To understand the complex regulation of genomic imprinting it is important to determine how early embryos establish imprinted gene expression across large chromosomal domains. Long non-coding RNAs (ncRNAs) have been associated with the regulation of imprinting domains, yet their function remains undefined. Here, we investigated the mouse Kcnq1ot1 ncRNA and its role in imprinted gene regulation during preimplantation development by utilizing mouse embryonic and extra-embryonic stem cell models. Our findings demonstrate that the Kcnq1ot1 ncRNA extends 471 kb from the transcription start site. This is significant as it raises the possibility that transcription through downstream genes might play a role in their silencing, including Th, which we demonstrate possesses maternal-specific expression during early development. To distinguish between a functional role for the transcript and properties inherent to transcription of long ncRNAs, we employed RNA interference-based technology to deplete Kcnq1ot1 transcripts. We hypothesized that post-transcriptional depletion of Kcnq1ot1 ncRNA would lead to activation of normally maternal-specific protein-coding genes on the paternal chromosome. Post-transcriptional short hairpin RNA-mediated depletion in embryonic stem, trophoblast stem and extra-embryonic endoderm stem cells had no observable effect on the imprinted expression of genes within the domain, or on Kcnq1ot1 imprinting center DNA methylation, although a significant decrease in Kcnq1ot1 RNA signal volume in the nucleus was observed. These data support the argument that it is the act of transcription that plays a role in imprint maintenance during early development rather than a post-transcriptional role for the RNA itself.
Genomic imprinting is a specialized transcriptional regulatory mechanism that restricts expression to the maternally or paternally inherited allele (Verona et al., 2003; Barlow and Bartolomei, 2007). Imprinted genes often cluster together in large domains that are coordinately regulated by cis-acting regions known as imprinting centers (ICs) or imprinting control regions (ICRs). ICRs harbor gamete-derived parental allelic marks and are responsible for imprinted gene regulation over hundreds of kilobases in a bidirectional manner. Interestingly, imprinting domains are associated with a non-coding RNA (ncRNA) that may regulate imprinted gene expression for the entire cluster (Barlow and Bartolomei, 2007). Generally, imprinted ncRNAs are transcribed from the unmethylated ICR and can range from 2.2 to possibly more than 1000 kb in length and thus are duly referred to as long or macro ncRNAs.
The Kcnq1ot1/KCNQ1OT1 (potassium voltage-gated channel, member 1, overlapping transcript 1) imprinting domain is located on mouse chromosome 7 and human 11p15.5 (Verona et al., 2003; Barlow and Bartolomei, 2007). This domain contains one paternally expressed ncRNA, Kcnq1ot1, and eight maternally expressed protein-coding genes, including Slc22a18 (solute carrier family 22a, member 18), Cdkn1c (cyclin-dependent kinase inhibitor 1c), Kcnq1 (potassium voltage-gated channel, KQT-like subfamily, member 1) and Ascl2 (achaete-scute homolog 2) (Fig. 1). The Kcnq1ot1 transcription start site (TSS) is located within the ICR (Lee et al., 1999; Mitsuya et al., 1999; Smilinich et al., 1999; Mancini-DiNardo et al., 2003). When methylated on the maternal allele, Kcnq1ot1 is silent (Mancini-DiNardo et al., 2003). On the paternal allele, the ICR is unmethylated and Kcnq1ot1 is transcribed. Paternal deletion of the Kcnq1ot1 ICR results in domain-wide loss of imprinting, demonstrating broad control by the ICR of imprinted gene regulation (Fitzpatrick et al., 2002). Activation of the normally silent paternal alleles of the protein-coding genes is also seen when Kcnq1ot1 is truncated, possibly indicating a functional role for the macro ncRNA in imprinted domain regulation (Mancini-DiNardo et al., 2006; Shin et al., 2008).
The mechanism by which Kcnq1ot1 represses the expression of genes located more than 300 kb away is not completely understood, although various models have been proposed. These include a role for the ncRNA in nucleating silent chromatin, similar to Xist in X inactivation; a role for Kcnq1ot1 transcription in initiating chromatin silencing through transcriptional interference, repressive chromatin compartmentalization or chromatin looping; or a combination of these mechanisms (Mancini-DiNardo et al., 2003; Lewis et al., 2004; Lewis et al., 2006; Pauler et al., 2007; Pandey et al., 2008; Shin et al., 2008; Terranova et al., 2008; Koerner et al., 2009; Nagano and Fraser, 2009). Important regulatory elements have been identified by mutational analysis but such studies do not differentiate between these various possibilities. Deletion studies of the Kcnq1ot1 ICR domain indicate that the ICR, a functional promoter and/or transcription are required for domain imprinting (Fitzpatrick et al., 2002; Mancini-DiNardo et al., 2006). Truncation studies suggest that it is the transcript or transcription that has a role in silencing, as premature termination of Kcnq1ot1 results in the derepression of imprinted protein-coding genes within their domains (Mancini-DiNardo et al., 2006; Shin et al., 2008). However, these studies are unable to differentiate the function of the transcript from that of transcription. Additional complexity arises from tissue-specific differential imprinted gene regulation (Caspary et al., 1998; Lewis et al., 2004; Umlauf et al., 2004; Shin et al., 2008; Weaver et al., 2010). Osbpl5, Tssc4, Cd81 and Ascl2 are imprinted only in the placenta, whereas Phlda2, Slc22a18, Cdkn1c and Kcnq1 are imprinted in embryonic and placental tissues (Fig. 1). Thus, functional studies should include these different lineages in the analyses.
Because long ncRNAs may be central regulators of imprinted domains, it is important to elucidate their modes of action. In this study, we investigated the mouse Kcnq1ot1 ncRNA and its role in imprinted gene regulation, specifically during preimplantation development by utilizing embryonic and extra-embryonic stem cells. Our findings demonstrate that the Kcnq1ot1 ncRNA terminates 471 kb from the TSS, exists primarily as a full-length transcript and originates from the Kcnq1ot1 ICR. The length of Kcnq1ot1 is conserved in various tissues and at different developmental time points, suggesting that Kcnq1ot1 length does not contribute to differential tissue-specific silencing. However, the length is significant as it raises the possibility that transcription through downstream genes might play a role in their silencing, including tyrosine hydroxylase (Th), which we show possesses maternal-specific expression during early development. To differentiate between the function of the transcript and that of transcription, we employed RNA interference (RNAi)-based technology to induce degradation of Kcnq1ot1 transcripts. We hypothesized that post-transcriptional depletion of Kcnq1ot1 ncRNA would lead to activation of maternally transcribed genes from the paternal chromosome. Short hairpin RNA (shRNA)-mediated Kcnq1ot1 RNA depletion in embryonic stem (ES), trophoblast stem (TS) and extra-embryonic endoderm stem (XEN) cells had no observable effect on imprinted gene expression, nor on ICR DNA methylation, although a significant decrease in Kcnq1ot1 nuclear volume was observed. These data suggest that it is the act of ncRNA transcription that predominantly influences imprint maintenance during early development, rather than there being a post-transcriptional role for the RNA itself.
MATERIALS AND METHODS
C57BL/6J (B6) and Mus musculus castaneus (CAST) mice were obtained from Jackson Laboratory (Bar Harbor, ME, USA). B6(CAST7) mice contain Mus musculus castaneus chromosome 7s on a B6 background (Mann et al., 2004). B6 females were mated to CAST males to obtain embryonic day (E) 7.5, 8.5 and 9.5 B6XCAST embryos and placentas. CAST7 females were crossed with B6 males to recover E12.5 CAST7XB6 embryos and placentas. E9.5 placentas with a paternal IC2 deletion were also analyzed (Fitzpatrick et al., 2002). Experiments were performed in compliance with guidelines set by the Canadian Council for Animal Care and policies and procedures approved by the University of Western Ontario Council on Animal Care.
Transduction and transfection of stem cells
B6XCAST XEN, ES and TS cells were generated as described (Golding et al., 2010). Stem cells were transduced with an shRNA targeting Kcnq1ot1 at 43 kb from the TSS (K43), a non-functional shRNA homologous to Kcnq1ot1 at 82 kb (K82), or with an shRNA targeting the luciferase gene (LUC) (Golding et al., 2010). K43 shRNA, which starts at Kcnq1ot1 position 43,804, 5′-TGCTGTTGACAGTGAGCGACCAGAGTTTGTCTTTCATAAATAGTGAAGCCACAGATGTATTTATGAAAGACAAACTCTGGGTGCCTACTGCCTCGGA-3′, was processed into the 22mer 5′-CCCAGAGTTTGTCTTTCATAAA-3′; and K82 shRNA, which begins at Kcnq1ot1 position 82,541, 5′-TGCTGTTGACAGTGAGCGATGGCTTAAGCTGATCAATTAATAGTGAAGCCACAGATGTATTAATTGATCAGCTTAAGCCACTGCCTACTGCCTCGGA-3′ was processed into the 22mer 5′-GTGGCTTAAGCTGATCAATTAA-3′. Production of recombinant virus, infection, puromycin selection and passage of cells were performed as described (Golding et al., 2010). Cells were collected between passages 7 and 27. For the shRNA depletion studies, multiple independent cell lines were generated and independent biological replicates of wild-type (WT), LUC, K82 and K43 samples were used; six to ten each for XEN cell samples, four each for ES cell samples, and four to six each for TS cell samples. Robust depletion was observed in K43-transduced stem cells (82-93% depleted) compared with control cells (see Fig. S1 in the supplementary material). Importantly, there was considerable sample overlap for Kcnq1ot1 length analysis, allelic expression analysis and fluorescence in situ hybridization.
NIH 3T3 cells were obtained from the American Type Culture Collection (Manassas, VA, USA). Custom siRNAs targeting the putative Kcnq1ot1 ncRNA were designed using RNAi Codex (http://hannonlab.cshl.edu/GH_siRNA.html) and synthesized by Dharmacon (Thermo Scientific, Lafayette, CO, USA). si7, 5′-GGCAUACUGUCCAUACGUAUU-3′, starts at Kcnq1ot1 position 6933, and si463, 5′-CCGAGCAGAUGAUACAGUAUU-3′, starts at Kcnq1ot1 position 463,032. Control siRNAs were an siGLO Green Transfection Indicator to monitor transfection success and a scrambled sequence non-targeting siRNA (Dharmacon). Lipofectamine 2000 transfection reagent (Invitrogen, Burlington, ON, Canada) was used to transfect 40 μM siRNAs into NIH 3T3 cells as per the manufacturer's protocol. Cell collection was performed at passage 4.
DNA and RNA isolation
DNA was extracted using the DNeasy Blood and Tissue Kit (Qiagen, Mississauga, ON, Canada) and total RNA was extracted either using the Roche High Pure RNA Tissue Kit (Roche Applied Science, Laval, QC, Canada) or Trizol (Invitrogen) according to the manufacturers' instructions.
cDNA was synthesized from total RNA using Superscript II or III reverse transcriptase (Invitrogen). Reactions were primed using both random hexamers and oligo(dT) primers (Invitrogen). It was essential to ensure that successful PCR amplification was not a result of contaminating genomic DNA nor was impeded by any remaining RNA in the cDNA pool. To eliminate DNA contamination, two DNase steps were performed, one during and one after RNA extraction. Only Invitrogen DNaseI was used, as DNases from other manufacturers (Sigma-Aldrich, Oakville, ON, Canada; Roche Diagnostics, Laval, QC, Canada) were unable to completely remove genomic DNA. After cDNA synthesis, samples were treated with RNaseA to remove any residual RNA. Once the no-reverse-transcriptase controls for each sample showed no RNA or DNA contamination following beta-actin amplification (Table 1), the sample was used for PCR amplification.
Approximately 660 kb of DNA sequence was obtained from Ensembl (http://www.ensembl.org/index.html) for distal chromosome 7 (150,418,839 to 149,758,443 bp). Primers were manually designed within intronic and intergenic regions, with confirmation of appropriate primer attributes by NetPrimer (http://www.premierbiosoft.com/netprimer/index.html) (Table 1), and were synthesized by Sigma Genosys (Oakville, ON, Canada). Primers were initially designed at ∼50 kb increments from the Kcnq1ot1 TSS to 619 kb downstream. Once amplification was narrowed to a ∼50 kb window, primer pairs were designed at ∼5 kb intervals, and then at 1 kb intervals to locate the end of the transcript. PCR amplification was performed using illustra Ready-To-Go PCR Beads (GE Healthcare Biosciences, Baie d-Urfe, QC, Canada) at 95°C for 2 minutes, 35 to 45 cycles of 95°C for 30 seconds, Tm (see Table 1) for 30 seconds, and 72°C for 50 seconds, followed by 7 minutes at 72°C, with a 4°C hold.
Quantitative (q) PCR analysis
qPCR analysis of mRNA levels was carried out using the iQ SYBR Green Supermix (Bio-Rad, Mississauga, ON, Canada) following the manufacturer's instructions. Reactions were performed in triplicate on three to six samples each for control and experimental groups in shRNA experiments, and on two samples each for control and experimental groups in siRNA experiments. Using an MJ Thermocycler Chromo4 Real-Time PCR System (Bio-Rad), qPCR was conducted at 95°C for 2 minutes, 35 to 40 cycles of 94°C for 20 seconds, 58°C for 30 seconds, 72°C for 45 seconds, then 72°C for 7 minutes. Samples were normalized to the reference gene beta-actin or to mitochondrial ribosomal protein L1 (Mrpl1). As controls for off-target effects, Airn and H19 expression was examined in the shRNA depletion study. For analyses, control samples (WT cells in shRNA studies and untreated cells in siRNA experiments) were set to a value of 1, and experimental reactions were normalized to the control. To determine relative levels of expression across the Kcnq1ot1 ncRNA, qPCR was performed for three E9.5 placental and two XEN cell cDNAs in duplicate and standardized to genomic DNA amplification to control for primer efficiency. qPCR was performed in triplicate using three different DNA samples. Mean expression levels for each primer set were then standardized to the 3k primer set. Standard error of the mean (s.e.m.) was calculated.
Allelic expression analysis
Primers (Table 1), polymorphisms between B6 and CAST, and allele-specific restriction enzymes were reported for Slc22a18 (Dao et al., 1998), Cdkn1c (Doherty et al., 2000), Kcnq1 (Gould and Pfeifer, 1998; Jiang et al., 1998), Kcnq1ot1 (Rivera et al., 2008) and Ascl2 (Mann et al., 2003). For Th, a polymorphism between B6 (G) and CAST (A) was identified at position 1046 (NM_009377). Restriction digestion with BsrI resulted in 124 and 71 bp fragments in B6, whereas the CAST amplicon was uncleaved. For Airn, a polymorphism at SNP rs6154084 on chromosome 17:12972645 between B6 (T) and CAST (C) was utilized in a restriction assay with AvaI to produce 220 and 156 bp fragments in CAST, whereas the B6 amplicon was uncleaved. To quantify the relative amounts of amplicons, computer-assisted densitometry was performed with QuantityOne 1-D Analysis Software (Bio-Rad) to calculate the adjusted intensity/mm2 for each band. The H19 allelic expression assay using the LightCycler Real-Time PCR System (Roche Molecular Biochemicals) was performed with hybridization probes (Table 1) as described (Mann et al., 2004). Parental allele-specific expression was calculated as the percentage B6 or CAST expression relative to the total expression. Monoallelic expression was defined as ≥90% from one parental allele.
3′ Rapid amplification of cDNA ends (3′RACE)
The 3′RACE System Kit (Invitrogen) was used for cDNA synthesis and for semi-nested PCR according to manufacturer's protocol with the 470F (first round) or 471F (second round) forward primer (Table 1) paired with the universal adapter primer on E9.5 placental and E12.5 embryo RNA. Following electrophoresis, bands of interest were recovered using the Qiaquick Gel Extraction Kit (Qiagen), DNA fragments were ligated into the pGEM-T EASY vector (Promega, Napean, ON, Canada), and the vector was transformed into competent Escherichia coli. DNA was extracted from bacterial colonies using the GenElute Plasmid Miniprep Kit (Sigma-Aldrich). Sequencing was performed at the London Regional Genomics Centre (London, ON, Canada), Nanuq Sequencing Facility (Montreal, QC, Canada) or BioBasic (Markham, ON, Canada). BLAST analysis (http://blast.ncbi.nlm.nih.gov/Blast.cgi) was performed to align resulting sequences with the GenBank database.
Bisulfite mutagenesis and sequencing with agarose embedding was performed as described (Market-Velker et al., 2010; Golding et al., 2010). Lysed cells (10 μl) were embedded in 20 μl 2% low-melting-point agarose (Sigma). Following bisulfite mutagenesis, 22 μl diluted agarose was added to Ready-To-Go PCR Beads containing Kcnq1ot1 BIS outer primers (Table 1) and 1 μl 240 ng/ml tRNA. PCR reactions were halved, allowing for two independent PCR reactions. The first round product (5 μl) was seeded into each second round PCR reaction with Kcnq1ot1 BIS inner primers (Table 1). Sequencing was performed at the Nanuq Sequencing Facility or BioBasic. Sequences with less than 95% conversion rates were not included. Percentage methylation was calculated as the number of hypermethylated DNA strands/total number of DNA strands. Hypermethylated DNA strands displayed ≥50% methylated CpGs.
RNA fluorescence in situ hybridization (FISH)
Kcnq1ot1 and Airn FISH probes were generated from a 32 kb region of fosmid Wl1-2505B3 in the Kcnq1 intronic region and a 42 kb region of fosmid Wl1-270O22, respectively (CHORI, Oakland, CA, USA), using the BioPrime DNA Labeling System (Invitrogen) and fluorescein-12-dUTP (Roche Diagnostics) with minor modifications; Invitrogen dNTPs were used and labeling reactions were incubated for 12 hours. Unincorporated dNTPs were removed using ProbeQuant G-50 Micro Columns (GE Healthcare). Probes were precipitated with Cot-1 DNA, yeast tRNA and salmon sperm DNA (Invitrogen) and washed with 75% and 100% ethanol. Probes were air dried then resuspended in 100% deionized formamide and denatured at 85°C for 10 minutes. After 2 minutes on ice, 2× hybridization mix (25% dextran sulfate, 4× SSC) was added, incubated at 37°C for 30-90 minutes, then stored at –20°C.
XEN cells were seeded on glass slides (VWR, Mississauga, ON, Canada), permeabilized with sequential transfers into ice-cold cytoskeletal extraction buffer (CSK) for 30 seconds, ice-cold CSK containing 0.25% Triton X-100 (Sigma-Aldrich) for 45 seconds and ice-cold CSK for 30 seconds (Kalantry et al., 2009), fixed in 4% paraformaldehyde at room temperature for 10 minutes (Panning and Jaenisch, 1996), washed in PBS, dehydrated in sequential washes of 85%, 95% and 100% ethanol, then air dried. RNA-FISH probes were hybridized overnight and washed as described (Murakami et al., 2007). Cells were mounted and stained with Vectashield augmented with DAPI (Vector Laboratories, Burlington, ON, Canada) and imaged using z-stacks on a FluoView FV1000 coupled to an IX81 motorized inverted system microscope (Olympus, Markham, ON, Canada). Fluorescence signal volumes were measured using intensity-based thresholds in Volocity (v5.2.0, PerkinElmer, Woodbridge, ON, Canada).
One-tailed Student's t-tests were performed on mean qPCR values and on mean fluorescence signal volumes. P-values were considered to be significant at P<0.05.
Kcnq1ot1 is a macro ncRNA
To better define the role of the Kcnq1ot1 ncRNA in imprinted gene regulation, we first delineated transcript length, as at least four conflicting reports exist regarding the ncRNA length (Horike et al., 2000; Yatsuki et al., 2000; Pandey et al., 2008; Redrup et al., 2009). Primer pairs were designed to amplify intronic and intergenic regions from E9.5 placental cDNA starting at 3 kb from the Kcnq1ot1 TSS and extending to 619 kb, within Ins2. Only primer sets that successfully amplified genomic DNA were utilized. A failure of primer pairs to amplify cDNA but not genomic DNA indicated that there was no transcript present at that location. Amplification with primer sets ∼50 kb apart revealed that the 3′ end of Kcnq1ot1 resides in a region between 449 to 486 kb (Fig. 1). Primer pairs were then designed every ∼5 kb within this 37 kb region. A fragment was detected at 457, 463 and 468 kb, but not at 475 and 479 kb. Finally, primer sets were designed at less than 1 kb intervals between 463 to 475 kb. Whereas primers between 463 and 470 kb successfully amplified cDNA, primers at 471, 474 and 475 kb failed to produce amplicons. To identify the 3′ end of the transcript, 3′RACE was performed on E9.5 placental and E12.5 embryonic cDNA. Sequence analysis of cloned 3′RACE products mapped the polyadenylation signal to 471 kb (Fig. 1). To confirm transcript termination at the polyadenylation signal, primer sets were designed between 470,857 and 471,291 bp (see Fig. S2 in the supplementary material). The end of the transcript resides at 471,164 bp.
Kcnq1ot1 transcript length is invariant
Midgestation embryos display differential regulation of imprinted genes in embryonic and placental tissue. One explanation for this differential regulation is that alternative Kcnq1ot1 transcripts exist with variant lengths. To determine whether Kcnq1ot1 length was tissue-dependent, B6XCAST ES cells, XEN cells, TS cells and neonatal brain samples, as well as CAST7XB6 E12.5 embryo and placenta, were compared with B6XCAST E9.5 placenta using selected primer sets along the putative Kcnq1ot1 ncRNA. Amplification was observed up to 470 kb in all tissues, but not beyond (Fig. 2A). Thus, in all three lineages and at all developmental time points examined, Kcnq1ot1 length was conserved. To determine relative levels of expression across the Kcnq1ot1 ncRNA, cDNA amplifications for individual primer sets were standardized to the 3k primer set (Fig. 2B). No statistical difference was observed in transcript abundance across the ncRNA, supporting the premise that the ncRNA is a single RNA moiety.
Kcnq1ot1 is primarily transcribed as a full-length transcript that originates from the Kcnq1ot1 promoter
To definitively demonstrate that the amplified products corresponded to the Kcnq1ot1 transcript, we employed RNAi technology, as well as a targeted deletion of the paternal Kcnq1ot1 ICR. Given the demonstrated ability of short interfering RNAs (siRNAs) and shRNAs to deplete nuclear transcripts (Robb et al., 2005; Willingham et al., 2005; Zhao et al., 2008), we first determined whether amplified PCR products were specific to the Kcnq1ot1 ncRNA using RNAi. As a shRNA targeting Kcnq1ot1 will result in degradation of the entire Kcnq1ot1 ncRNA (Hannon, 2002), it was expected that amplicons specific to Kcnq1ot1 would exhibit reduced RNA levels compared with controls, whereas those from another undefined transcript would be unaffected by RNA depletion. Real-time qPCR was performed using cDNA from transgenic XEN cells containing a shRNA targeting Kcnq1ot1 at 43 kb from the TSS (K43), control XEN cells harboring an ineffectual shRNA (K82), and WT cells. XEN cells were chosen because they are functionally relevant as preimplantation extra-embryonic cells, exhibit abundant expression of Kcnq1ot1, and are effectively transduced with the shRNA targeting vectors (Golding et al., 2010). qPCR was performed with primers at intronic and intergenic regions: 3k, 65k, 202k, 307k, 392k, 463K and 470k (Fig. 3A). Amplification of Kcnq1ot1 was significantly decreased (88-97%) in K43-transduced XEN cells compared with controls for all primer pairs examined (P<0.05) (Fig. 3C), indicating that all amplicons represent the Kcnq1ot1 ncRNA. As controls for off-target effects, the expression of two other ncRNAs, Airn and H19, was examined. Expression of Airn and H19 was not statistically different among WT, K82 and K43 XEN cells (Fig. 3C).
If a single full-length RNA is primarily produced, we would expect that targeting near the terminal region of the transcript would yield a similar depletion of amplified regions as targeting the ncRNA near the TSS. If multiple transcripts are produced of varying length, we would anticipate greater depletion for more terminal regions than for more proximal regions when the ncRNA is targeted at its distal end. To test this, siRNAs were designed to target Kcnq1ot1 at the proximal and distal ends at 7 and 463 kb from the TSS (si7 and si463) (Fig. 3A). Owing to the ease of transfection and the availability of established protocols, mouse embryonic fibroblast (NIH 3T3) cells were transfected with a scrambled, non-targeting control siRNA, si7 or si463, and were compared with non-transfected cells. Real-time RT-PCR analysis was performed with primers 3k, 94k, 118k, 392k and 470k along the length of the putative transcript. NIH 3T3 si7- and si463-targeted cells had significantly reduced transcript abundance (for si7 and si463, respectively: 3k, 70% and 76%; 94k, 84% and 93%; 118k, 98% and 100%; 392k, 93% and 97%; 470k, 84% and 82%) compared with control cells (P<0.05) (Fig. 3D). As depletion of Kcnq1ot1 at the proximal and distal ends of the transcript were not statistically different for each primer set, and as distal amplicons were no more reduced than proximal regions for si463, these results indicate that the majority of Kcnq1ot1 transcripts are full length and extend at least 470 kb.
As a final confirmation that the Kcnq1ot1 transcript extends 471 kb, a targeted deletion of the paternal Kcnq1ot1 ICR (Fitzpatrick et al., 2002) was employed (Fig. 3B). qPCR was performed using cDNA from E9.5 B6XCAST WT and paternal Kcnq1ot1 ΔIC2 placenta using primer sets 3k, 65k, 118k, 202k, 307k, 392k, 463k and 470k (Fig. 3E). For all primer sets, paternal ΔIC2 placenta failed to produce an amplification product, in contrast to control placenta, verifying that Kcnq1ot1 transcript length is at least 470 kb and that the amplicons are specific to a transcript originating from the Kcnq1ot1 ICR/promoter.
Imprinted status of Th
Th is located between the Kcnq1ot1 and H19 imprinting domains, but is not recognized as a member of either domain as targeted disruption of Th failed to demonstrate any parental origin effects (Zhou et al., 1995). However, another study using microarray technology to identify imprinted genes found that Th is maternally expressed in placenta (Schulz et al., 2006). As we found that Kcnq1ot1 was transcribed through Th, this warranted further analysis of its imprinted status. RT-PCR was performed on B6XCAST E7.5, E8.5, E9.5 and CAST7XB6 E12.5 embryonic and placental cDNA. The relative maternal and paternal transcript abundance was determined by BsrI allelic restriction digestion (Fig. 4). Th displayed maternal expression in E7.5 embryo and placenta and E8.5 placenta, and preferential maternal expression in E9.5 and E12.5 placenta, but was biallelically expressed in the corresponding embryonic tissues. These results demonstrate that Th shares a similar expression pattern with the neighboring genes Tssc4 and Cd81, with maternal expression in extra-embryonic lineages but progressive biallelic expression in the embryo (Caspary et al., 1998; Lewis et al., 2004; Umlauf et al., 2004).
Imprinting is maintained upon Kcnq1ot1 RNA depletion
Previous studies have shown that truncation of Kcnq1ot1 or deletion of the Kcnq1ot1 promoter results in derepression of imprinted genes within the domain. However, in these studies there was no delineation between transcription of Kcnq1ot1 and a functional role for the transcript itself (Fitzpatrick et al., 2002; Mancini-DiNardo et al., 2006; Shin et al., 2008). If the transcript has a function in silencing genes in this region, its depletion would be expected to result in activation of maternally transcribed genes from the paternal chromosome. To determine whether Kcnq1ot1 depletion results in loss of imprinting within the Kcnq1ot1 domain, imprinted methylation and expression were assayed in ES, TS and XEN cells. The DNA methylation status of the Kcnq1ot1 ICR was determined by the bisulfite mutagenesis and sequencing assay. Results showed no change in Kcnq1ot1 ICR methylation patterns in B6XCAST K43-transduced cells as compared with B6XCAST WT and K82-transduced control cells. Maternal DNA strands were hypermethylated, whereas paternal DNA strands remained hypomethylated (Fig. 5), indicating that depletion of the Kcnq1ot1 transcript in stem cells had little effect on ICR methylation.
To assess the effects of Kcnq1ot1 depletion on imprinted expression, allelic expression analysis was performed for the control genes H19 and Airn and for five imprinted genes in K43-transduced cells, as well as WT, K82-transduced and LUC-transduced control ES, TS and XEN cells (Fig. 6). Depletion of Kcnq1ot1 by shRNA K43 had no effect on H19 or Airn imprinted expression. As bidirectional silencing of imprinted genes occurs within the Kcnq1ot1 imprinted domain, two genes upstream (Slc22a18 and Cdkn1c) and three genes downstream (Kcnq1, Ascl2 and Th) of the Kcnq1ot1 ICR were analyzed. Similar to control ES, TS and XEN cells, K43-transduced cells displayed primarily monoallelic expression from the maternal Kcnq1 and Cdkn1c alleles, as previously observed for WT ES and TS cells (Umlauf et al., 2004; Lewis et al., 2006). For Slc22a18 and Th, K43-transduced and control TS and XEN cells were found to have monoallelic expression, whereas ES cells exhibited biallelic expression. Ascl2 expression was monoallelic in TS cells but biallelic in ES and XEN cells, with little difference between K43-transduced and control cells. Thus, imprinted gene expression was maintained in K43-transduced stem cells despite a ≥82% reduction in the Kcnq1ot1 ncRNA. Thus, depletion of Kcnq1ot1 RNA in embryonic and extra-embryonic stem cells does not result in domain-wide loss of imprinting.
The Kcnq1ot1 ncRNA localizes to the nuclear compartment as a strong signal that overlaps the paternal Kcnq1ot1 domain (Pandey et al., 2008; Terranova et al., 2008; Redrup et al., 2009). Given that imprinted expression and methylation were maintained despite a substantial decrease in Kcnq1ot1 ncRNA, we next determined whether Kcnq1ot1 ncRNA localization was abrogated by K43 RNAi depletion. RNA FISH was performed using probes for Kcnq1ot1 and, as a control, Airn. Cells were analyzed by confocal microscopy, and volumetric measurements were taken (Fig. 7). For the purposes of comparison, signal volumes were classified according to Redrup et al. (Redrup et al., 2009). Airn showed similar volume distributions for WT and K43-transduced XEN cells: 47.9% and 50.4% for 0-0.4 μm3, 21.4% and 20.9% for 0.4-0.8 μm3, 16.3% and 12.5% for 0.8-1.2 μm3, 8.3% and 5.1% for 1.2-1.6 μm3, 3.3% and 5.1% for 1.6-2.0 μm3 and 2.6% and 6.0% for >2.0 μm3, respectively (Fig. 7). No significant difference was found for mean Airn volume in WT and K43-transduced cells, at 0.58 and 0.61 μm3, respectively. This contrasts with Kcnq1ot1, where volume size was significantly decreased in K43-transduced cells compared with WT XEN cells: 68.8% versus 53.1% for 0-0.4 μm3, 19.9% versus 24.9% for 0.4-0.8 μm3, 8.4% versus 11.8% for 0.8-1.2 μm3, 2.3% versus 6.1% for 1.2-1.6 μm3, 0.3% versus 2.4% for 1.6-2.0 μm3 and 0.3% versus 1.7% for >2.0 μm3, respectively (Fig. 7). Overall, mean Kcnq1ot1 volume was significantly decreased in K43-transduced XEN cells: 0.30 μm3, as compared with 0.49 μm3 for WT cells (P<6×10–7). Thus, we conclude that imprinted methylation and expression are maintained even though the majority (69%) of K43-transduced cells exhibit very little or no RNA accumulation.
In this study, we found that the Kcnq1ot1 macro ncRNA extends to 471 kb from the TSS, exists predominantly as a full-length transcript, and originates from the Kcnq1ot1 ICR. Kcnq1ot1 length was conserved in embryonic and placental tissues, as well as in embryo-derived stem cells and postnatal brain. The extensive length of Kcnq1ot1 is significant as it suggests that transcription through downstream genes might play a role in their silencing. Post-transcriptional shRNA-mediated Kcnq1ot1 RNA depletion in embryonic and extra-embryonic stem cells had no effect on the imprinted expression of genes within the domain, nor on DNA methylation at the Kcnq1ot1 ICR, although a significant decrease in Kcnq1ot1 signal volume was observed. These data support the argument that it is the act of transcription that plays a role in imprint maintenance during early development, rather than there being a post-transcriptional role for the RNA itself.
Kcnq1ot1 length and function
Original estimates placed Kcnq1ot1 at ∼60 kb based on sequence homology between mouse and human (Horike et al., 2000; Yatsuki et al., 2000), 91.5 kb from the TSS (Pandey et al., 2008), or between 80 kb and 120 kb with a polyadenylation site located 121 kb downstream (Redrup et al., 2009). In these studies, the Kcnq1ot putative stop site falls short of the Kcnq1 TSS. Here, we demonstrate using RNAi technology as well as a paternal IC2 deletion that the Kcnq1ot1 ncRNA extends through downstream imprinted genes to 471 kb from the TSS.
The mouse and human Kcnq1ot1/KCNQ1OT1 domains possess a high degree of synteny (Paulsen et al., 1998; Engemann et al., 2000; Onyango et al., 2000; Paulsen et al., 2000; Yatsuki et al., 2000). Both domains are ∼1 Mb and the maternal ICR is methylated and the Kcnq1ot1/KCNQ1OT1 ncRNA is expressed from the paternal allele (Verona et al., 2003; Barlow and Bartolomei, 2007). However, notable differences also exist. In mouse, the region between Th and Ins2 is over 200 kb and contains an extraordinarily high concentration of tandem repeats, long interspersed nuclear element (LINE) retrotransposons and endogenous retroelements (Shirohzu et al., 2004). This repeat-rich region also exhibits asynchronous replication, similar to the rest of the domain, and contains 18 matrix attachment regions (Yatsuki et al., 2000; Purbowasito et al., 2004). In human, the distance between TH and INS is considerably shorter (∼10 kb). However, a highly repetitive region that is enriched for LINEs and SINEs (short interspersed elements) and retroelements is found between ASCL2 and TH (a ∼115 kb region) in humans (Shirohzu et al., 2004) and in the marsupial tammar wallaby (Ager et al., 2008). If repetitive regions play a role in boundary function, this might indicate that the boundary between the KCNQ1OT1 and H19 domains in humans is within the ASCL2 and TH repetitive region. Current evidence argues against this, however, as ASCL2 apparently escapes genomic imprinting in humans (Miyamoto et al., 2002). Alternatively, if repetitive elements are unrelated to boundary function, the boundary might lie within the conserved region upstream of INS/Ins2. A recent investigation of a maternally expressed GFP reporter located more than 600 kb downstream of the Kcnq1ot1 ICR and 2.6 kb upstream of Ins2 (Jones et al., 2011) found that the reporter was regulated in the same manner as other maternally expressed genes within the Kcnq1ot1 imprinted domain, possibly placing the boundary within 2.6 kb of Ins2.
Imprinted Kcnq1ot1 domain regulation
Precedent exists for macro ncRNAs residing within imprinted domains (Koerner et al., 2009), ranging from Xist as a 17 kb ncRNA to Ube3a-ats (Snrpnlt, Lncat), which possibly extends more than 1000 kb (Koerner et al., 2009). If the Kcnq1ot1 ncRNA functioned similarly to Xist, a ncRNA involved in X chromosome inactivation (Pauler et al., 2007), then an excessively long RNA would be unnecessary as the 17 kb Xist ncRNA is more than sufficient to coat and effectively silence an entire chromosome (Brockdorff et al., 1992). In addition, if transcript length is a determining factor in imprinted gene regulation, we would expect that differential regulation of the Kcnq1ot1 domain in embryonic and extra-embryonic tissues would correlate with shorter and longer transcript lengths, respectively. Instead, we observed that the Kcnq1ot1 transcript length remained the same in embryonic and extra-embryonic lineages and at all developmental stages examined, indicating that Kcnq1ot1 mRNA length does not contribute to differential silencing.
Various models have been proposed for Kcnq1ot1 imprinted domain regulation, including a role for the ncRNA itself and a role for Kcnq1ot1 transcription in initiating chromatin silencing (Mancini-DiNardo et al., 2003; Lewis et al., 2004; Lewis et al., 2006; Pauler et al., 2007; Shin et al., 2008; Koerner et al., 2009; Nagano and Fraser, 2009). To differentiate between functions of the transcript and of transcription, we employed RNAi technology that preserves transcription but degrades the macro ncRNA post-transcriptionally. Kcnq1ot1 transcripts were depleted in both embryonic and extra-embryonic stem cells, in case the Kcnq1ot1 transcript or its transcription had different regulatory roles in these cell types. Despite significant depletion (>80%), the imprinted expression of genes within the domain and the imprinted DNA methylation of the Kcnq1ot1 ICR were maintained in all three lineages of the early embryo. These results suggest that it is the act of transcription that plays a role in imprinted gene regulation rather than there being a post-transcriptional role for the RNA itself. If the ncRNA had played a post-transcriptional role in imprinted gene regulation, we would have anticipated the activation of the normally silent paternal alleles of protein-coding genes in these cells, which would have been detected in the allelic expression assays.
RNA-depletion studies have also been performed for the Gtl2 (Meg3 – Mouse Genome Informatics) and Xist ncRNAs. Depletion of Gtl2 results in a ∼50% reduction in histone H3 lysine 27 trimethylation (H3K27me3) and in the overexpression of neighboring imprinted genes (Zhao et al., 2010); this study also demonstrates that 50-70% ncRNA depletion is sufficient to produce a change in imprinted gene regulation. Depletion of the 1.6 kb RepA ncRNA within the Xist locus in female ES cells abolishes Xist localization and H3K27me3 (Zhao et al., 2008). This contrasts with female ES cells transduced with a shRNA targeting Xist exon 1, where a 70-80% depletion of Xist reduced the number of cells exhibiting Xist localization but had no perceptible effect on H3K27me3. Furthermore, compacted preimplantation embryos lacking Xist were proficient at silencing X-linked genes on the paternal chromosome, indicating that initiation of X inactivation can occur in the absence of the Xist ncRNA (Kalantry et al., 2009). Together, these studies suggest a limited post-transcriptional role for the long ncRNAs Xist and Kcnq1ot1.
One caveat of these RNA-depletion experiments is that they represent a snapshot in time. Thus, the Kcnq1ot1 ncRNA might be required later in development to maintain domain imprinting, similar to the Xist ncRNA, which stably silences paternal X-linked genes in postimplantation extra-embryonic tissues (Kalantry et al., 2009). As Slc22a18 has still to acquire imprinted expression in ES cells, this indicates that imprinting within the Kcnq1ot1 domain has not yet been completely established. Consistent with this, parental-specific histone modifications have not yet been acquired in ES and TS cells at imprinted genes that exhibit postimplantation placenta-specific imprinted expression (Lewis et al., 2006). Thus, further studies are required during late embryogenesis to delineate the role of the Kcnq1ot1 ncRNA in imprinted gene regulation. Alternatively, the Kcnq1ot1 ncRNA might function in imprinted domain establishment. As depletion occurred after fertilization in late preimplantation stem cells (i.e. not in germline-inherited cells), imprinting of the Kcnq1ot1 domain may have already been largely established, after which point the Kcnq1ot1 ncRNA might have little or no role to play in imprint maintenance. This would argue for an examination of Kcnq1ot1 function by RNAi-based methods prior to the two-cell stage, when the RNA is first transcribed. Either way, our data, produced by shRNA-targeted depletion of the Kcnq1ot1 ncRNA, suggest that any post-transcriptional role for the ncRNA in imprint maintenance during early development is limited.
One can envisage three transcription-based regulatory models for the Kcnq1ot1 imprinted domain. First, transcription of Kcnq1ot1 might cause interference with RNA polymerase II access to other promoters, thereby preventing paternal protein-coding gene transcription (Pauler et al., 2007; Koerner et al., 2009). A 471 kb macro ncRNA could effectively silence genes as it is transcribed through them. However, we found that the Kcnq1ot1 ncRNA is of invariant length. This means that the Kcnq1ot1 ncRNA would be transcribed through all promoters in both embryonic and extra-embryonic tissues, including those in the embryo/embryonic stem cells that are not silenced, such as Ascl2 and Th. Second, the ncRNA might not be required post-transcriptionally but is instead needed during the act of transcription, with the ncRNA acting as a tether or a cis-guide for recruiting repressive complexes (Lee, 2009; Tsai et al., 2010). However, the transcriptional interference and RNA-tethering mechanisms fail to explain how genes on the opposite side of the Kcnq1ot1 promoter are silenced and thus cannot be the sole mechanism of imprinted gene regulation. Third, transcription through one or more cis-regulatory elements might be necessary to evoke domain-wide silencing. In this case, ncRNA transcription might function to set up a repressive nuclear compartment (Shin et al., 2008; Nagano and Fraser, 2009), possibly via retroelement transcription and Argonaut1 repressive remodeling complex recruitment (Golding et al., 2010) or by chromatin looping (Koerner et al., 2009). The fact that Kcnq1 and Th displayed imprinted expression in XEN cells but Ascl2 was biallelically expressed might support this model, with Ascl2 excluded from the repressive compartment or loop. Interestingly, each of these models is congruent with the fact that the majority of WT XEN cells possessed very low to no Kcnq1ot1 nuclear signal, an observation that might be reflective of a more transient role for the ncRNA or for its transcription.
Much further investigation will be required to unravel the functional complexities of ncRNAs and their transcription in imprinted gene regulation. Novel technologies in transcription biology, such as the use of synthetic pyrrole-imidazole polyamide gene silencers, as well as the identification of a larger number of chromatin remodelers/modifiers involved in imprinted gene regulation, will lead to advances in our understanding of these mechanisms.
We thank Rosemary Oh and Louis Lefebvre for ΔIC placenta; Andy Fedoriw and Terry Magnuson for FISH reagents, protocols and advice; Liana Kaufman, Morgan McWilliam and Malaika Roussouw-Miles for technical assistance. The authors acknowledge the late Galina V. Fitzpatrick, who generated the Kcnq1ot1 ICR deletion mice. She will be sorely missed. This work was supported by research grants from NSERC 326876-06, Lawson Health Research Institute and Department of Obstetrics and Gynecology, University of Western Ontario to M.R.W.M., and by NCI/NIH grant 2RO1 CA089426 to M.J.H. M.R.W.M. was supported by the Ontario Women's Health Council/CIHR Institute of Gender and Health New Investigator Award. M.C.G. was supported by the Ontario Women's Health Council/CIHR Institute of Gender and Health Fellowship Award and the Dr David Whaley Postdoctoral Fellowship in Maternal/Fetal and Neonatal Research. Deposited in PMC for release after 12 months.
Competing interests statement
The authors declare no competing financial interests.
Supplementary material for this article is available at http://dev.biologists.org/lookup/suppl/doi:10.1242/dev.057778/-/DC1
↵* Present address: Department of Veterinary Physiology, College of Veterinary Medicine, Texas A&M University, College Station, TX 77843, USA
↵† These authors contributed equally to this work
- Accepted June 17, 2011.
- © 2011.