Multiomic characterization of RNA microenvironments by oligonucleotide-mediated proximity-interactome mapping

Abstract

RNA molecules form complex networks of molecular interactions that are central to their function and to cellular architecture. But these interaction networks are difficult to probe in situ. Here, we introduce Oligonucleotide-mediated proximity-interactome MAPping (O-MAP), a method for elucidating the biomolecules near an RNA of interest, within its native context. O-MAP uses RNA-fluorescence in situ hybridization-like oligonucleotide probes to deliver proximity-biotinylating enzymes to a target RNA in situ, enabling nearby molecules to be enriched by streptavidin pulldown. This induces exceptionally precise biotinylation that can be easily optimized and ported to new targets or sample types. Using the noncoding RNAs 47S, 7SK and Xist as models, we develop O-MAP workflows for discovering RNA-proximal proteins, transcripts and genomic loci, yielding a multiomic characterization of these RNAs' subcellular compartments and new regulatory interactions. O-MAP requires no genetic manipulation, uses exclusively off-the-shelf parts and requires orders of magnitude fewer cells than established methods, making it accessible to most laboratories.

Extended Data Figure 1.. Alternative O-MAP designs.

a, summary of RNA-targeted HRP-recruitment strategies tested. Design 1 uses biotinylated primary probes to recruit a streptavidin-HRP conjugate. Designs 2 and 3 use Digoxigenin (DIG)-labeled primary or secondary probes to recruit an HRP–conjugated anti-DIG antibody. Our final O-MAP design, which uses HRP-conjugated oligo probes, is shown for comparison (see also, Fig. 1a). The same anti-DIG antibody is used in designs 2 and 3; the same “universal landing pad” sequences are used in designs 3 and 4. b–e, limitations of Designs 1–3. In all cases biotin was visualized by staining with a fluorescent neutravidin conjugate. b, Design 1 was disfavored because in situ biotinylation cannot be distinguished from biotinylated probes. HeLa cells over-expressing NPM1~eGFP were probed using the same 47S-targeting probes as in (Supplementary Fig. 1), bearing 3´–terminal biotin. c, Design 2—analogous to HyPro—in HeLa cells. This sometimes produced well-resolved nucleolar-targeted biotinylation, but was disfavored due to antibody irreproducibility (see below). The high cost of DIG-labeled oligos would also limit its use with low-abundance transcripts, which can require dozens-to-hundreds of probes. d, Design 3 overcomes the oligo cost issue but still suffers from antibody background binding and irreproducibility. HeLa cells were probed with the same 47S-targeting primary probe set and DIG-labeled secondary oligo, using four different lots of commercial HRP-conjugated antibodies. In some cases, we observed well-resolved RNA-targeted biotinylation (panel i), though other lots from the same vendor exhibited significant off-target labeling (panel ii, arrows). Regents from other vendors exhibited varying degrees of blurring (panel iii), or excessive off-target biotinylation (panel iv). e, Design 3 is particularly problematic with lower-abundance RNA targets. Patski cells were probed with the same Xist-targeting, landing-pad-extended probes as used in the main text, divided into two sub-pools. Anti-DIG-HRP was from Vendor 1. Note that all conditions—both probe sub-pools and the omit-primary negative control–induced substantial off-target biotinylation. All scale bars, 20 μm.

Extended Data Fig. 2:. More O-MAP controls and imaging analysis.

a, O-MAP and negative control experiments were performed in HeLa cells, as indicated. Biotin was imaged using a fluorescent neutravidin conjugate; NPM1 via immunofluorescence. Note that omitting any component of the O-MAP pipeline ablated biotinylation signal. The 47S-O-MAP, 7SK-O-MAP, omit primary and scrambled primary conditions (left four columns) are the same images presented in (Fig. 1b). Scale bars, 20 μm. b, Representative line traces of each condition. Note the overlap in NPM1 and O-MAP signals in the 47S O-MAP condition; the lack of overlap in 7SK O-MAP, and the nearly complete absence of biotinylation under all other conditions. Scale bars, 10 μm. The 47S O-MAP image are the same data presented in (Fig. 1b). In all “omit-H2O2” conditions (marked *), cells were pre-quenched with sodium azide and ascorbic acid prior to the addition of biotin-phenol and H2O2. Simply removing H₂O₂ from the O-MAP protocol still resulted in targeted in situ biotinylation (that is nucleolar labeling using 47S probes), presumably due to photoactivation by HRP.

Extended Data Fig. 3:. O-MAP is more spatially precise than genetically encoded proximity-labeling systems.

a, Qualitative assessment of off-target biotinylation by 47S-targeted O-MAP (left) and nucleolar-targeted APEX2 ~ GFP (right). NPM1 was visualized by immunofluorescence; APEX2 ~ GFP by autofluorescence. O-MAP data are the same as presented in (Fig. 1b and Extended Data Fig. 1); APEX2 data are the same as in (Fig. 1c). Note prominent nucleoplasmic (off-target) labeling by APEX2, indicated by arrowheads. Scale bars, 10 μm. b, More head-to-head comparisons between 47S O-MAP and APEX2~NoLS, quantified in line-traces. Green: APEX2 ~ GFP biotinylation; Purple: 47S O-MAP biotinylation. Scale bars: 10 μm. c, Quantifying 47S O-MAP background biotinylation. Left: DAPI signal was used to identify Regions of Interest (ROI; yellow outlines), corresponding to individual nuclei, and the biotinylation intensities for all pixels within each ROI were measured. Right: histogram summarizing the distribution of biotinylation intensities observed within Cell 1. Note the prominent peak near zero, indicating that most pixels contained little or no signal. Inset highlights a minor peak of intense biotinylation (>17,500). d, The same approach was applied to APEX2~NoLS cells. Left: ROIs defined; Right: Histogram of observed biotinylation intensities within cell 1. Note the lower proportion of pixels near zero, fewer pixels at high signal intensities (inset) and the prominent peak between 1500–3600—suggesting significant background labeling. In c, d, a single z-slice is shown; scale bars, 20 μm. e, Cumulative distribution functions of biotinylation intensities for O-MAP and APEX2~NoLS (n = 4). Mean values over all ROIs (dark lines). Lighter bands correspond to the maximum and minimum values observed. Larger plot highlights the distributions at intensities below 10,000; inset shows all data. 59% of pixels in the O-MAP experiment were below 1000; only 14% of APEX2 pixels fell within this range, suggesting that O-MAP’s lower background biotinylation. Furthermore, 57% of APEX2 pixels fell between 1500—3600, while only 8% of O-MAP data fall in this range. All images under direct comparison (for example, all images in a; all in b; and the images in c and d) were collected on the same day, using the same acquisition parameters.

Extended Data Fig. 4:. Probing the nucleolar proteome with 47S O-MAP-MS.

a, Receiving-Operator Characteristic (ROC) analysis of the 47S vs 7SK O-MAP-MS experiment. True Positive and False Positive proteins were defined using lists of exclusively nucleolar and exclusively nucleoplasmic proteins, respectively, as reported by the Human Protein Atlas (HPA). An Area Under the Curve (AUC) of nearly 1.0 suggests strong and highly sensitive selectivity for nucleolar proteins over the nucleoplasmic proteome. b, These data were used to derive an optimal Log2(fold change, 47S/7SK) cutoff value of 0.523, and to define a putative list of 258 O-MAP core nucleolar proteins, as described in (Fig. 3). c–e, Parallel analysis using (47S/Scramble controls), instead of (47S/7SK). c, Volcano plot and histograms of showing the enrichment of HPA-nucleolar, HPA-Nucleoplasmic, and HPA-bilocalized proteins, using the same protein marker reference lists as in (a, b), and (Fig. 3c, d). Benjamini Hochberg corrected p-values (FDR). d, ROC analysis of the (47S/Scramble) data demonstrates slightly lower sensitivity than that of the (47S/7SK) analysis, though still exceptionally. e, As in (b), these data were used to determine an optimal Log2(fold change, 47S/Scramble) cutoff value of 2.201, defining a putative cohort of 286 O-MAP core nucleolar proteins. f, The putative nucleolar proteomes derived from the (47S/7SK) and (47S/Scramble) ROC analyses show considerable overlap (66–73%). Outliers were used for Gene Ontology (GO)-term analysis. Factors uniquely captured by the (47S/7SK) analysis were highly enriched for ribosome biogenesis factors, while those unique to the (47S/Scramble) analysis were enriched for nucleoplasmic functions. This suggests that the (47S/7SK) comparison more precisely captures the nucleolar proteome. Hypergeometric test with Benjamini-Hochberg correction.

Extended Data Fig. 5:. Coverage of the nucleolar proteome during the 47S O-MAP labeling time course.

a, Volcano plots for each 47S O-MAP labeling point, calculated relative to the 7SK/10-min label condition. Enrichment of nucleolar proteins derived from our k-medoid analysis (Fig. 3h–j; Supplementary Fig. 10) are highlighted in pink. Significance cutoffs were assigned at p_adj ≤ 0.05 and Fold Change (47S/7SK) ≥ 2.0. (dotted lines). Benjamini Hochberg corrected p-values (FDR) b, Table summarizing the recovery of the nucleolar proteome at each labeling time point. Note that coverage appears to plateau at 10 min.

Extended Data Fig. 6:. Comparison of 47S-targeted O-MAP-MS and HyPro-MS.

Top: Venn diagram of enriched proteins from 47S-targeted HyPro-MS (gray) and the k-medoid-derived high-confidence O-MAP nucleolar proteome (Fig. 3j; Supplementary Table 3). Note that O-MAP enriches more than 50% of the HyPro nucleolar proteome. Bottom: Gene Ontology analysis of the outlier proteins observed by HyPro-MS (101 proteins; left) and by O-MAP-MS (194 proteins, right). The top 10 most highly enriched terms are shown. Note that the O-MAP-MS outliers are predominantly nucleolar proteins and ribosome biogenesis factors, suggesting that HyPro has undersampled the nucleolar proteome. In contrast, HyPro outliers exhibit little association with the nucleolus or its central functions, suggesting that they may be noise. Hypergeometric test with Benjamini-Hochberg correction.

Extended Data Fig. 7:. Probing the Xi compartment with O-MAP-MS.

a, Receiving-Operator Characteristic (ROC) analysis of the Xist O-MAP-MS experiment. True Positives (183 proteins; Supplementary Table 2) were defined as Xist-interacting proteins observed previously via antisense oligonucleotide purification methods. False Positives (104 proteins; Supplementary Table 2) were defined using the Jackson Laboratory Mouse Genome Informatics portal (JAX MGI; https://www.informatics.jax.org/vocab/gene_ontology) list of tRNA modifying factors, manually curated to remove duplicate entries and factors that lack nuclear localization. An Area Under the Curve (AUC) of 0.92 suggests strong and highly sensitive selectivity for known Xist interactors relative to the broader proteome. b, These data were used to derive an optimal Log2(fold change, Xist/mock) cutoff value of 0.916, and to define a putative list of 621 O-MAP Xi-proximal proteins, as described in (Fig. 4). c, Behavior of other prominent gene sets in the Xist O-MAP experiment, based on their subnuclear localization (top) or molecular function (bottom). All sets were defined using JAX MGI, manually curated to remove duplicate entries and factors in our Xist True Positive list. None of these gene sets is more highly enriched than known Xist interactors en toto (Fig. 4c). However, each set contains a subset of factors that are enriched by Xist O-MAP-MS, suggesting a functional role for these factors within the Xi compartment. Student’s two-tailed T-Test, with a permutation-based FDR adjustment.

Extended Data Fig. 8:. 47S O-MAP-Seq enriches known and novel nucleolar transcripts.

a, the 47S pre-rRNA. Note prominent enrichment for the 5´–ETS, IT1, and ITS2 “transcribed spacer” domains. Sequences corresponding to the mature 18S, 5.8S, and 28S rRNAs are removed during sequencing library preparation. Reads are aligned to a custom genome assembly containing a single copy of the rDNA consensus sequence (courtesy of T. Moss, U. Laval) annotated as a unique chromosome. b, the U3 noncoding RNA, which directs key cleavage events during ribosomal biogenesis, c, d, exemplar Box C/D (c) and H/ACA (d) small nucleolar RNAs (snoRNAs). SnoRNAs are often expressed within the introns of protein-coding genes (gray). e, RNase MRP (enriched) upstream of the CCDC107 gene (not enriched). f, lncRNA SLERT, which is processed from the two H/ACA snoRNAs embedded in the TBRG4 gene. g, Example of a novel nucleolar transcript—a processed pseudogene—discovered by O-MAP-Seq.

Extended Data Fig. 9:. Comparison of 47S O-MAP-Seq and HyPro-Seq.

a, Heatmap of Pearson correlations, comparing three biological replicates of 47S O-MAP-Seq and two biological replicates (four technical replicates each) of the analogous HyPro-Seq experiment. Comparisons were made between log2(fold change, enriched/input) for each replicate. Note the high inter-replicate reproducibility of O-MAP (r = 0.87–0.93), and that its correlation to HyPro-Seq Replicate A (r = 0.53–0.58) is comparable to the correlation between HyPro-Seq biological replicates (r = 0.55–0.56). b, Example scatter plots of the data used in (a), highlighting O-MAP-Seq’s superior reproducibility. c, O-MAP-Seq and HyPro-Seq enrich many of the same transcripts. Volcano plot of 47S O-MAP-Seq data, equivalent to (Fig. 5b), with coding and noncoding isoforms merged. 47S O-MAP-Seq prominently enriches many of the transcripts that were enriched by 47S HyPro-Seq (red), in addition to a smaller set of unique RNAs (blue). Note that relatively few of the transcripts that were uniquely enriched by HyPro-Seq (dark gray) were de-enriched by O-MAP; the majority of these RNAs fell below the significance cutoff of detection, presumably due to the O-MAP-Seq dataset’s lower sequencing depth. Significance testing: Wald’s test. d, Breakdown of 47S-proximal RNAs by transcript class, for those enriched by O-MAP-Seq (blue), HyPro-Seq (gray), and both methods (red). Although the overall composition is similar, HyPro-Seq enriches more protein coding genes and mitochondrial tRNAs, which are not thought to be nucleolar, and fewer lncRNAs and pseudogenic transcripts. e, Isoform-level analysis of enriched transcripts arising from protein-coding genes, analogous to (Fig. 5d). Note that HyPro-Seq enriches a higher proportion of protein-coding isoforms, and lower proportion of retained intron-containing transcripts. Collectively, these data suggest that O-MAP-Seq largely recapitulates the data obtained by analogous methods, but with lower noise and higher reproducibility.

Extended Data Fig. 10:. Xist O-MAP-Seq enriches nascent transcripts of XCI-escape genes.

a, Enrichment of the Xist gene itself. Note different scales for input RNA and O-MAP-Seq tracks. The lack of intronic reads suggests that O-MAP-Seq has predominantly targeted and captured the mature Xist transcript. The absence of reads mapping to the antisense noncoding RNA TsiX, a lncRNA that is monoallelically expressed from the Xa (gray), confirms that O-MAP-Seq is precisely labeling the Xi. b, Enriched XCI-escape genes appear to be nascent transcripts. In all cases, read densities for both Input RNA (gray) and O-MAP-Seq (red) are shown, using matched scales. Note prominent intronic read density for all XCI-escape genes (Shroom4, Pbdc1, Magee1, Mid1, Erdr1). Transcript structures denoting the most prominent isoforms are displayed; other nearby genes not known to escape XCI are denoted in gray. c, Xist O-MAP-Seq does not appear to preferentially capture nascent transcripts of non-XCI-escape genes. Two examples (Tspan6 and Rps4x) are shown, neither of which was enriched by Xist O-MAP-Seq. Note the absence of prominent intronic read density, suggesting that the intronic signatures observed in (b) are not general artifacts of the O-MAP-Seq pipeline.

Figure 1.. O-MAP Design and Implementation.

a, Overview of O-MAP. Specimens are chemically fixed, and pools of antisense DNA probes are hybridized to the target RNA. These probes recruit a common, HRP-conjugated secondary probe that catalyzes in situ proximity-biotinylation. b, O-MAP enables precise RNA-targeted proximity labeling, here, in HeLa cells. In situ biotinylation imaged by neutravidin staining; NPM1 by immunofluorescence. Note nucleolar biotinylation in 47S O-MAP, nucleoplasmic biotinylation in 7SK O-MAP. c, Nucleolar-targeted APEX2 exhibits substantial off-target, nucleoplasmic labeling. d, O-MAP Probe Validation Assay. Primary probes are split into sub-pools that enable O-MAP and RNA-FISH to be performed simultaneously. Lack of co-localization suggests probe off-targeting. e, Probe Validation Assays on 47S-pre-rRNA and 7SK (HeLa), and Xist, in mouse Patski cells. f, Recovery of O-MAP-biotinylated proteins. Top: Streptavidin-HRP blot of whole cell lysates. Note ladder of biotinylated proteins from 47S O-MAP Bottom: Ponceau stain. All scale bars: 20 μm.

Figure 2.. O-MAP is portable across specimen types and applicable to diverse RNA targets.

a, 47S-O-MAP in cultured mammalian cell lines. NPM1 immunofluorescence denotes nucleoli. All human-derived lines used the same probe set and hybridization conditions as HeLa cells (Fig. 1); MEFs used analogous mouse-targeting probes. b, 47S O-MAP in human patient-derived pancreatic ductal adenocarcinoma (PDA) organoids. FBL immunofluorescence denotes nucleoli. c, 47S O-MAP in cryo-preserved mouse tissue slices. d, O-MAP Probe Validation Assay (Fig. 1d) applied to a compendium of target transcripts. Note conspicuous overlap between RNA-FISH (green) and O-MAP (magenta) signals. Images from HeLa (47S, 7SK, MALAT1, NEAT1), Patski (Xist, Kcnq1ot1, Firre), patient-derived fibroblast (WDR7), and U2OS cells ((CxG)n and (G4C2)n RNAs). Firre is expressed from a single-copy transgene; (CxG)n and (G4C2)n RNAs are artificial constructs under doxycycline-inducible expression; all other targets are endogenous transcripts. Insets show zoomed-in sections of the same images, to highlight signal overlap. e, O-MAP at nascent transcripts probes subnuclear neighborhoods. Left: Mature Xist coats the entire inactive X-chromosome (Xi), while nascent Xist transcripts uniquely denote the X-inactivation center (XIC). Right: O-MAP targeting Xist introns induces biotinylation at confined foci within the Xi “cloud.” All scale bars: 20 μm.

Figure 3.. O-MAP-MS for probing RNA-proximal proteomes.

a, Approach. Parallel O-MAP experiments targeted the 47S (nucleolar), 7SK (nucleoplasmic) or used scrambled probes (background), in HeLa cells. b–g, Analysis of a single-shot 47S/7SK O-MAP-MS experiment. b, Recovery of known nucleolar proteins and 7SK interactors. Two different probe sets, requiring different labeling times, were used in 7SK experiments; these were time-matched to scrambled-probe controls. Inclusive-median box-whiskers with quartiles shown, centered on the mean; whiskers denote the minimum and maximum observed values; n = three biological replicates. c–d, Global enrichment of known nucleolar, nucleoplasmic, and bi-localized marker proteins, defined by the Human Protein Atlas (HPA). Dotted line denotes the optimal threshold separating the compartments, determined by Receiver-Operating Characteristic (ROC) analysis (Extended Data Fig. 4). c, Histograms plotting the enrichment of each sub-compartmental proteome. Only proteins with q-value ≤ 0.05 are shown. d, Volcano plot showing all data. Benjamini Hochberg corrected p-values (FDR) e, Gene Ontology analysis for the O-MAP-nucleolar and O-MAP-Nucleoplasmic proteomes, defined from ROC analysis. The top eleven most enriched terms are shown. Hypergeometric test with Benjamini-Hochberg correction. f, 7SK O-MAP enriches the Nuclear Speckle proteome. Histograms and volcano plots as in c,d, showing enrichment of speckle proteins in both 7SK-vs-47S and 7SK-vs-Scramble comparisons. g, Gene Set Enrichment Analysis of the “O-MAP 7SK-proximal” proteome. GOBP and GOCC: Gene Ontology Biological Process and Cellular Component, respectively. h–j, Higher coverage of the nucleolar proteome using an O-MAP in situ biotinylation time course. h, 47S O-MAP was performed in parallel for the indicated times, enriched, and TMTPro-labeled (n = 3). A single 7SK probe set and time point was used for normalization. In situ biotinylation visualized by neutravidin staining. Scale bars: 20 μm. i, k-medoid clustering yields a clade of 313 high-confidence nucleolar proteins. Notable marker genes are indicated. j, Nearly all members of the nucleolar medoid group corresponds to annotated or manually curated nucleolar proteins.

Figure 4.. Proteomics of the inactive X-chromosome by Xist O-MAP-MS.

a, Approach. Xist-targeted O-MAP and negative controls (omitting primary probes) were processed as indicated, in mouse Patski cells. Protein abundances were measured by Label-Free Quantification (LFQ; n = 4 biological replicates). b, Xist O-MAP-MS enriches established Xist interactor, but not negative control proteins. Box whisker plots with individual data points shown. Inclusive-median box-whiskers with quartiles shown, centered on the mean; whiskers denote the minimum and maximum observed values; n = 4 biological replicates. c, O-MAP-MS recovers Xist interactors at higher depth than do oligo capture-based methods. Interaction graphs for three representative complexes are shown; factors significantly enriched by each technique are colored as indicated. d, Global enrichment of known Xist interactors (magenta) and negative controls (nuclear-localized tRNA-biogenesis factors, blue). Dotted line denotes the optimal enrichment threshold, determined by ROC analysis (Extended Data Fig. 7). Top: Histograms plotting the enrichment and de-enrichment of each gene set. Only proteins with q-value ≤ 0.05 are shown. Bottom: Volcano plot showing all data. Note conspicuous separation between Xist interactors and negative controls. Student’s two-tailed T-Test, with a permutation-based FDR adjustment. e, GO analysis, Significance: hypergeometric test with Benjamini-Hochberg correction. f, GSEA of the O-MAP Xi-proximal proteome reveal prominent enrichment in factors involved in chromatin organization, negative regulation of gene expression, and RNA processing, as expected.

Figure 5.. O-MAP-Seq for probing RNA-proximal transcripts.

a, Approach. In situ biotinylated proteins are used to enrich nearby transcripts, which are then quantified by RNA-sequencing. b–f, O-MAP-Seq characterization of the HeLa nucleolar transcriptome. b, Volcano plots of 47S O-MAP-Seq data, demonstrating enrichment of snoRNAs, lncRNAs, pseudogenes (left), and de-enrichment of mRNAs (right) (n = 3) Wald’s test. c, Summary of enriched and de-enriched RNA classes in the nucleolar transcriptome. d, Nucleolar transcripts expressed from protein-coding genes are predominantly noncoding variants; de-enriched transcripts are predominantly mRNAs. NSD: nonstop decay; NMD: nonsense-mediated decay. e, Nearly half of the nucleolar transcriptome is encoded from loci within Nucleolar-Associated chromatin Domains (NADs; Fig. 6). One-sided Fisher’s exact test. f, Nucleolar transcripts are enriched in Transposable Element (TE) domains. Left: Z-scores of variance stabilizing transformed (VST) data, corresponding to major TE families. Right: volcano plot of individual TE classes in 47S O-MAP-Seq data. Wald’s test. g, O-MAP-Seq identifies novel nucleolar-localized transcripts. Note co-localization between ENSG00000–286147.1 RNA-FISH and NPM1 Immunofluorescence (arrows; highlighted in zoom insets), quantified on right. h–j, Characterizing transcripts near the inactive X-chromosome (Xi), in mouse Patski cells. h, Volcano plots of Xist O-MAP-Seq data. Left: enrichment for X-linked transcripts that escape X-chromosome inactivation (XCI, red) and autosomal transcripts, including several chromatin-regulatory lncRNAs (indicated). Right: enrichment for several classes of X-linked transposable elements. Wald’s test. i, Gm14636 is a novel XCI-escape gene. Left: Gm14636 enrichment by O-MAP-Seq. Middle: co-localization of Gm14636 RNA-FISH and Xist O-MAP Note penetrant mono- or bi-allelic expression from the Xa (gold arrows) and/or Xi (white arrowheads), quantified on the right. j, Kcnq1ot1 localizes near the Xi. Panel arrangement parallels that of i.

Figure 6.. O-MAP-ChIP for probing RNA-proximal genomic loci.

a, Approach. In situ biotinylated proteins are used to enrich nearby chromatin loci, which are then quantified by DNA-sequencing. b–d, O-MAP-ChIP characterization of Xist genomic interactions, in Patski cells. b, Xist O-MAP-ChIP predominantly labels the X-chromosome (right, and inset). Data for the entire mouse genome are shown; Xist genomic locus is noted below. c, Xist O-MAP-ChIP is specific to the inactive X-chromosome (Xi). Histogram of Allelic proportions for ChIP data, quantified using SNPs specific to the Xi and Xa. d, Putative interactions between autosomal loci and the Xi. The kcnq1ot1 locus—but not MALAT1 and NEAT1 loci—appear enriched in Xist O-MAP-ChIP data. L2FC: log₂(Fold change, Enriched/Input). e–i, O-MAP-ChIP characterization of Nucleolar Associated Domains (NADs). e, 47S-targeted O-MAP in HeLa cells recapitulates the known human NAD architecture. Most of chromosome 8 is shown. Nucleolar fractionation data taken from (Ref. ). f, Conservation of NAD architecture between HeLa and HT1080 cells. One-sided Fisher’s exact test. g, Parallelized analysis of NAD architecture across four Pancreatic Ductal Adenocarcinoma (PDA) cell lines (ASPC1, SUIT2, 8988T, Panc3.27). Upset Plot summarizing NAD conservation, or lack thereof, between lines. The total number of NADs in each line appears nearly invariant. h, O-MAP-ChIP identifies NADs that are differentially regulated between Classical and Basal PDA subtypes. Examples of constitutive (left) and Differential (right) NADs on Chromosome 14 are shown. i, ChromHMM analysis reveals differential enrichment of chromatin signatures among HeLa and PDA cell line NADs.