De novo annotation of lncRNA HOTAIR transcripts by long-read RNA capture-seq reveals a differentiation-driven isoform switch

Abstract

Background: LncRNAs are tissue-specific and emerge as important regulators of various biological processes and as disease biomarkers. HOTAIR is a well-established pro-oncogenic lncRNA which has been attributed a variety of functions in cancer and native contexts. However, a lack of an exhaustive, cell type-specific annotation questions whether HOTAIR functions are supported by the expression of multiple isoforms.

Results: Using a capture long-read sequencing approach, we characterize HOTAIR isoforms expressed in human primary adipose stem cells. We find HOTAIR isoforms population displays varied splicing patterns, frequently leading to the exclusion or truncation of canonical LSD1 and PRC2 binding domains. We identify a highly cell type-specific HOTAIR isoform pool regulated by distinct promoter usage, and uncover a shift in the HOTAIR TSS usage that modulates the balance of HOTAIR isoforms at differentiation onset.

Conclusion: Our results highlight the complexity and cell type-specificity of HOTAIR isoforms and open perspectives on functional implications of these variants and their balance to key cellular processes.

Keywords: Adipose differentiation; Adipose stem cells; Capture-seq; HOTAIR; Long-read sequencing; lncRNA isoform.

Fig. 1

HOTAIR expression during adipogenic differentiation. a Quantitative RT-PCR analysis of HOTAIR expression in HeLa, MCF7, MDA-MB-231, and ASCs from two independent donors (ASC-1 and ASC-2). b Differential expression of HOTAIR between differentiation time points analyzed by short-read RNA-seq (mean ± SD; *p < 0.05, **p < 0.01, ***p < 0.001, limma moderated t-statistic; n = 3). c RT-qPCR analysis of HOTAIR relative expression, normalized to D0 ASCs, during adipogenic and osteogenic differentiation (mean fold difference ± SD; ****p < 0.0001, two-way ANOVA; n = 3)

Fig. 2

Identification of HOTAIR isoforms in differentiating ASCs. a Bioinformatic pipeline for identification of HOTAIR isoforms. b Representative integrative genomics viewer (IGV) tracks for short-read Illumina RNA-seq (upper) and PacBio capture-seq (lower) coverage tracks on HOTAIR (D0). c Venn diagram showing the overlap between HOTAIR isoforms identified using TAMA and Cupcake ToFU. d-h SQANTI characterization of 34 HOTAIR transcripts, with: (d) Number of isoforms per SQANTI categories and presence of non-canonical splice sites. e Distance in base pairs from the isoform start to the nearest CAGE peak summit from Ref. [34]. f HOTAIR exon E7 diagram showing 3’ end polyA tails. g Cumulative read counts of the isoforms shown by SQANTI categories. h Cumulative read number per isoform vs. number of samples in which the isoform was detected (cutoff: 0.1 × 10³ reads per isoform). i Exon structure of 23 high confidence HOTAIR isoforms colored by SQANTI categories. FSM: full splice match; ISM: incomplete splice match; NIC: novel in catalog; NNC: novel not in catalog

Fig. 3

HOTAIR splicing across functional domains during adipogenesis. a Schematic representation of HOTAIR 3’ exons (exons E4 to E7) containing LSD1 and PRC2 binding domains. Alternative splicing events and polyA (pA) sites usage detected in 5 representative HOTAIR high confidence isoforms are represented. b Proportion of long-read Capture-Seq reads for each E7 length across differentiation time points. c Schematic representation of RT-qPCR amplicons (upper panel) and qRT-PCR analysis of HOTAIR E7 length variation during adipose differentiation (lower panel). d PacBio Capture-Seq analysis of HOTAIR exon E5 alternative splice site usage and e of the proportion of major exon E5 splice variants during adipogenic differentiation. f Schematic representation of PCR amplicons (upper panel) and semi-quantitative RT-PCR analysis of HOTAIR expression using primers located in exons E3 and E5 (E3-E5; lower panel). SF3A1 is shown as a loading control. RT-: no reverse transcriptase control; NTC: no template control. Full-length gels are presented in Additional file 4, Fig. S4

Fig. 4

Remodeling of the HOTAIR isoform pool during adipogenesis. a Schematic representation of HOTAIR 5’ exons (exons E1 to E5) showing novel HOTAIR exons and 9 alternative TSSs detected by PacBio Capture-Seq. b Cumulative number of long-reads for each HOTAIR starting exon. c Heatmap of the proportion of isoforms expressed at each time in the adipogenesis time course. Matching full length reads from PacBio long-read Capture-Seq are compared with each time point. d, e Proportion of read coverage for each start exon quantified from long-read data for (d) HOTAIR isoforms cumulating > 500 reads over the time course and (e) E3- and E3.1-starting isoforms. f Normalized short-read RNA-seq read count per kilobase of exon sequence for E3 and E3.1 exons during differentiation, quantified with featureCounts with strict parameters (-f -g exon_id -p -s 2 -O –fraction -B -C) (n = 3, ***p < 0.001 paired t-test with Benjamini–Hochberg adjusted p value) g Semi-quantitative RT-PCR analysis of HOTAIR expression using primers located in HOTAIR exons E3.1-E4 and E3-E5. SF3A1 is shown as a loading control. A representative image from one of three independent experiments is shown for adipogenesis (left) and osteogenesis (right); see also Additional file 1 Fig. S3b,c. RT-: no reverse transcriptase control; NTC: no template control. Full-length gels are presented in Additional file 4 Fig. S4,5,6

Fig. 5

HOTAIR isoform expression is highly cell-type specific. a Number of SRA samples with 10 or more junction spanning reads for exons E3.1-E4 and E3-E4. b Semi-quantitative RT-PCR analysis of HOTAIR expression in ASCs, primary myoblasts, BJ fibroblasts, HEK293T and HeLa cells using primers located in HOTAIR exons E3.1-E4 and E3-E5. SF3A1 is shown as a loading control. Full-length gels are presented in Additional file 4 Fig.S7. c IGV tracks over HOTAIR locus showing predicted HOTAIR regulatory elements (REMs) from the EpiRegio database [42, 43] (red: activating, blue: repressing), average ATAC-Seq tracks from ASCs [44, 45], Myoblasts [46] and HeLa cells [47], and ChIP primers location. d ChIP-qPCR analysis of histone modifications during adipogenesis of ASCs (mean ± SEM of n ≥ 3 independent differentiation experiments; **p < 0.01 Mixed effect analysis)

Fig. 6

HOTAIR stability increases upon growth arrest. RT-qPCR analysis of a HOTAIR, b CEBPD and c GAPDH levels after flavopiridol treatment (mean ± SD; *p < 0.05, **p < 0.01, two-way ANOVA with Šídák's multiple comparisons test; n = 3 independent experiments)