HIF1α-AS1 is a DNA:DNA:RNA triplex-forming lncRNA interacting with the HUSH complex

Abstract

DNA:DNA:RNA triplexes that are formed through Hoogsteen base-pairing of the RNA in the major groove of the DNA duplex have been observed in vitro, but the extent to which these interactions occur in cells and how they impact cellular functions remains elusive. Using a combination of bioinformatic techniques, RNA/DNA pulldown and biophysical studies, we set out to identify functionally important DNA:DNA:RNA triplex-forming long non-coding RNAs (lncRNA) in human endothelial cells. The lncRNA HIF1α-AS1 was retrieved as a top hit. Endogenous HIF1α-AS1 reduces the expression of numerous genes, including EPH Receptor A2 and Adrenomedullin through DNA:DNA:RNA triplex formation by acting as an adapter for the repressive human silencing hub complex (HUSH). Moreover, the oxygen-sensitive HIF1α-AS1 is down-regulated in pulmonary hypertension and loss-of-function approaches not only result in gene de-repression but also enhance angiogenic capacity. As exemplified here with HIF1α-AS1, DNA:DNA:RNA triplex formation is a functionally important mechanism of trans-acting gene expression control.

Fig. 1. HIF1α-AS1 is a triplex- and DNA-associated RNase H-insensitive lncRNA in endothelial cells.

a Overview of the identification of endothelial-expressed triplex-forming lncRNAs. LncRNAs from a previous Triplex-Seq study in HeLa S3 and U2OS were overlapped, filtered with high stringency and analyzed for nuclear expression in endothelial cells with ENCODE and FANTOM5 CAGE data followed by analyses for noncoding probability and enriched peaks in the Triplex-Seq data. b RNA-immunoprecipitation with anti-dsDNA antibody followed by qPCR (RIP-qPCR) targeting the lncRNA candidates in HUVEC. Samples were treated with or without RNase H. βAct served as control for RNase H-mediated degradation. n = 3. c Scheme of the human genomic locus of HIF1α-AS1. d RT-qPCR after anti-dsDNA-RIP in HUVEC. HIF1α and 18 S rRNA served as negative control. One-way ANOVA with Tukey’s test, n = 3. *(p = 0.0002). e RIP-qPCR with anti-histone3 (H3) in HUVEC. Data was normalized against GAPDH. Paired t test, n = 4. *(p = 0.0364). f RT-qPCR of HIF1α-AS1 in HUVEC treated with hypoxia (0.1% O₂) for the indicated time points. Normoxia served as negative control (CTL). n = 3, One-Way ANOVA with Bonferroni test. *6 h (p = 0.0216), * 24 h (p = 0.0035). g RT-qPCR of HIF1α-AS1 in HUVECs treated with hypoxia (0.1% O₂) followed by reoxygenation with normoxia (after 24 h of hypoxia) for the indicated time points. n = 6, paired t test. *Hypoxia-6h (p = 0.0001), *Hypoxia-24h (p = 0.002), *Reoxygenation-24h (p = 0.0189), *Reoxygenation-48h (p = 0.04). h RT-qPCR of HIF1α-AS1 in lungs from control donors (CTL, n = 6) or patients with idiopathic pulmonary arterial hypertension (IPAH, n = 6) or chronic thromboembolic pulmonary hypertension (CTEPH, n = 8). One-Way ANOVA with Tukey’s test. *IPAH (p = 0.0063), *CTEPH (p = 0.0005). For b, d–g, n is defined as number of independent experiments. Data are presented as mean values ± SEM.

Fig. 2. HIF1α-AS1 potentially forms DNA:DNA:RNA triplexes.

a Overview of the identification of HIF1α-AS1 triplex forming regions (TFR) and their DNA triplex target sites (TTS) with triplex domain finder (TDF). HIF1α-AS1 pre-RNA and ATAC-Seq of HUVECs treated with or without LNA GapmeRs against HIF1α-AS1 were used as input for TDF. RIP and LNA GapmeRs were used to validate the findings of TDF. b All three TFRs of HIF1α-AS1 have a significantly higher number of DNA triplex target regions (blue dots) than the random background (boxplots in gray). Boxplot visualizes the median, first and third quartiles. The whiskers present the 1.5 interquartile range. External gray dots represent outliers. Numbers in brackets are positions of the individual TFR within HIF1α-AS1 pre-RNA. Analyzed with TDF, n=200 randomizations. c Circos plot showing the localization of the individual TFR within HIF1α-AS1 pre-RNA and its interaction with the chromosomal TTS. d Overlap of TTS of the individual TFRs of HIF1α-AS1. e Identification of RNase H-resistant TFRs. RIP with anti-S9.6 with/without RNase H treatment in HUVEC followed by qPCR for the TFRs. Ratio of %-input recovery with/without RNase H treatment is shown. n = 8 independent experiments, paired t test. Dotted line represents normalized values without RNase H treatment. The asterisk indicates that the %-recovery after RNase H is significantly different for TFR2 compared to TFR1 (p = 0.0452) and TFR3 (p = 0.0244). f HIF1α-AS1 TFR1-3 overlapping top target genes, their genomic location and number of TTS identified by TDF. g RT-qPCR of triplex target genes of TFR2 after knockdown of HIF1α-AS1 in HUVEC. n = 6 independent experiments. One-Way ANOVA with Holm’s Sidak test. *ADM (p = <0.0001), *HIF1α-AS1 (p = <0.0001), *PLEC (p = 0.0238), *RP11-276H7.2 (p = 0.0238), *EPHA2 (p = 0.0238), *MIDN (p = 0.023). h Different triplex target regions of HIF1α-AS1 are shown. Triplex target regions are highlighted in gray, triplex target sites are shown in blue. Arrows indicate TTS at the 5ʹend. i ChIP-qPCR with S9.6 antibody with/without RNase H treatment. Dotted line represents normalized values without RNase H. QPCR was performed against EPHA2 or ADM TTS or regions up- and downstream of the individual TTS. One-Way ANOVA with Bonferonni test. n = 5 independent experiments. *EPHA2:up/TTS (p = <0.0001), *ADM:up/TTS (p = <0.0001), *ADM:TTS/down (p = <0.0001), *EPHA2:TTS/down (p = 0.0321). Data are presented as mean values ± SEM.

Fig. 3. HIF1α-AS1 TFR2 RNA forms in vitro DNA:DNA:RNA triplexes with the predicted DNA target region in EPHA2.

a Electromobility shift assay of EPHA2 DNA duplex and EPHA2:HIF1α-AS1_TFR2 triplex with 15 and 25 equivalents of RNA (TFO2-23). b Circular dichroism spectra of the EPHA2 DNA duplex (black), the heteroduplex (dark gray) and EPHA2:HIF1α-AS1-TFR2 (red) measured at 298 K. c Thermal melting assay of the EPHA2 DNA duplex (black), the heteroduplex (dark gray) and EPHA2:HIF1α-AS1-TFR2 (red). d Sequence of EPHA2 DNA (black) and HIF1α-AS1-TFR2 RNA (red). Watson-Crick base pairing is indicated with | and the Hoogsteen base pairing is indicated with ^*. Changes in the DNA duplex were quantitatively analyzed using NOESY spectra of duplex and triplex. Imino protons with strong attenuation (dark blue arrows) or medium attenuation (light blue arrows) of cross peak intensities in the imino-imino region were observed. e ¹H-1D NMR spectra of the EPHA2 DNA duplex (black), HIF1α-AS1 TFR2 RNA (blue), heteroduplex (dark gray) and EPHA2:HIF1α-AS1-TFR2 triplex (red) at 288 K. f Assignment of the imino region of the ¹H,¹H-NOESY spectrum of EPHA2 DNA duplex measured at 800 MHz and 288 K in NMR buffer with 5% D₂O. g Cartoon representation of DNA:DNA:RNA triplex docking studies with the following color code: DNA strand (blue and gray) and RNA strand (red). This shows an ensemble of the 20 top-ranked modeled structures for a DNA:DNA:RNA triplex. The figure was generated by using PyMol 2.5 (Schrödinger, LLC). Source data (for a) are provided as a Source Data file.

Fig. 4. HIF1α-AS1 limits EPHA2 and ADM expression through TFR2.

a, b CRISPRa (a, n = 6 independent experiments) or CRISPRi (b, n = 3 independent experiments) targeting HIF1α-AS1 in HUVECs followed by RT-qPCR for HIF1α-AS1, EPHA2 and ADM. Paired t test. A non-targeting gRNA served as negative control (CTL). a: *AS1 (p = 0.0009), *EPHA2 (p = 0.0335), *ADM (p = 0.0359); b: *AS1 (p = 0.0012), *EPHA2 (p = 0.008), *ADM (p = 0.0428). c Western blot with (AS1)/without (-, CTL) LNA knockdown of HIF1α-AS1 in two independent experiments using two different batches of HUVEC. GAPDH served as control. d–g Representative images (d, f) of a spheroid assay and quantification (e, g) of the cumulative sprout length of HUVECs treated with/without siRNAs against EPHA2 (d and e) or LNA GapmeRs targeting HIF1α-AS1 (f and g). Scale bar, 200 µm. One-Way ANOVA with Bonferroni test. e: CTL-VEGF-A (n = 12), CTL+VEGF-A (n = 15), EPHA2-VEGF-A (n = 13), EPHA2+VEGF-A (n = 12); *CTL-/+VEGF-A (p = <0.0001), *EPHA2-/+VEGF-A (p = <0.0001), *CTL/EPHA2+VEGF-A (p = <0.0001), *CTL/EPHA2-VEGF-A (p = 0.0079); g: CTL-basal (n = 21), AS1-basal (n = 26), CTL+VEGF-A (n = 19), AS1+VEGF-A (n = 23), CTL+bFGF (n = 12), AS1+bFGF (n = 32); *VEGF-A (p = 0.0495), *bFGF (p = 0.0012). h–k Quantification of the cumulative sprout length from the spheroid assays with siRNA targeting the HIF1α-AS1 intron (h, n = 14 replicates), with CRISPRi (i, CTL-basal(n = 10), AS1-basal (n = 11), CTL+VEGF-A (n = 11), AS1+VEGF-A (n = 14), CTL+bFGF (n = 10), AS1+bFGF (n = 13)), with CRISPRa (j, CTL-basal (n = 8), AS1-basal (n = 7), CTL+VEGF-A (n = 10), AS1+VEGF-A (n = 7), CTL+bFGF (n = 5), AS1+bFGF (n = 10)) or after overexpression (k, CTL-basal (n = 21), 1200-basal (n = 22), CTL+VEGF-A (n = 20), 1200+VEGF-A (n = 18), CTL+bFGF (n = 21), 1200+bFGF (n = 19)) of the first 1200 nt of the HIF1α-AS1 gene (included TFR2, named as 1200). One-Way ANOVA with Bonferroni test. h: *VEGF-A (p = 0.0109); i: *VEGF-A (p = <0.0001), *bFGF (p = 0.0006); j: *VEGF-A(p = 0.0429), *bFGF(p = 0.0489); k: *VEGF-A(p = <0.0001), *bFGF(p = 0.0004). l, m Western blot with (si) or without (CTL) siRNA-mediated knockdown of HIF1α-AS1 targeting the intron (l) or with (OE) or without (CTL) overexpression of the first 1200 nt of HIF1α-AS1 (m) in three independent experiments using three different batches of HUVEC. NONO served as control. n, o, p RT-qPCR of TGFBR1 (n), EPHA2 (o) or ADM (p) after replacement of HIF1α-AS1-TFR2 with single-stranded oligodeoxynucleotides (ssODN) containing MEG3-TFR or a DNA fragment of a luciferase negative control (Luc). -, no ssODN. n = 5 independent experiments, Paired t test. n: *(p = 0.0018); o: *MEG3(p = 0.0187), *Luc (p = 0.015); p: *(p = 0.0106). q, r ChIP with anti-S9.6 after replacement of HIF1α-AS1-TFR2 and qPCR for EPHA2 (q) or ADM (r) TTS. -, no ssODN. One-Way ANOVA with Bonferroni test. n = 3 independent experiments. q: *MEG3 (p = 0.0025), *Luc (p =0 .0024); r: *Meg3 (p = 0.0006), *Luc (p = 0.0006). Data are presented as mean values ± SEM. AS1, HIF1α-AS1. M, marker. Source data (for c, l, m) are provided as a Source Data file.

Fig. 5. HIF1α-AS1 interacts directly with the HUSH complex member MPP8.

a Volcano plot of HIF1α-AS1 protein interaction partners after RNA pulldown assay and ESI-MS/MS measurements with fold enrichment and p value. n = 5. Significant proteins are shown above the line (p < 0.05). b List of proteins enriched after RNA pulldown assay, their p value (p, unpaired t test, two-tailed) and absolute fold change (FC). c RIP with MPP8 antibodies and qPCR for HIF1α-AS1 TFR2. IgG served as negative control. n = 4 independent experiments, Mann–Whitney test. *(p = 0.0286). d RIP with histone3-lysine9-trimethylation antibodies and qPCR for HIF1α-AS1 TFR2. IgG served as negative control. n = 3 independent experiments, Paired t test. *(p = 0.0162). e Scheme of the different HIF1α-AS1 RNAs used for in vitro RNA immunoprecipitation. E1, Exon1; E2, Exon2. f Scheme of the different MPP8 mutants used for in vitro RNA-immunoprecipitation. IDR intrinsically disordered region, ANK Ankyrin repeat; Chromo, Chromodomain. g RT-qPCR after in vitro binding assay of purified MPP8 with in vitro transcribed HIF1α-AS1 RNAs. MPP8 antibodies were used for RNA immunoprecipitation (RIP). An T7-MCS in vitro transcribed RNA served as negative control (CTL). FL, full length; E1, Exon1; E2, Exon2. Δ indicates the deleted nt from HIF1α-AS1 full length. n = 4 independent experiments, One-Way ANOVA with Dunnett’s test. *FL (p = 0.018), *ΔE2 117-652 (p = 0.0453). h RT-qPCR after in vitro binding assay of different in vitro translated His-tagged MPP8 mutants with in vitro transcribed HIF1α-AS1 full length. Anti-His was used for RNA immunoprecipitation. n = 4 independent experiments, One-Way ANOVA with Dunnett’s test. *(p = 0.0029). i, j Proximity ligation assay of HUVECs with antibodies against MPP8 and H3K9me3 (i) or MPP8 and SETDB1 (j). The individual antibody alone served as negative control. Red dots indicate polymerase amplified interaction signals. Scale bar indicates 20 µm (i) or 10 µm (j). Images were representative of three independent experiments. Data are presented as mean values ± SEM.

Fig. 6. HIF1α-AS1 directs the HUSH complex member MPP8 and SETDB1 to triplex target sites.

a Chromatin immunoprecipitation (ChIP) with MPP8 antibodies with or without RNase A treatment and qPCR for the triplex target sites of EPHA2 and ADM. Primers against a promoter sequence of GAPDH served as negative control. n = 4 independent experiments, paired t test. *EPHA2 (p = 0.0223), *ADM (p = 0.0221). b, c ChIP with antibodies against SETDB1, MPP8 or NP220 in HUVECs treated with (AS1) or without (CTL) LNA GapmeRs against HIF1α-AS1. QPCR was performed for EPHA2 TTS (b) or ADM TTS (c). n = 5 independent experiments, paired t test. b: *SETDB1 (p = 0.0487), *MPP8 (p = 0.0287); c: *SETDB1 (p = 0.0358), *MPP8 (p = 0.0109). d, e ChIP with SETDB1, MPP8 or NP220 antibodies after CRISPRi (d) or CRISPRa (e) for HIF1α-AS1 and qPCR for EPHA2 TTS. n = 3 independent experiments, paired t test. d: *SETDB1 (p = 0.0371), *MPP8 (p = 0.0117); e: *SETDB1 (p = 0.0159), *MPP8 (p = 0.0465), *NP220 (p = 0.0202). f Genome tracks for EPHA2 of ATAC-Seq in HUVECs separately and as an overlay after knockdown of HIF1α-AS1 (black), SETDB1 (gray), MPP8 (blue) or the negative control (red), after CRISPRi and CRISPRa of HIF1α-AS1 or after LentiCRISPR-mediated deletions of HIF1α-AS1 TFR2, EPHA2 TTS or ADM TTS. ChIP-Seq data (H3K4me3, H3K27Ac, H3K9Ac) in HUVECs was derived from ENCODE. Numbers in square brackets indicate data range values. Red arrow indicates the TTS analyzed in this study, orange arrows indicate strong changes. The red box indicates the relevant location. NTC non-targeting control, TFR triplex forming region, TTS triplex target site. Data are presented as mean values ± SEM.