc and Rluc activities from cell lysates. Cells were seeded in 96-well plates (for luminescence) or 60 mm dishes (for protein and gDNA analyses) in the absence of doxycycline and harvested at the indicated time points. Error bars represent mean±SE (n = 6). All readings were compared with the 0 h control (**P <0.01). (B) Time-dependent increase of ORF1p expression. ORF1p and β-actin were detected by western blot. Murine embryonal carcinoma cells (F9) were used as a positive control for ORF1p. The parental HeLa-tTA cells and uninduced HeLa Tet-ORFeus cells were used as negative controls. (C) Confirmation of L1 retrotransposition by end-point PCR. Genomic DNA was amplified by an intron-flanking primer pair. The presence of a band of 250 bp is diagnostic for intron removal; the intron-containing donor DNA is amplified as a band of 1150 bp. NTC, no template control. Dox+, gDNA from cells cultured in the presence of doxycycline for 48 h. Fluc plasmids with or without the intron were used as controls. Molecular weight was indicated by the 1 kb Plus DNA Ladder (Invitrogen). (D) Quantification of L1 insertions by qPCR. The number of L1 insertions in gDNA was determined by a TaqMan-based qPCR assay. qPCR signals were normalized by setting signals from the 48 h time point to 1. The normalized signals from each time point were then compared with the 0 h time point by two-tailed Student’s t-test. P values are indicated (**P <0.01). Error bars represent mean±SE (n = 3).

Full size image

Cell-cycle arrest inhibits L1 retrotransposition

To test the effect of cell-cycle arrest in HeLa Tet-ORFeus cells, cells were treated with three different inhibitors in the absence of doxycycline. Cell-cycle analysis showed that cells were arrested in S phase by aphidicolin and hydroxyurea and in S+G2/M phase by thymidine (Figure 3A). The presence or absence of doxycycline had no effect on cell-cycle status (compare Dox+ with Dox- in Figure 3A). As compared with control cycling cells (that is, Dox-), arrested cells showed 7.6% to 9.4% Rluc expression, indicating PTight was suppressed in the arrested cells (Figure 3B; P <0.001, detailed in Additional file5: panel A). Indeed, western blot analyses confirmed that, as compared with the Dox- group, the level of ORF1p was reduced to 6% to 15% in the arrested cells (Figure 3C). If the frequency of retrotransposition was a simple function of L1 expression, we would expect a proportional reduction of retrotransposition in arrested cells (that is, approximately 10% of cycling cells). However, the Fluc signal in arrested cells was at most 0.8% of the Dox- group (Figure 3B; P <0.001, detailed in Additional file5: panel B), indicating the presence of an additional approximate 10-fold reduction in retrotransposition that cannot be explained by the decrease in L1 expression. A potential caveat for these results is that the level of retrotransposition was indirectly measured by the expression of Fluc from integrated L1 insertions; this may cause an ascertainment bias between control and treatment groups if the inhibitors affect Fluc expression. Thus, we directly quantified retrotransposition by qPCR, a method that is independent of Fluc expression. Results from these qPCR experiments confirmed the additional reduction in retrotransposition in cell-cycle arrested cells (Figure 3D): in all three treatment groups, the magnitude of decease in qPCR signal was greater than the fold reduction in L1 expression, regardless whether L1 expression is measured as Rluc or ORF1p. On the other hand, two out of the three inhibitors displayed an inhibitory effect on Fluc expression when the Fluc data and qPCR data were compared (0.09% versus 0.4% for hydroxyurea treated cells and 0.8% versus 3.8% for thymidine treated cells, respectively; compare Figure 3B and3D). As a result, we compared the correlation between ORF1p and qPCR data. After adjusting the decrease in ORF1p, qPCR showed additional 8.7-, 27.5-, and four-fold reductions in retrotransposition in aphidicolin, hydroxyurea, and thymidine treated cells, respectively (compare Figure 3C and3D).

Figure 3
figure 3

Cell-cycle arrests inhibit L1 retrotransposition in HeLa Tet-ORFeus cells. (A) Cell-cycle analysis. HeLa Tet-ORFeus cells were cultured in doxycycline-free medium (Dox-) or supplemented with 5 μg/mL aphidicolin, 75 μg/mL hydroxyurea, or 2 mM thymidine. Cells cultured in 100 ng/mL doxycycline were used as control (Dox+). The distribution of cells in different phases of the cell cycle and their corresponding DNA content histograms are shown. (B) Normalized Fluc and Rluc activities. HeLa Tet-ORFeus cells were treated as in panel A for 48 h. Raw luminescence readings were normalized by cell viability first and then to those from Dox- cells (Flucmean = 425,000 and Rlucmean = 314,000). Error bars represent mean±SE (n=6). Statistical analyses are presented in Additional file5. (C) The effect of cell-cycle arrest on ORF1p expression. Representative western blots were shown for ORF1p and β-actin; quantitative data were calculated from three biological replicates and had been normalized by β-actin. F9 cells were used as a positive control for ORF1p. The parental HeLa-tTA cells and uninduced HeLa Tet-ORFeus cells (Dox+) were used as negative controls. (D) Quantification of L1 insertions by qPCR. The number of L1 insertions in gDNA was determined by a TaqMan-based qPCR assay. qPCR signals were normalized by setting signals from the Dox- cells to 1 (equivalent to 4.9 copies per cell as estimated from plasmid DNA dilution series). Error bars represent mean±SE (n=3).

Full size image

L1 retrotransposition in synchronized HeLa Tet-ORFeus cells

To exclude the possibility that the observed inhibition of L1 retrotransposition is caused by unknown side effects of inhibitors used, we wished to test the effect of cell division on L1 retrotransposition in cycling cells. To this end, we synchronized HeLa Tet-ORFeus cells by double-thymidine block and then allowed cells to enter normal cell cycling by removing thymidine from the culture medium (Figure 4A). According to cell-cycle analysis, these cells would complete two full cell cycles in 44 h (Figure 4A; Additional file6). We compared retrotransposition under two experimental conditions (b and d in Figure 4A). L1 transcription was induced for the same amount of time in both conditions (that is, 37 h). However, the withdrawal of doxycycline was timed so that L1 expression was activated at different cell-cycle phases. As a result, when L1 expression was induced, cells in experiment b would undergo G2/M phase once whereas cells in experiment d would undergo G2/M phase twice (Figure 4A). After 37 h induction, dual-luciferase readouts were taken from both conditions. Both showed similar levels of promoter activities (Figure 4B) but the level of retrotransposition was 2.6-fold higher in experiment d than in experiment b (P <0.05; Figure 4C). In control experiments, we demonstrated that the difference in assay duration did not alter the assay background (Figure 4A to C; conditions a and c, where cells were released from the double-thymidine block but remained in doxycycline-supplemented medium). Thus, these data from synchronized cells further support the conclusion that cell division promotes L1 retrotransposition, and thus is a potential means of regulating L1 activity.

Figure 4
figure 4

L1 retrotransposition in synchronized HeLa Tet-ORFeus cells. (A) Synchronized cells were released from double-thymidine block at time 0. Cells were incubated in the presence (solid line) or absence (dotted line) of doxycycline for varied time periods as indicated. (B) Rluc readouts. (C) Fluc readouts. Error bars represent mean±SE (n=6).

Full size image

In summary, our data provide additional support for an important role of cell division in L1 retrotransposition and argue that restricting the accessibility of L1 RNP to nuclear DNA could be a post-translational regulatory mechanism for retrotransposition (Table 1). As compared with the two previous studies[8, 9], our experimental approach has several advantages for assessing the role of cell division in retrotransposition. First, the dual-luciferase system provided an efficient means of simultaneous quantification of both L1 expression and L1 retrotransposition. Second, the use of an inducible, integrated reporter not only allowed us to avoid variation in gene transfer efficiency between experimental conditions but also to better resemble the replication cycle of endogenous L1 elements, which express from chromosomal rather than episomal DNA. Indeed, it allowed us to separate two layers of regulation in cell-cycle arrested cells: one layer is at the transcriptional level, which was highlighted by Shi et al.[9]; the other layer is downstream and independent of L1 transcription, as indicated by both Rluc signals and ORF1p levels (discussed below). Lastly, our inducible system permitted the comparison of retrotransposition in synchronized cell populations where the major difference was the number of mitoses completed.

Integrating our data and those of previous studies[8, 9], we propose that active cell division promotes retrotransposition. All three studies showed strong inhibition of retrotransposition when cells were arrested. Shi et al.[9] analyzed L1 RNA levels in their assay system and attributed the inhibitory effect on retrotransposition largely to reduced L1 transcription (Table 1). The assay systems used by this study and Kubo et al.[8] enabled retrotransposition at larger dynamic ranges, permitting the evaluation of additional layers of regulation. In this study, after Fluc signals were normalized to the co-expressed Rluc, we observed an additional 10-fold reduction in retrotransposition in cell-cycle arrested HeLa Tet-ORFeus cells and a 2.6-fold reduction in synchronized cells undergoing one fewer round of cell division. Thus, after factoring in the effect of drug treatment on L1 expression, our data support an important role of cell division in promoting efficient L1 retrotransposition in a manner independent of L1 expression. It is noteworthy that, even when the variable infection rate was not taken into consideration, Kubo et al.[8] found a three-fold reduction of retrotransposition in G1/S arrested cells in addition to a 16-fold reduction of retrotransposition in G0 arrested Gli36 cells (Table 1). On the other hand, both Kubo et al.[8] and the current study showed substantial retrotransposition in cell-cycle arrested cells (for two of the three inhibitors tested, we observed statistically significant Fluc signals at approximately 10-fold above the assay background). Currently, we cannot exclude the possibility that the residual retrotransposition observed in arrested cell populations in both studies originates from a minor population of cycling cells. An alternative explanation for such residual retrotransposition is that L1 retrotransposition may also be facilitated by a yet uncharacterized active nuclear import mechanism (Table 1). Indeed, the control experiments performed by Kubo et al.[8] in G1/S arrested cells, showing differential transduction by retroviral and lentiviral vectors, support the presence of an active nuclear import mechanism for L1 retrotransposition. It is noteworthy that some non-LTR (long terminal repeat) retrotransposons have evolved active nuclear import strategies for their propagation in respective host species. A precedent of active nuclear import has been reported for the telomeric repeat-specific SART1 retrotransposon from Bombyx mori: its ORF1p contains functional nuclear localization signals (NLSs), which are required for active retrotransposition[14]. Thus far, no NLS has been reported in mammalian L1 proteins although both ORF1 and ORF2 proteins have highly basic regions, which is a common feature of nuclear localized proteins. Alternatively, it is possible active nuclear import is mediated by other host-derived components of L1 RNP. Recently, two poly(A) binding proteins, PABPN1 and PABPC1, were found to be associated with L1 RNP[15]. Of particular interest, PABC1 was found to be critical for RNP formation; as it can shuttle between the cytoplasm and the nucleus, it would be interesting to determine whether PABPC1 mediates RNP nuclear import[15].

A caveat shared by all three studies is that the role of cell division in retrotransposition was mainly assessed in cancerous cell lines (Table 1; but note Shi et al.[9] also tested normal human fetal lung fibroblasts). Additionally, two of these studies ([8] and the current study) used non-native promoters to drive L1 expression (Table 1), which precludes the study of the endogenous transcriptional regulation of L1 with these systems. Nevertheless, a unifying view from these and other studies of L1 variants, mutations, and host factors ([15–19] and citations therein) is that retrotransposition is not a simple function of L1 expression. By extension, the level of L1 expression cannot be equated with the frequency of retrotransposition and the evaluation of retrotransposition should take into consideration the cell-cycle status. For example, in normal individuals, endogenous L1 expression has only been confirmed at protein level in testicular and ovarian tissues ([20–22]; reviewed in[23]). It is noteworthy that L1 ORF1p is detected in two distinct stages of male germ cell development, namely, in gonocytes (embryonic stage) and in meiotic/post-meiotic germ cells (prepuberal and through adulthood)[20–22]. However, during both stages cell division is limited: gonocytes are mitotically arrested in G0 phase[24] while spermatocytes only divide twice before becoming haploid spermatids. Similarly, in the female germline, L1 ORF1p is detected during the meiotic prophase I in embryonic oocytes[21], which are subsequently arrested in the diplotene stage of the meiotic prophase I and do not divide until puberty[25]. Therefore, both male and female germline development may have been programmed in a way that restricts excessive retrotransposition by avoiding frequent nuclear membrane breakdown when L1 is expressed. Thus, to understand the developmental timing of retrotransposition, it is imperative to measure the level of retrotransposition directly.