The molecular dynamics of long noncoding RNA control of transcription in PTEN and its pseudogene

Significance

In recent years, noncoding RNA transcripts have been found to interact with genes and modulate their ability to be transcribed and made into protein. Here we uncover many of the mechanistic underpinnings involved in how noncoding RNAs control gene transcription. Notably, we find that noncoding RNA control of gene transcription is based on a combination of structural and sequence components of the noncoding RNA and targeted gene. Collectively, the observations presented here suggest that a much more complex and vibrant RNA regulatory world is operative in gene expression and evolution of the genome.

Abstract

RNA has been found to interact with chromatin and modulate gene transcription. In human cells, little is known about how long noncoding RNAs (lncRNAs) interact with target loci in the context of chromatin. We find here, using the phosphatase and tensin homolog (PTEN) pseudogene as a model system, that antisense lncRNAs interact first with a 5′ UTR-containing promoter-spanning transcript, which is then followed by the recruitment of DNA methyltransferase 3a (DNMT3a), ultimately resulting in the transcriptional and epigenetic control of gene expression. Moreover, we find that the lncRNA and promoter-spanning transcript interaction are based on a combination of structural and sequence components of the antisense lncRNA. These observations suggest, on the basis of this one example, that evolutionary pressures may be placed on RNA structure more so than sequence conservation. Collectively, the observations presented here suggest a much more complex and vibrant RNA regulatory world may be operative in the regulation of gene expression.

The phosphatase and tensin homolog (PTEN) is a protein encoded on chromosome 10 by the PTEN gene and is a negative regulator of the oncogenic PI3K-protein kinase B (Akt) pathway. The PTEN gene is mutated and epigenetically inactivated in a diverse range of cancers (1). This gene is of particular interest, as emerging studies have shown that a pseudogene, PTENpg1, is actively involved in the regulation of PTEN. These observations suggest an underappreciation of the complexity involved in gene regulation. To date, thousands of pseudogenes have been identified in humans, including many disease-associated genes such as TP53 (2), BRCA1 (3), OCT4 (4–6), and PTEN (7). PTEN has a single pseudogene in the human genome, PTENpg1 (also called PTENp1, PTENΨ), which is encoded by chromosome 9 (8). PTENpg1 posttranscriptionally regulates PTEN expression by acting as a miRNA sponge to PTEN-targeting miRNAs (9). Recent studies have indicated the presence of antisense RNAs (asRNAs) derived from the PTENpg1 promoter locus (10). Several different isoforms of this antisense, named α and β, have been identified with transcription arising from the bidirectional PTENpg1 promoter, and one variant, PTENpg1 asRNA α, is found to modulate PTEN transcription via the recruitment of chromatin-modifying complexes EZH2 and DNMT3a (10). These proteins are actively recruited to the promoter by PTENpg1 asRNA α and cause chromatin condensation, and subsequently a reduction in PTEN expression (10).

It is noteworthy that PTEN and PTENpg1 are localized on different chromosomes, and the putative in trans acting mechanism by which the PTENpg1 asRNA α interacts with the PTEN promoter has not been determined. We find here that PTENpg1 asRNA α targeting DNMT3a to the PTEN promoter requires transcription of PTEN, specifically at the 5′UTR region containing homology to the PTENpg1 asRNA α transcript, and that this RNA can target the PTEN promoter in the absence of DNMT3a.

Results

Detection and Function of 5′ UTR PTEN Promoter Transcripts.

Previous studies with small noncoding RNAs (ncRNAs) demonstrated that an expressed low-copy transcript spanning the 5′ UTR of protein coding genes, designated a promoter-associated RNA (paRNA), was required for small ncRNA-directed epigenetic regulation in human cells (11–13). Transcriptomic data suggest there are several expressed sequence tags (ESTs) spanning the PTEN promoter [including BG772190 (Fig. 1A and Table S1) and DA005202, DA676942, and CN413383]. To detect the presence of these transcripts at the PTEN promoter, and to examine to what extent that they may play a role in PTENpg1 asRNA α regulation of PTEN transcription, strand-specific directional RT-PCR was performed on HeLa total RNA, using the PTENproR1 reverse primer only (Fig. S1A). This assay allows for any RNA transcripts spanning the PTEN promoter in a sense orientation to be discerned. Notably, a product was observed spanning this region of PTEN (Fig. 1B). Unfortunately, it is virtually impossible to disentangle to what extent this PTEN promoter-spanning transcript (Fig. 1B) is either a unique low-copy transcript or the 5′ UTR of PTEN mRNAs (Fig. 1A). Next, to interrogate to what extent the PTEN 5′ UTR-containing transcripts (paRNA/5′UTR) are involved in PTENpg1 asRNA α regulation of PTEN, suppression of this transcript was carried out using single-stranded antisense phosphorothioate oligodeoxynucleotides (ODNs) (14) (Fig. S1B). ODNs allow strand-specific interactions to be targeted and blocked, thereby disrupting only the sense-stranded paRNA/5′UTR for PTEN (10–12). We observed here that targeting the paRNA/5′UTR with ODN2 resulted in suppression of the PTEN promoter-associated 5′ UTR transcripts (Fig. 1C), but had little to no effect on downstream PTEN mRNA expression (Fig. 1D), suggesting there may be unique 5′ UTR-associated transcripts overlapping the PTEN promoter, similar to previous observations with small RNA-targeted transcriptional regulatory mechanisms (11, 12).

Fig. 1.

The PTEN promoter transcriptional landscape. (A) A screen shot from the University of California, Santa Cruz (UCSC) genome browser depicting several ESTs at the PTEN promoter/5′ UTR. EST BG772190 is also shown, which directly overlaps the PTENpg1 asRNA exon 1 target site (10). (B) Detection of a paRNA (e.g., EST BG772190) by directional RT-PCR were run on a 2% agarose gel. (Left) A 100-bp ladder. PCR products are HeLa directional RT (diRT), HeLa no primer (noPrimer), and No template RT (noRT). (C and D) The effects of antisense phosphorothioate oligodeoxynucleotide targeting of the PTEN pRNA (EST BG772190). (C) Expression of the PTEN 5′UTR (e.g., paRNA/EST BG772190) and (D) PTEN mRNA 48 h posttransfection with ODNs. (E) Biotin ODN2 binds predominantly PTEN 5′UTR-containing transcripts and not the PTEN gene. Biotin ODN2 IP was performed on 293HEK cells, and RNase A vs. RNase H sensitivity was determined. (F) Effects of ODN treatment on DNMT3a localization to the PTEN promoter, as determined by ChIP 48 h posttransfection with ODNs. (G) PTEN promoter CpG methylation, as determined by methyl-cytosine-specific restriction enzyme McrBc treatment and qPCR in ODN2 and ODN2 Scrambled Control-treated 293HEK cells 24 and 48 h posttransfection. The averages of triplicate-treated cultures are shown, and the P value from a paired t test, a single representative experiment, is shown. (H) The effects of ODN treatment on expression of unspliced variants of PTEN in stable 293HEK cells that overexpress PTENpg1 asRNA exon1 (PG1 cells). PTEN RNA expression was determined by qRT-PCR with PTEN F3/R3 primers, which specifically detect unspliced forms of PTEN. For C, D, F, and H, triplicate-treated samples are shown as a fraction of the control ODN ± SE of mean (SEM) after normalization to RPL10, and for E, duplicates with the ranges shown from a single representative experiment. P values from a paired two-tailed t test are also shown.

Table S1.

Reported EST BG772190 found on chr10:89623172–89623881

Construct name

Long noncoding RNA genetic sequence

EST BG772190

5′TACCGCCCCCTGCCCTGCCCTGCCCTCCCCTCGCCCGGCGCGGTCCCGTCCGCCTCTCGCTCGCCTCCCGCCTCCCCTCGGTCTTCCGAGGCGCCCGGGCTCCCGGCGCGGCGGCGGAGGGGGCGGGCAGGCCGGCGGGCGGTGATGTGGCGGGACTCTTTATGCGCTGCGGCAGGATACGCGCTCGGCGCTGGGACGCGACTGCGCTCAGTTCTCTCCTCTCGGAAGCTGCAGCCATGATGGAAGTTTGAGAGTTGAGCCGCTGTGAGGCGAGGCCGGGCTCAGGCGAGGGAGATGAGAGACGGCGGCGGCCGCGGCCCGGAGCCCCTCTCAGCGCCTGTGAGCAGCCGCGGGGGCAGCGCCCTCGGGGAGCCGGCCGGCCTGCGGCGGCGGCAGCGGCGGCGTTTCTCGCCTCCTCTTCGTCTTTTCTAACCGTGCAGCCTCTTCCTCGGCTTCTCCTGAAAGGGAAGGTGGAAGCCGTGGGCTCGGGCGGGAGCCGGCTGAGGCGCGGCGGCGGCGGCGGCACCTCCCGCTCCTGGAGCGGaGGGGAGAAGCGGCGGCGGCGGCGGCCGCGGCtGGCTGCAGCTCCAGGGAGGGGGTCTGAGTCGCCTGTCACCATTTCCAGGGCTGGGAACGCCGGAGAGTcGGTCTCTCCCCTTCTACTGnCTCCAACACGGCCGGCGGCTGGCggcggcaACATCCAGGGACCCGGG3′

Fig. S1.

Location of directional RT-PCR primers and antisense ODNs relative to the PTEN promoter. (A) A schematic of the PTEN promoter is shown depicting the PTENpg1 asRNA exon1 binding site and location of antisense ODNs, as well as RT-PCR primers used in directional RT analysis of the PTEN-associated paRNA. (B) The sequence of the ODNs and control ODN (miRN367, ref. 1) used in the presented study.

To distinguish whether ODN2 was targeting the PTEN gene or a unique 5′ UTR-associated transcript, we used biotin-labeled ODN2 to immunoprecipitate (IP) ODN2-bound nucleic acids. By using RNase A and RNase H treatment, we determined that ODN2 binds a PTEN 5′ UTR-associated transcript, which interestingly extends the full length of the PTEN mRNA (Fig. 1E), similar to previous observations involved in the mechanism of small noncoding RNA transcriptional gene silencing (11, 12). Notably, although ODN2 also appeared to bind to the PTENpg1 transcript (Fig. 1E), this interaction did not appear to affect PTENpg1 sense and antisense expression (Fig. S2), suggesting ODN2 interacts with a full-length 5′ UTR-containing PTEN transcript, which we refer to as paRNA/5′UTR.

Fig. S2.

PTENpg1 and PTENpg1 asRNA expression after ODN2 treatment. The 293HEK cells were transfected in triplicate with ODN2, control ODN miRN37 target site (1, 2). The averages of triplicate-treated cultures are shown with the SDs.

The pathway of PTENpg1 regulation of PTEN involves the recruitment of DNMT3a by exon 1 of the PTENpg1 asRNA α transcript to the PTEN promoter (10). We find here that ODN2 targeting of the PTEN paRNA/5′UTR results in a loss of DNMT3a at the PTEN promoter (Fig. 1F) and a reduction in CpG methylation at one locus in the PTEN promoter (Fig. 1G). This effect of ODN2 on PTEN appeared to be the result of ODN2 blocking the activity of the PTENpg1 asRNA α exon 1. Blocking this transcript results in preventing the transcriptional regulation of the PTEN promoter by PTENpg1 asRNA α, as ODN2 treatment in stable PTENpg1 asRNA α exon 1 overexpressing cells demonstrated increased expression of unspliced forms of PTEN (Fig. 1H). Collectively, these observations, along with others (10), suggest PTEN promoter-associated 5′ UTR-containing transcripts are required for PTENpg1 asRNA α- and DNMT3a-based epigenetic regulation of PTEN.

PTEN paRNA/5′UTR Interacts Directly with PTENpg1 asRNA α Exon 1.

Previous studies with small antisense RNAs targeted to gene promoters have demonstrated that the small antisense RNAs interact directly with a transcript at the promoter (11, 12). The observations presented here (Fig. 1), along with previous studies (10), suggest the long noncoding RNA (lncRNA) PTENpg1 asRNA α regulates PTEN transcription by interactions with a sense-stranded transcript or elongated 5′UTR (paRNA/5′UTR) that essentially spans the PTEN promoter. To interrogate whether the PTEN promoter-associated transcript interacts directly with PTENpg1 asRNA α and DNMT3a, an immunoprecipitation of DNMT3a was carried out, followed by qRT-PCR with primers specific to each transcript (Fig. 2A). We find, using this technique, that DNMT3a interacts with both PTENpg1 asRNA alpha exon 1 and a PTEN paRNA/5′UTR promoter-associated transcript (Fig. 2 A and B). To interrogate whether the PTEN promoter-associated transcript interacts directly with PTENpg1 asRNA α, biotin-labeled PTENpg1 asRNA α exon1 was generated and cotransfected with ODN2 into 293HEK cells, and the effects on biotin-labeled PTENpg1 asRNA α exon1 binding to the PTEN promoter were determined by RNA immunoprecipitation (RIP; ref. 4 and Fig. 2 D–I). ODN2 treatment, relative to controls, abrogated the binding of PTENpg1 asRNA α exon 1 to the PTEN promoter (Fig. 2 D–F). To more clearly map the interacting domain of the PTEN 5′ UTR promoter-associated transcript (paRNA/5′UTR) and PTENpg1 asRNA α, RNase A treatment was carried out on the ODN2 and biotin PTENpg1 asRNA α exon 1 cotransfected 293HEK cells. Of those regions bound to the PTEN promoter in the presence of ODN2, one particular region spanning the PTENpg1 asRNA α exon 1 homologous target region and ∼150 nucleotides upstream in the PTEN promoter (Fig. 2G) appeared to exhibit the most abundant RNA:RNA interactions, based on RNase A sensitivity, relative to the other loci assessed (Fig. 2 H and I), suggesting this region (Fig. 2C; PTENproF1/R1) may be where the strongest localization of those RNA:RNA interactions required for PTENpg1 asRNA α regulation of PTEN transcription occur.

Fig. 2.

The effects of ODN2 treatment on PTENpg1 asRNA exon1 binding to and regulation of PTEN. (A) A schematic is shown depicting the predicted PTEN paRNA and PTENpg1 asRNA exon 1 interacting region relative to PTEN, along with the RNA:RNA interacting domain and presumed DNMT3a interaction (10). (B) ChIP was carried out for DNMT3a in 293HEK cells, and those transcripts associated with DNMT3a were determined by semiqRT-PCR. DNMT3a was observed to interact with both PTENpg1 asRNA (asF3R0) and paRNA/EST BG772190 (PTENpro F2R2) loci, but not the 3′ UTR of PTEN. The PTENpro F2/R2 primer set is specific for PTEN-5′UTR paRNAs (e.g., EST BG772190), whereas PTENpg1 asF3R0 is specific for PTEN pg1-encoded transcripts. A primer set targeting the PTEN 3′UTR is provided as a control. (C) A UCSC genome browser screenshot of the PTENpg1 antisense RNA exon 1 target site in the PTEN promoter is shown along with various primers used to interrogate PTENpg1 asRNA exon 1 interactions with the PTEN promoter. (D–I) RIP analysis of biotin PTENpg1 asRNA α interactions with the PTEN promoter in the presence of ODN2. (D–F) ODN2 treatment blocks biotin dUTP-containing PTENpg1 asRNA exon1 binding to the PTEN promoter. Biotin-labeled PTENpg1 asRNA exon 1 was cotransfected with ODN2 into 293HEK cells, and localization to the PTEN promoter determined by ChIP. (D) Primer set PTENproF1/R1, (E) PTENpro F2/R2, and (F) PTENproF2/R1 were used to detect the loss of PTENpg1 asRNA α exon 1 binding after ODN2 treatment to various loci in the PTEN promoter. (G–I) Mapping of RNA:RNA interacting loci at the PTEN promoter. ODN2 treatment results in reduced PTENpg1 asRNA exon1 interactions with PTEN paRNA/EST BG772190 at the (G) PTENproF1/R1 and (H) PTENproF2/R2 loci, but not the (I) PTENproF2/R1 locus. For D–I, the averages of triplicate-treated cultures are shown with the SE of mean and P values from a paired two-tailed t test.

Previous observations have indicated PTENpg1 asRNA α exon 1 binds and directs DNMT3a to the PTEN promoter (10). To determine the parameters and particular region in PTENpg1 asRNA α exon 1 involved in binding to the PTEN promoter, various truncations of PTENpg1 asRNA α exon 1 were generated as biotin-labeled transcripts (Fig. S3 and Table S2). These truncated PTENpg1 asRNA α exon 1 variants (Fig. S3A and Table S2) were transfected into 293HEK cells, and localization to the PTEN promoter determined by chromatin immunoprecipitation (ChIP). Two truncated fragments appeared to localize to the PTEN promoter, F4R1 and F5R2, as well as the control full-length PTENpg1 asRNA α exon 1 (Fig. 3A and Fig. S3). However, when these deletion constructs were assessed for their ability to direct DNMT3a to the PTEN promoter, only the full-length PTENpg1 asRNA α exon 1 and F4R1 variant was functionally capable of directing DNMT3a to the PTEN promoter (Fig. 3B). Interestingly, the F4R1 truncated variant was able to repress PTEN mRNA expression (Fig. 3C), similar to previous observations with PTENpg1 asRNA exon 1 (10), whereas the F5R2 variant resulted in a dose-dependent increase in PTEN expression (Fig. 3D). Indeed, both PTENpg1 asRNA α exon 1 and F4R1 were found to interact directly with DNMT3a/DNMTL-CD in vitro (Fig. 3 E and F and Figs. S4 and S5), relative to the control GFP RNA (Fig. 3H), as observed in electrophoretic mobility shift assays (EMSA). Interestingly, the F5R2 variant was also found to bind DNMT3a/DNMTL-CD in vitro (Fig. 3G), but, unlike the full-length PTENpg1 asRNA α exon 1 or F4R1 variants, was unable to direct DNMT3a to the PTEN promoter (Fig. 3B). This is an interesting observation, as the F5R2 variant appears to bind the PTEN promoter (Fig. 3A) or DNMT3a (Fig. 3G), but not both (Fig. 3B), suggesting, based on this single observation, that some antisense lncRNAs may target genes in the absence of DNMT3a or bind DNMT3a and block endogenous recruitment to their intended target. In the case presented here with the PTEN locus, the dose-dependent overexpression of the F5R2 variant appeared to result in active increases in PTEN expression (Fig. 3D).

Fig. S3.

Analysis of the PTENpg1 asRNA alpha exon 1 and various truncations of PTENpg1 asRNA exon1 target sites in the PTEN promoter. (A) A screen shot from UCSC genome browser is shown depicting the PTEN promoter and homologous target sites for PTENpg1 asRNA exon1 and the various truncations of PTENpg1 asRNAs assessed. (B–D) The predicted DNMT3a binding bulge in both (B) full-length and (C) F4R1 variant of PTENpg1 asRNA exon 1, and lost in (D) F5R2 variant, as determined from M-Fold. The conserved sequence that appears required for DNMT3a interactions with PTENpg1 asRNA exon 1 at the PTEN promoter (ACAUUCAAAC UUCCAUCACGGC) is shown in PTENpg1 asRNA exon 1 and F4R1, but lost in F5R2.

Table S2.

Truncated PTEN alpha exon 1 variants

Construct name	Long noncoding RNA genetic sequence
PTENpg1 antisense alpha exon 1 (described in ref. 3)	5′GGAAGAGGCTGCACAGTTAGAAAAGACGAAGAAGAAGCGAGAAACGCCGCCGCTGCCGGCGCCGCCCCCGCGGATGCTCACGGGTTGCTGAGAGGGGCTTCAGGCCGGGCCGGGCCGGGCCGCCGCCGCCGCCGCCGTCTCTGTCTCTCATCTCCCTCGCCTGAGCCCGGCCTCGCCTCACAGCGGCTCAACATTCAAACTTCCATCACGGCTGCAGCTTCTGAGAGGAGAGCATCTTCTAG3′
F2R1 PCR set 2 (variant 1 PTENpg1asRNA alpha, 220 bp fragment)	5′CCGGCGCCGCCCCCGCGGATGCTCACGGGTTGCTGAGAGGGGCTTCAGGCCGGGCCGGGCCGGGCCGCCGCCGCCGCCGCCGTCTCTGTCTCTCATCTCCCTCGCCTGAGCCCGGCCTCGCCTCACAGCGGCTCAACATTCAAACTTCCATCACGGCTGCAGCTTCTGAGAGGAGAGCATCTTCTAGTTTTT3′
F3R1 PCR set 3 (variant 2 PTENpg1asRNA alpha, 164 bp fragment)	5′GGGCCGGGCCGCCGCCGCCGCCGCCGTCTCTGTCTCTCATCTCCCTCGCCTGAGCCCGGCCTCGCCTCACAGCGGCTCAACATTCAAACTTCCATCACGGCTGCAGCTTCTGAGAGGAGAGCATCTTCTAGTTTTT3′
F4R1 PCR set 4 (variant 2 PTENpg1asRNA alpha, 107 bp fragment)	5′GGGTAGTGAAGGCTAGCGGGCCTCGCCTCACAGCGGCTCAACATTCAAACTTCCATCACGGCTGCAGCTTCTGAGAGGAGAGCATCTTCTAGTTTTTGGTACCTATCTGA3′
F5R2 PCR set 4 (variant 2 PTENpg1asRNA alpha, 217 bp fragment)	5′TAGTGAAGGCTAGCGCGCCGCCGCTGCCGGCGCCGCCCCCGCGGATGCTCACGGGTTGCTGAGAGGGGCTTCAGGCCGGGCCGGGCCGGGCCGCCGCCGCCGCCGCCGTCTCTGTCTCTCATCTCCCTCGCCTGAGCCCGGCCTCGCCTCACAGCGGCTCAACTCTCAAACTTCCATCATGGCTGCAGCTTCCGAGAGGTTTTTGGTACCTATCTGA3′

The experimentally determined 3a binding bulge is underlined in the F4R1 and full-length PTENpg1 alpha exon 1 and the AT insertion in bold that disrupts the 3a binding bulge for the F5R2 variant. Underlined segments represent predicted loop promoter-associated RNA-binding loci.

Fig. 3.

Characterization of PTENpg1 asRNA a exon 1 and truncated variants involvement in the recruitment of DNMT3a to the PTEN promoter. (A) Enrichment of biotin-labeled PTENpg1 asRNA alpha full-length and truncated variants by RIP at the PTEN promoter (F2/R2 locus) after subtraction of beads alone and standardized to inputs. (B) Ability of PTENpg1 asRNA exon 1 and truncated variants to direct DNMT3a to the PTEN promoter, as determined by ChIP. (C) The PTEN F4R1 variant suppresses PTEN mRNA expression. Either the control pcDNA3.1-GFP or pcDNA-F4R1 were transfected in triplicate into 293HEK cells and PTEN mRNA expression determined 72 h later by qRT-PCR. (D) The PTEN F5R2 variant activates PTEN expression based on qRT-PCR analysis of transcript expression. Dose-dependent transfection of F5R2 activation of PTEN mRNA expression was observed in 293HEK cells transfected in triplicate with increasing concentrations of either pcDNA3.1-F5R2 or control pcDNA3.1-GFP. (E–H) EMSA analysis of (E) PTENpg1 asRNA exon 1, (F) truncated F4R1, (G) truncated F5R2, and (H) Control GFP binding to DNMT3a in the presence of DNMTCD. The reconstituted system Dnmt3a/DnmtL-CD in vitro requires the presence of DNMTL-CD, which alone is unable to bind RNA (even at 50× molecular excess), indicating that the observed RNA binding in the DNMT3a/DNMTL-CD complex comes from DNMT3a. For A and B, the averages of triplicate transfected 293HEK cells from a single representative experiment are shown with the SEM and P values from a two-sided t test. *P < 0.05; **P < 0.01. For D (**), a one-way ANOVA was carried out with the f-ratio value of 10.38725; the P value is 0.00913.

Fig. S4.

EMSA analysis of PTENpg1 asRNA truncated F4R1 binding to DNMT3a in the presence of DNMTL-CD. Protein concentration is indicated above, and the free RNA and bound RNA-DNMT3a are indicated.

Fig. S5.

Analysis of the bound fraction of EMSA for PTENpg1 asRNA exon 1, F4R1 and F5R2 binding to Dnmt3a in the presence of DnmtL-CD. Bands were quantified using ImageJ and plotted in a log scale with a curve fit using Prism 7.

To interrogate this notion further and determine the requirement of an observed major loop (5′-ACAUUCAAACUUCCAUCACGGC-3′) that was found in both the PTEN pg1 asRNA exon 1 (Fig. S3B) and F4R1 variant (Fig. 4A and Fig. S3C), but not in the F5R2 variant (Fig. S3D), in binding to the PTEN promoter, we generated several truncations of F4R1 (Fig. 4A and Table S3) and determined their respective ability, along with the full-length F4R1 control, to bind the PTEN promoter in the absence of DNMT3a. Interestingly, only the controls full-length F4R1 and PTENpg1 asRNA exon 1 were able to bind the PTEN promoter in the absence of DNMT3a (Fig. 4B), suggesting the major loop observed in both the PTENpg1 asRNA exon 1 and the F4R1 variant is required for PTENpg1 asRNA targeting of the PTEN promoter in the absence of DNMT3a (Fig. 4 C and D).

Fig. 4.

Characterization of the PTENpg1 asRNA F4R1 variant functional domain. (A) Various truncations of F4R1 were developed and cloned into the pcDNA3.1 expression vector and screened for binding to the PTEN promoter. M-Fold analysis of the predicted RNAs is shown. (B) Biotin-labeled PTENpg1 asRNA exon1, F4R1, and F4R1 variants C1-F4R1 PTENpg1 asRNA Δ major loop, C2-F4R1 PTENpg1 asRNA Δ lower stem, and C3-F4R1 PTENpg1 as RNA Δ upper stem and major loop binding to the PTEN promoter in the DNMT3a deleted HCT116 cultures. The average of triplicate-treated cultures are shown with P values from a one-way ANOVA (nonparametric) analysis. ***P < 0.0001; **P < 0.003269. (C) PTENpg1 asRNA exon 1 and F4R1 major loop interacting locus in the PTEN paRNA (EST BG772190)/5′ UTR of PTEN). (D) A screen shot from the UCSC genome browser showing the PTENproF1/R1 region in which PTENpg1 asRNA exon 1 and F4R1 interact with EST BG772190. The major loop interacting region is also shown.

Table S3.

The F4R1 functional domains

F4R1 variants	Long noncoding RNA genetic sequence
F4R1	5′GGGTAGTGAAGGCTAGCGGGCCTCGCCTCACAGCGGCTCAACATTCAAACTTCCATCACGGCTGCAGCTTCTGAGAGGAGAGCATCTTCTAGTTTTTGGTACCTATCTGA3′
C1 Construct	5′TAGTGAAGGCTAGCGGGCCTCGCCTCACAGCGGCTCACGAATGCAGCTTCTGAGAGGAGAGCATCTTCTAGTTTTTGGTACCTATCTGA3′
C2 Construct	5′CTCGCCTCACAGCGGCTCAACATTCAAACTTCCATCACGGCTGCAGCTTCTGAGAGGAGAGC3′
C3 Construct	5′GGGTAGTGAAGGCTAGCGGGCGAATCTTCTAGTTTTTGGTACCTATCTGA3′

Several variations of the F4R1 transcript were derived. Deletion construct C1 contains a change to the major loop of F4R1, resulting in the replacement of the major loop with another loop sequence. Construct C2 contains a deletion of ”lower” stem and construct C3 contains a deletion of the “upper” stem. Underlined segments represent predicted loop promoter-associated RNA-binding loci.

Discussion

The observations presented here suggest PTEN 5′ UTR promoter-associated transcripts are involved in PTENpg1 asRNA exon 1-directed epigenetic regulation of PTEN. Previous studies have detected promoter-associated transcripts (11, 15), which are required for small antisense RNAs to guide DNMT3a and direct transcriptional gene silencing in human cells (11–13). These promoter-associated transcripts are thought to be low abundant mRNAs that contain elongated 5′ UTRs that can be detected through directional RT-PCR (11, 12) and public deep sequencing (Fig. 1A and Table S1). The data presented here juxtaposed with previous observations suggest that a mechanism of action is active in human cells whereby RNA:RNA interactions occur at chromatin to facilitate the recruitment of epigenetic regulatory protein complexes (16). Building on observations that PTENpg1 antisense RNA α exon 1 is an active transcriptional and epigenetic modulator of PTEN (10), we find here that the localization of PTENpg1 asRNA α exon 1 and a truncated variant F4R1 to the PTEN promoter requires a PTEN promoter-associated RNA (paRNA/5′UTR) and involves a conserved major loop domain (5′-ACAUUCAAACUUCCAUCACGGC-3′) to successfully direct DNMT3a to the PTEN promoter (Fig. 4). Interestingly, through deletion studies, one sequence appears to be the main modulator involved in the ability of PTENpg1 asRNA exon1 or F4R1 to target DNMT3a to the PTEN promoter, which was not retained in the defective F5R2 variant. This sequence (Fig. 4C) appears in both the PTENpg1 asRNA exon 1 and F4R1 variants and maps directly to a region that was observed previously to exhibit high levels of DNMT3a and PTENpg1 asRNA exon 1 binding (10) (Fig. 4D).

The observations presented here suggest that an RNA:RNA interaction is involved in PTENpg1 asRNA exon1 targeting of the PTEN promoter and that the conserved domain required for this interaction consists of the major loop domain (5′-ACAUUCAAACUUCCAUCACGGC-3′) interacting with the PTEN paRNA/5′UTR (Fig. 4 C and D). This loop appears to be required for localization of the PTENpg1 asRNA exon 1 and F4R1 variant transcripts to the PTEN promoter, along with the entire stem present in the F4R1. When this loop is altered, as is the case with the F5R2 variant, there appears to be a loss of localization of the transcript to the PTEN promoter, whereas an ability to interact with DNMT3a remained intact. Collectively, the dichotomous observations presented here among PTENpg1 asRNA exon1, F4R1, and the F5R2 variants suggest the major loop domain (Fig. 4 A and C and Fig. S3 B and C), in combination with a longer stem found in both PTENpg1 alpha asRNA exon 1 and the F4R1 variant, interacts directly with DNMT3a to direct transcriptional and epigenetic silencing of PTEN. Such observations may support the notion that lncRNAs and their putative evolutionary conservation may be more contingent on a combination of both structure and sequence (17). Collectively, the observations presented here expand our understanding of endogenous lncRNA networks in human cells and suggest RNA:RNA interactions, particularly at gene promoters, may be mechanistically relevant in lncRNA regulation of protein-coding gene expression. An understanding of this emerging mode of gene and epigenetic regulation could prove useful in the development of targeted therapeutics to disrupt or augment transcriptional regulatory networks in humans.

Materials and Methods

Biotin ODN2 Immunoprecipitation.

As described (12, 18), 5′ biotin-labeled ODN2 or scrambled control (Table S4) was transfected into 293HEK cells. Cells were harvested 48 h posttransfection and permeabilized in 0.25% Triton-X/PBS and then washed in PBS. Cells were treated with RNase A, RNase H, or no RNase control at 37 °C for 15 min. RNase inhibitors were added, and cells were cross-linked with formaldehyde. Cells were resuspended in lysis buffer (1% SDS, 50 mM Tris⋅HCl at pH 8, 10 mM EDTA, RNase inhibitors) and sonicated. Immunoprecipitation was performed using Dynabeads MyOne Streptavidin Beads (Thermo Fisher). Beads were incubated with the samples for 30 min and then washed five times in lysis buffer. Samples were eluted at 95 °C for 15 min in elution buffer (1% SDS, 10 mM EDTA, 0.1 mM NAHCO₃ at pH 8). Samples were then divided and treated with either RNase or DNase. Enrichment of DNA in RNase-treated samples was determined by qPCR, using the following primer sets: PTENP1_F/R, paRNA_F/R, PTENex1_F/R, PTENex9_F/R (Table S4). The RNA in DNase-treated samples was quantified by RT-qPCR, using the same primer sets as earlier. Scramble control was subtracted from the immunoprecipitate, and data were normalized to the sample input.

Table S4.

Primers and ODN sequences used in analysis (5′-3′ shown)

Name	Long noncoding RNA genetic sequence
PTENproF1	GCTGCAGCCATGATGGAAGTTTGA
PTENproR1	AAAGACGAAGAGGAGGCGAGAAAC
PTENproF2	TGATGTGGCGGGACTCTTTATGC
PTENproR2	TCACAGCGGCTCAACTCTCAAACT
PTENF3	AGAAAGCTTACAGTTGGGCCCTGT
PTENR3	GCCACAGCAAAGAATGGTGATGCT
PTENP1ex1_F	GGAAGAGGCTGCACAGTTA
PTENP1ex1_R	CTAGAAGATGCTCTCCTCTCA
paRNA_F	ATGTGGCGGGACTCTTTATG
paRNA_R	GCGGCTCAACTCTCAAACT
PTENex1_F	TGCCATCTCTCTCCTCCTT
PTENex1_R	CGAATCCATCCTCTTGATATCTCC
PTENex9_F	TGTAATCAAGGCCAGTGCTAAA
PTENex9_R	AGCATCCACAGCAGGTATTATG
PTEN unspliced F	AAAGCTGGAAAGGGACGAACTGGT
PTEN unspliced R	TCTCAGATCCAGGAAGAGGAAAGG
mirN367 (control) ODN	GTGTGGGGTTTTAGCTTCGTGAA
ODN1	GTCTCTCATCTCCCTCGCCT
ODN2	GCTTCCACCTTCCCTTTCAG
Scrambled ODN2	CTAACTCTCCGTGTCCTCCT
pcDNA3.1 F	CCCACTGCTTACTGGCTTATC
pcDNA3.1 R	CAGATGGCTGGCAACTAGAA
Killin F	ACACAAGCACCCACATCCAAA
Killin R	AGTCCTTTGGCTTGCTCTTAG

PTENpg1asRNA Exon1 and Truncated Variant Transcription.

PTENpg1asRNA exon one and truncated variants (Table S2) were in vitro transcribed from linearized pcDNA 3.1 plasmids (containing the various inserts), using T7 RNA polymerase (19). RNA constructs were purified from transcription components by denaturing gel electrophoresis.

Electrophoretic Mobility Shift Assays.

Protein lncRNA binding reactions were performed in a final volume of 35 μL and contained 637, 500, 700, or 600 ng PTENpg1asRNA exon1, F4R1, F5R2, or GFP RNA, respectively. All lncRNA variants were refolded by heating at 90 °C for 80 s before snap cooling on ice. Refolding was performed in a buffer containing 50 mM Hepes at pH 7.5, 150 mM KCl, 1.5 mH MgCl₂ 1 mM TCEP, and 20% glycerol for all lncRNAs. After refolding, protein was incubated with lncRNA for 30 min on ice. Dnmt3a/DnmtL-CD binding reactions contained between 1× and 10× molecular excess of the protein to RNA. DnmtL-CD binding reactions between 5× and 50× the molecular excess of the protein to RNA. Samples were loaded onto a 4.5% TBE acrylamide gel (containing 2.5 mM MgCl₂) and run at 250 V for 4 h at 4 °C. Gels were stained with SYBR gold (Invitrogen) and visualized using a Typhoon FLA 900 biomolecular imager.

SI Materials and Methods

ODN Transfections.

Control (mirN367), ODN1, and ODN2 were transfected into cells to a final concentration of 100 nM using Lipofectamine 2000 (Life Technologies). ODNs were transfected into both 293HEK and the PTENpg1 asRNA exon1 overexpressing 293-HEK PG1 cells.

qRT-PCR.

qRT-PCR analysis was performed using KAPA SYBR FAST qPCR Master Mix (KAPA Biosystems). Plate was placed in a ViiA 7 Real-Time PCR system (Life Technologies). Cycling conditions: 95 °C for 2 min and then 40 cycles of 95 °C for 3 s and 60 °C for 30 s.

Directional RT Analysis of Gene Expression.

Directional reverse transcription (directional RT) was performed on HeLa total RNA, using the PTENproR1 primer (Table S4) to strand-specifically convert any RNA transcripts originating from the PTEN 5′ UTR into ssDNA. The resulting product was PCR amplified using primer set PTENproF2/R2 and run on a 2% agarose gel (Fig. 1B).

ChIP.

ChIP analysis was carried out in 293HEK cells using anti-DNMT3a (Abcam, cat. no. ab2850). The ChIP was performed 48 h posttransfection with ODNs (Fig. 2 D–I) or various biotin-labeled truncations of the full-length PTENpg1 alpha exon 1 transcript (Fig. 3A) following previously described techniques (4–6). The relative enrichment of DNMT3a was determined at the PTEN promoter using primer sets PTENproF2/R2 (Fig. 1F and Table S4) or PTENproF1/R1 (Fig. 3B and Table S4). Any IgG or no antibody values are first subtracted from the resultant IP and input values, and then each sample is standardized relative to the sample input.

Dnmt3a/DnmtL-CD Purification.

Dnmt3a (residues 284–910) and DnmtL-CD (residues 178–379) were cloned into a modified petDuet vector. Dnmt3a/DnmtL-CD proteins were coexpressed in Escherichia coli strain Rosetta 2 (DE3). The transformants were grown at 37 °C in LB medium and induced at an OD₆₀₀ of 0.6 with IPTG and further incubated for 20–24 h at 18 °C. Dnmt3a/DnmtL-CD was purified from the supernatant of the cell lysate by three-step liquid chromatography. Nickel affinity, heparin affinity, and gel filtration chromatography were used and the purified protein complex stored in 50 mM Hepes at pH 7.5, 150 mM NaCl, 10% glycerol, and 1 mM TCEP. The purified proteins were estimated to be >90% pure by Coomassie blue-stained SDS/PAGE and were concentrated to 2–3 mg/mL for electrophoretic mobility shift assays. DnmtL-CD (residues 178–379) was cloned into pLIC-HK vector and overexpressed using the same methods as Dnmt3a/DnmtL-CD. DnmtL-CD was purified from cell lysate using Ni2+ affinity and gel filtration chromatography and stored in 50 mM Hepes at pH 7.5, 150 mM NaCl, 10% glycerol, and 1 mM TCEP. Purified DnmtL-CD was estimated to by >95% pure and was concentrated to 9–10 mg/mL.

T7-Transcribed Synthetic RNA Pulldown in Presence of ODNs.

Synthetic biotinylated ncRNAs were generated by T7 transcription using the Ampliscribe T7-Flash Biotin –RNA Transcription Kit (Epicentre Biotechnologies) according to the manufacturer’s instruction. Templates for T7 transcription were prepared by PCR of pcDNA3.1 plasmids expressing the relevant ncRNAs. Primers pcDNA3.1 F/R were used for PCR amplification (Table S4). PTENpg1 alpha exon 1 was in vitro transcribed using Durascribe, with biotin-linked dCTPs (described in detail in ref. 5). The resultant biotin PTENpg1 α exon 1 transcript was transfected into 293HEK cells at a concentration of 100 nM 18 h after transfection of either control ODN mirN367 or ODN2 transfection (SI Materials and Methods). Thirty hours after transfection of biotin conjugated transcripts, cells were cross-linked with formaldehyde at 1% final concentration for 10 min at room temperature followed by the addition of glycine to a final concentration of 0.125 M and a further incubation for 10 min at room temperature. Cells were then lysed with ChIP lysis buffer (5 mM Pipes, 85 mM KCl, and 0.5% Nonidet P-40) supplemented with PMSF on ice for 20 min. Chromatin was sheared by sonication. Cell lysates containing sheared chromatin were incubated with Dynabeads M280 Streptavidin (Life Technologies) prepared according to the manufacturer’s instructions for 1 h on a rotating platform. Beads were pulled down with a magnet for 3 min and washed with low salt immune complex wash buffer (0.1% SDS, 1% Triton X-100, 2 mM EDTA, 20 mM Tris⋅HCl at pH 8.1, 150 mM NaCl), high salt immune complex wash buffer (0.1% SDS, 1% Triton X-100, 2 mM EDTA, 20 mM Tris⋅HCl at pH 8.1, 500 mM NaCl), LiCl Immune complex wash buffer (0.25 M LiCl, 1% Nonidet P-40, 1% sodium deoxycholate, 1 mM EDTA, 10 mM Tris⋅HCl at pH 8.1), and TE buffer (10 mM Tris⋅HCl, 1 mM EDTA at pH 8.0). Each wash step was carried out for 3 min on a rotating platform. Streptavidin bead-biotinylated RNA–DNA complexes were resuspended in Elution buffer (1% SDS, 0.1 M NaHCO₃) and heated at 95 °C for 5 min to dissociate biotin RNA from beads. Eluted biotinylated RNA complexes were removed from beads via magnet pull-down, and supernatants were analyzed by qPCR for enrichment at the PTEN promoter. Samples were then DNase, RNase A, and proteinase K treated so as to remove everything from elutes except dsRNAs. These were reverse transcribed and analyzed by qPCR for PTEN promoter enrichment. Primer sets used for analyzing the PTEN promoter were PTENproF1/R1, PTENproF2/R2, and PTENproF2/R1 (Fig. 2C and Table S4).

Truncated PTEN Alpha Exon 1 and F4R1 Deletions ChIP.

Various truncated versions of PTEN alpha exon 1 (Table S2) and mutants of F4R1 (Table S3) were generated to be expressed from the CMV promoter in the context of pcDNA3.1 (Genewiz). The constructs (Tables S2 and S3) were in vitro transcribed using Durascribe with biotin-linked dCTPs (described in detail in ref. 5). The resultant biotin-labeled transcripts and a GFP-biotin control were transfected into 293HEK cells (50 nM) and assessed by RIP 48 h later using PTENproF2/R2 (Table S4) primers for detection at the PTEN promoter.

PTEN Promoter CpG Methylation Post-ODN2 Treatment.

Genomic DNA was isolated from Hek293 cells after 48 h treatment with ODN scrambled control and ODN2 (Table S4). Briefly, 200 ng DNA was digested with the methyl-cytosine-specific restriction enzyme McrBc (New England Biolabs) overnight at 37 °C. Twenty-four hours later, the enzyme was heat inactivated at 65 °C for 1 h and qRT-PCR performed (Kapa Biosystems), using Killin F/R primers (Table S4), and standardized to uncut input. The delta CT values were converted to fold-change values, and the ratio between ODN2/ODN scrambled control treated cells was calculated.