RNA pull-down was conducted in three replicates according to the protocol described in Desideri et al. ¹⁴ To enrich the nuclear RNA fraction, cells were permeabilised for 15 minutes on ice in the 50 ul/cm ² of the following buffer: Triton-X100 0.5% (v/v), NaCl 150 mM, Tris pH 7.5 50 mM, MgCl ₂ 3 mM, DTT 1 mM, Protease inhibitor cocktail 1x, RNAse inhibitor 0.2 U/ul. This buffer was carefully aspirated, and remaining nuclei were gently rinsed with the same buffer without Triton-X100 for 1 minute on ice. Primers used for CHASERR pull-down are listed in Supplementary Table 1. ¹⁵

The reads for three replicates were mapped to the hg38 genome using STAR ¹⁶ with the option clip5pNbases 0 3. The reads were counted in gene features from gencode v48 annotation using STAR quantMode and in promoter regions identified by FANTOM6 ^{5, 17} using Rsubread featureCounts v2.16.1. ¹⁸ The features with 0 expression in the pull-down were excluded from the further analysis. For gene feature quantification, the fibroblast expression table was downloaded from FibroDB ^{19, 20} for background normalisation; samples with TGF-beta treatment were excluded. TPM normalisation was applied to each gene, and log10 values with pseudo count were used for further analysis. For promoter quantification, FANTOM6 expression for the control samples was used as a normalisation background; FANTOM6 expression and pull-down expression were converted to CPM, and log10 values with pseudo count were used for further analysis.

To assess the potential RNA-RNA interaction targets, metric A was calculated for each gene as the absolute deviation metric of pull-down normalised expression from CHASERR pull-down normalised expression. A = | 1 − ( pulldown ( gene ) − background ( gene ) ) ( pulldown ( CHASERR ) − background ( CHASERR ) ) | , where pulldown() and background() are mean log10 TPM expression for each gene.

Gene Ontology Biological Process enrichment analysis was performed using clusterProfiler v4.14.3 ²¹ with a significance threshold of q-value < 0.05, genes with non-zero expression in pull-down were used as the universe set.

Differential expression analysis to compare the CHASERR knockdown and control states was performed at the level of individual promoter regions separately for the ASO_G0272888_AD_07 (ASO 07) and ASO_G0272888_AD_10 (ASO 10). The analysis was performed using DESeq2, ²² and regions with an FDR < 0.05 and a |logFC|> 0.1 were selected for further analysis.

To define the CHASERR interactome, we performed RNA pull-down followed by sequencing in primary fibroblasts. This revealed a network highly enriched for non-coding RNAs. CHASERR itself accounted for 36% of genome-mapped reads, confirming specific enrichment.

Since highly abundant RNAs are more likely to be recovered non-specifically, we normalised pull-down expression to a fibroblast transcriptome baseline. ¹⁹ This revealed a significant correlation between pull-down and baseline expression (Pearson r = 0.5033, p-value < 2.2e-16), establishing a general relationship, while CHASERR itself showed 5-fold specific enrichment ( Figure 1A). To distinguish genuine interactors from this abundant background, we reasoned that true binding partners should be co-enriched to a level approximating CHASERR’s own abundance. We therefore selected RNAs whose pull-down abundance most closely matched that of CHASERR; the 1% of genes with the smallest deviation from CHASERR’s level were classified as high-confidence interactors ( Figure 1B, Supplementary Table 2 ¹⁵ ).

The top-ranking candidate was RMRP, the RNA component of the mitochondrial RNA processing endoribonuclease. The high-confidence interactor set also comprised multiple lncRNAs (including LINC-PINT and FTX) and microRNA host genes (e.g., MIR4435-2HG, MIR100HG), consistent with CHASERR functioning as a regulator for diverse non-coding RNA species. The enrichment in miRNA genes supports prior evidence of CHASERR’s role as a microRNA sponge. ¹⁰ Gene Ontology enrichment analysis of the network strongly implicated processes of miRNA- and ncRNA-mediated post-transcriptional silencing ( Figure 1C), corroborating this molecular role.

To probe functional connectivity, we tested whether interactors were altered upon CHASERR knockdown. The lncRNA CRNDE — the lncRNA that stabilises SIRT1 deacetylase ²³ — the only interactor, besides CHASERR itself, that was significantly downregulated following knockdown, revealing a specific regulatory connection (FDR < 0.05). Moreover, a direct CHASERR–CRNDE RNA–RNA interaction was predicted computationally using ASSA. ¹³ Together, these findings establish CHASERR as a hub for a defined non-coding RNA network, with RMRP and CRNDE as core components.

We next tested whether CHASERR controls transcription via direct promoter binding. We quantified CHASERR pulldown reads across FANTOM6 promoter regions, normalising to CAGE-seq expression from control fibroblasts. Promoters were stratified based on enrichment: those with a pull-down signal exceeding baseline (normalised expression > 1) and those without (normalised expression < 1).

Promoters enriched for CHASERR binding showed a strong, specific association with transcriptional changes upon its depletion ( Figure 2A-E, Supplementary Tables 3, 4 ¹⁵ ). When compared to promoters differentially expressed after CHASERR knockdown, we found a highly significant overlap between upregulated genes and promoters without pull-down signal exceeding baseline (p-value < 2.2e-16 and p-value < 1.056e-05 for two distinct ASOs). Downregulated genes were significantly enriched for CHASERR-bound promoters with pull-down signal exceeding baseline (p-value < 3.371e-14 for ASO 07). Thus, promoter binding by CHASERR is functionally linked to its role in gene regulation.