A key step in the RNA silencing pathway is the cleavage of double-stranded RNA precursors by Dicer to produce functional siRNAs. In a number of organisms, siRNAs have been implicated in silencing retrotransposable elements [7, 14]. To understand the biological function of the RNAi pathway against the chicken CR1 elements, we used a conditional loss-of-Dicer function in a chicken-human hybrid DT40 cell line that containing a single copy of human chromosome 21 [15]. A conditional knockout of Dicer was generated in which a tetracycline-repressible promoter controlled the expression of Dicer. In the presence of doxycycline (Dox), a gradual loss of Dicer protein was observed, with complete loss 48 h after addition of Dox (Figure 1A). Quantitative RT-PCR with primers specific for chicken Dicer revealed that mRNA levels were up to 92% lower in Dox-induced DT40 cells than in the control cells (Figure 1B; P = 0.001). Moreover, the Dicer-deficient chicken cells survived for up to 6 days and their phenotype was almost indistinguishable from that of wild-type DT40 cells (data not shown). To further confirm the loss of Dicer function, we analysed the aberrant accumulation of transcripts from α-satellite DNA repeats in Dicer-deficient DT40 cells (see Additional file 1, Figure S1). Consistent with previous research [15], loss of Dicer in chicken cells resulted in increased levels of transcripts from α-satellite sequences through the disruption of Dicer-related RNAi silencing machinery.
Upregulation of mammalian L1 transcripts in Dicer-deficient chicken cells
Dicer-mediated gene silencing is an evolutionally conserved system across eukaryotes. Previous studies in cultured human cells have identified that RNAi machinery elicited by the antisense transcripts of L1 retrotransposons can suppress L1 expression and retrotransposition efficiency [11]. To demonstrate the direct role of chicken Dicer in transposon silencing, and also to rule out the possibility that chicken may have multiple Dicer enzymes (as seen in Neurospora and Aspergillus [16, 17]) that are redundantly responsible for transposon-specific silencing, we used human chromosome 21 as a source of mammalian L1 elements in chicken-human hybrid DT40 cells. This system has two unique advantages: first, it allows us to study L1 expression from a natural chromosomal environment rather than from transfected plasmids; second, as human L1 sequences are completely different from chicken CR1 elements, it is easy to identify the transcript abundance of L1 retrotransposons by quantitative RT-PCR analysis without interference from chicken CR1 transposable elements.
Human chromosome 21 contains 65 copies of a full-length L1 element determined by TBLAST and L1Base searches [18] using the consensus sequence of the active L1 element (accession no. AF148856). Although the majority of predicted L1 elements contain both ORF1 and ORF2, all but one of them carry mutations and frame-shift interruptions or in-frame stop codons. Only one copy of the L1 element in human chromosome 21 (position 17669229 – 17676620) appeared to be functional with an ability to produce both ORFs active proteins. Using real-time quantitative RT-PCR, we first compared the abundance of L1 transcripts in Dicer-deficient and control DT40 cells. Remarkably, Dicer knockdown in chicken cells resulted in a significant increase (around 1.6-fold) in the levels of both ORF1 and ORF2 transcripts derived from L1 element of human chromosome 21 (Figure 1D; P = 0.005 and P = 0.006 for ORF1 and ORF2, respectively). In addition, we also confirmed the increased expression of L1 transcripts upon loss of Dicer by strand-specific RT-PCR analysis (Figure 1C), suggesting that Dicer may be required for silencing of human L1 retrotransposons. These data also imply that the Dicer-mediated RNAi silencing pathway is functional in chicken DT40 cells and enables them to control the levels of human L1 retrotransposons.
Since Dicer is a crucial component in the RNAi pathway, which processes dsRNA precursors to functional siRNAs, loss of Dicer may contribute to an increased expression of L1 transcripts due to the absence of, or poorly processed, siRNAs against L1 elements. An equally important possibility is that loss of Dicer also affects miRNA processing, and thus the increased expression of L1 transcripts in the absence of Dicer could be an effect of miRNA dysregulation that targets L1 expression directly, or indirectly through miRNA-mediated regulation of cellular factors that suppress L1 expression. Although bona fide siRNAs or miRNAs that control the human L1 elements are yet to be isolated, a previous study in cultured human cells [11] has shown that the Dicer knockdown or deletion of L1's antisense promoter increases the levels of L1 expression, suggesting that the siRNAs derived from the active L1 promoter control L1 expression. To confirm that the Dicer knockdown in chicken cells affects L1-derived small RNAs processing and thus the increased expression of L1 transcripts, we analysed the accumulation of small RNAs in control and Dicer-deficient DT40 cells (Figure 1E). Northern blot analysis shows reduced accumulation of L1-specific siRNAs in Dicer-depleted cells (up to 61 ± 3% as measured by signal density quantification) compared with control cells, but not their complete absence, even though the level of Dicer was almost completely lost 48 h after addition of Dox. It is not clear whether the leftover small RNAs in Dicer-depleted cells are pre-existing siRNAs or poorly processed siRNAs upon the loss of Dicer. Nonetheless, this study shows that the reduced levels of L1-specific small RNAs in Dicer-deficient cells correlate with upregulation of the human L1 transcripts.
Dicer knockdown activated mammalian L1 retrotransposition
To further confirm that the Dicer is indeed required for controlling human L1 expression and retrotransposition, we introduced L1 expression cassettes harbouring a retrotransposition indicator for cell culture-based assay [11, 19]. This cassette consists of a full-length human L1 tagged at its 3'UTR with an antisense enhanced green fluorescent protein (EGFP) gene, which is driven by a cytomegalovirus (CMV) promoter (Figure 2A). The EGFP gene is disrupted by a γ-globin intron in the same orientation as the L1 transcript. This arrangement ensures that EGFP expression occurs only after L1 transcription, splicing of the intron, reverse-transcription, and insertion of the L1 copy back into chromosomal DNA (that is, after the retrotransposition event).
The chicken DT40 cells were electroporated with an active L1 element (pCMV-RP99-eGFP) and assayed for EGFP expression in the presence or absence of Dox. An inactive L1 (pCMV-ΔRP99-eGFP) that contained two missense mutations in ORF1 [19] to abrogate retrotransposition activity was used as a negative control. Cells expressing L1 were screened for EGFP expression by flow cytometry in order to compare the L1 retrotransposition events between Dicer-deficient and control DT40 cells. EGFP-positive cells were not detected in any of the control cells, even after several passages; PCR analysis also confirmed the absence of the retrotransposed or spliced EGFP in control cells (Figure 2B). This observation suggests that no retrotransposition events occurred in the chicken cells that contained intact Dicer and RNAi pathways. Strikingly, Dicer knockdown in chicken cells resulted in a significant increase in EGFP-positive cells (0.8 ± 0.07%) after 48 h addition of Dox, indicating that Dicer is indeed required for controlling the expression of human L1 elements. PCR analysis of chicken genomic DNA also confirmed the splicing of the intron in the EGFP gene in Dicer-deficient cells; these cells remained EGFP-positive for 5 days in culture and then ceased to proliferate upon loss of Dicer and subsequently died before 7 days by apparent cell death or apoptosis [15]. Furthermore, we also observed a correlation between the relative abundance of L1 transcripts and retrotransposition events by real-time quantitative RT-PCR analysis (Figure 2C). The level of L1 transcripts was almost doubled after Dicer knockdown (P = 0.004 and P = 0.007 respectively for L1's ORF1 and ORF2) suggesting that chicken Dicer gives rise to an inhibitory effect on retrotransposition by reducing the levels of L1 transcripts, most probably via L1-derived siRNAs.
Dicer is not responsible for silencing of chicken CR1 transposons
If siRNAs were responsible for controlling the copy numbers of chicken CR1 transposons, one would expect that depletion of siRNAs by Dicer knockdown would lead to a higher level of endogenous chicken CR1 transcripts in the chicken genome. To examine whether chicken Dicer is responsible for processing siRNAs acting against the chicken CR1 transposons, we used both strand-specific RT-PCR (targeting the sense message) and real-time quantitative RT-PCR to assess the transcript levels of CR1-B elements in DT40 cells in the presence or absence of Dox for 2 days. Unexpectedly, there was no significant difference in transcript accumulation of chicken CR1 elements in Dicer-deficient cells compared with control DT40 cells (Figure 3A). The mRNA levels of both ORF1 and ORF2 of CR1-B elements in Dicer-deficient and control DT40 cells were close to background levels, as determined by real-time quantitative RT-PCR (Figure 3B; P = 0.004 and P = 0.006 for ORF1 and ORF2, respectively). Consistently, regardless of whether Dicer was weakly or strongly depleted, Dicer knockdown did not increase CR1 mRNA levels (data not shown). To further confirm this finding, we analysed transcript levels of the functionally active CR1-F element in chicken chromosome 6 (positions 2162180 – 2166909) [4]. Again, we found a basal level of CR1 transcription that was similar in both control and Dicer-deficient DT40 cells (Figure 3D). These data were further confirmed by strand-specific RT-PCR analysis targeting the sense message of CR1 elements (Figure 3C), suggesting that the Dicer-related RNAi machinery may not be responsible for controlling the endogenous chicken CR1 elements.
Next, to determine whether the CR1 elements are able to produce any transcription, we transfected the chicken DT40 cells with luciferase reporter constructs whose expression is driven by either functionally active CR1-B 5'UTR or L1 5'UTR as a promoter and performed the Dual-Luciferase assay with the Renilla luciferase plasmid. In this way, we measured promoter activity in the presence or absence of Dicer. As expected, transcript level of the human L1 5'UTR promoter in Dicer-depleted DT40 cells was around twofold greater than that of control cells but not chicken CR1-B 5'UTR promoter, whose level of expression was almost comparable to that of the control DT40 cells (Figure 3F). This observation suggests that the chicken CR-1 elements may not be controlled by the Dicer-mediated RNAi pathway. Interestingly, when we reprobed the northern blot of small RNAs against the mixture of synthetic sense-strand CR1-B 5'UTR sequences, we were not able to detect any CR1-specific antisense transcripts, regardless of whether the Dicer was present or absent (data not shown). This suggests that the CR1 promoter may not produce any siRNAs like the human L1 5'UTR promoter. Remarkably, according to the luciferase reporter assay, L1 promoter seems to be at least sixfold stronger than that of the CR1 element (Figure 3E), indicating the existence of major differences between chicken CR1 and other mammalian L1 elements in their 5'UTRs, which conceivably function as promoters.
Very few CR1 5'UTR promoters are functional
Evidence from RNAi silencing of human L1 retrotransposons suggests that the first step for RNAi machinery is synthesis of dsRNAs from the transcriptionally active 5'UTRs of retrotransposons [11, 14, 20]. The human L1 5'UTR is known to contain a bidirectional promoter (both sense and antisense promoters) that leads to the production of dsRNAs, which are processed by Dicer into siRNAs. To investigate the possibility that chicken 5'UTRs might transcribe dsRNAs, perhaps through DNA hypomethylation, we analysed all putative 5'UTRs of chicken CR1 elements identified through literature and database searches. A BLASTN search against a CR1 consensus sequence (GenBank Acc. U88211) suggests that the chicken genome probably contains around 26,650 copies of the CR1 element scattered throughout the genome. To identify potentially active CR1 elements with intact promoters, the 5'UTRs of the previously identified active F and B subfamilies [4, 6] were initially used for a Blast search against the chicken genome. About 148 UTRs (136 from the CR1-F and 12 from the CR1-B subfamilies) were identified and used in the promoter analysis. Surprisingly, except for 15 UTRs (seven CR1-F and eight CR1-B), the vast majority of CR1 5'UTRs (>90%) are found as short fragments (less than half of the expected size) and are severely truncated or mutated to varying degrees. Notably, most of the mutations or deletions occurred at putative transcription factor binding sites as determined by TRANSFAC searches [21] against the promoter sequences of the CR1 elements. Unlike the functionally active CR1 sequences, these elements do not contain cis-acting elements or putative promoter-like sequences upstream of the start codon, suggesting that the overwhelming majority of chicken CR1 elements do not have the necessary functional promoter sequences for initiation of RNA transcription.
The 5'UTR of the potentially active CR1-B element is approximately 240 bp long and contains two putative E boxes, the cis-element for binding of the basic helix-loop-helix family of proteins. E boxes have been noted as an obvious feature of vertebrate CR1-elements [22]. In addition to the E boxes, other potential binding sites for C/EBP, USF and Sp1 are also found in the promoter sequence of CR1-B element (Figure 4B and Additional file 1, Table S4). Unlike other vertebrate CR1-elements such as those in the turtle and puffer fish [22, 23], there is no obvious evidence of a 32-bp direct repeat sequence within the CR1-B promoter sequence. Several deletions or insertions, ranging from several nucleotides to several dozen nucleotides, were found in these regions of the CR1-B direct repeats suggesting that this region might have undergone frequent recombinational events. Direct repeat sequences are often thought to be involved in dsRNA production, which are processed by Dicer into functional siRNAs. Interestingly, the 5'UTR sequence of CR1-F element shows no resemblance to the 5'UTR of CR1-B elements. Unlike the CR1-B element, there is no sequence corresponding to canonical E boxes (Figure 4C). Promoter analysis of the CR1-F elements shows the presence of at least three regions corresponding to the binding of Oct-1/AP-1 (Fos/Jun activating protein-1) core elements within the region of promoter. In addition, a potential binding site for GR (glucocorticoid receptor) is also located at upstream of the start codon. GR is a zinc-finger DNA-binding protein likely to be involved in the activation of CR1-F elements through a Fos/Jun complex similar to that found in the chicken vitellogenin and ovalbumin genes [24]. This observation raises the possibility that cellular transcription factors that bind to these genes might act in concert to regulate the expression of CR1-F elements.
Sequence divergence of 5'UTR within each CR1 subfamily
The intact and functionally active CR1 elements identified thus far belong to the CR1-B and CR1-F subfamilies. Our study shows that the 5'UTR promoter sequence of B and F elements are distinct and unrelated to each other, suggesting the existence of sequence diversity in the 5'UTR promoters in the chicken genome. The relationships among the 5'UTRs of all other CR1 subfamilies (including the previously described CR1 elements C, D, E, and H) are poorly understood. To evaluate the biological significance of all these CR1 promoters, we downloaded the 5'UTR sequences of all CR1 subfamilies in the chicken genome and aligned them using ClustalX (see Additional file 1, Tables S2 and S3). Sequences that contained large truncations or deletions in the alignment of the 5'UTR sequences were removed, resulting in a final set of 169 sequences that were used for the construction of a phylogenetic tree (Figure 5). The 5'UTRs of chicken CR1 elements cluster into six distinct subfamilies. The largest group of 5'UTRs belongs to subfamilies D and G (with 75 and 51 elements, respectively) scattered throughout the chicken genome. Except for a very few copies of B and F subfamily members, none of the CR1 elements representing the C, D, G, and H subfamilies contain sequences that might provide functionally active CR1 elements (data not shown). This indicates that they are not retrotransposition competent. The most notable finding that emerged from our sequence analysis is that the 5'UTR sequence of the functionally active CR1-F element has diverged from that of the non-functional CR1-F elements (Additional file 1, Table S5) indicating the existence of distinct promoter sequences for functional elements within the CR1 subfamily of the chicken genome.
Functional CR1 elements are rare in the chicken genome
Next, we analysed whether the intact 5'UTR promoter contains the genomic DNA sequence of full-length ORF1 and ORF2 sequences. For this analysis, the consensus protein sequence of ORF1 was initially used in a protein BLAT search against the chicken genome. The regions that showed similarity were isolated along with 6 kb of their flanking sequences. In a second step, the resulting genomic DNA sequences were screened for the presence of an intact 5'UTR promoter. We identified only six full-length CR1 elements with intact promoters (four belonging to the CR1-F subfamily and two to the CR1-B subfamily) that have a size of >4,200 bp; the others were severely truncated at either the 5' or 3'region of ORFs. Out of four CR1-F elements, three were found to contain intact promoters capable of RNA transcription, but these elements are defective in the coding region of ORF1 and ORF2 due to frame-shift mutations (see Additional file 1, Table S6). Only one copy of the CR1-F element found on chromosome 6 with an intact promoter (chr. 6: position 2161469 – 2166909) appears to be functionally active with the ability to produce both ORF proteins. Its sequences are also in agreement with previous report of a functional 'mother' CR1 element [4]. Similarly, out of two intact CR1-B elements, only one copy a of CR1-B element with intact promoter and ORFs for both proteins was found on chromosome 5 (position 5715277 – 5719735). This element is 98% identical with the consensus sequence of CR1-B (accession no. U88211), whereas the other CR1-B element on chromosome 2 (51121971 – 51126431) was truncated at ORF1. Interestingly, two additional CR1-B elements on chromosomes 5 and 2 contained intact ORF regions (Additional file 1, Table S6) but these elements do not contain promoter sequences that can serve as an initiation point for transcription. In summary, the overwhelming majority of chicken CR1 elements appear to be non-functional and thus unable to replicate. Nonetheless, our findings suggest that there is at least one copy of the CR1-F and B subfamily still potentially capable of active in the chicken genome.