DDX18 is required for hESC maintenance through the regulation of RiBi and protein synthesis
DDX18 is highly enriched in pluripotent hESCs relative to differentiated27 or somatic28 cells (Fig. S1a) and abruptly downregulated upon embryoid body (EB) differentiation at kinetics even faster than NANOG and OCT4 (Fig. S1b), suggestive of its specific function in hESC maintenance. shRNA-mediated DDX18 knockdown (KD) (shDDX18) at both protein (Fig. 1a) and RNA (Fig. 1b) levels, relative to control KD (shLuc), functionally validated its critical roles in maintaining hESCs, manifested by downregulation of the core pluripotency and stem cell surface marker genes and upregulation of lineage-specific and developmental genes (Fig. 1a–c), as well as reduced alkaline phosphatase staining (Fig. 1d). RNA sequencing (RNA-seq) in control and DDX18 KD cells (Fig. 1e) revealed 789 significantly downregulated genes for “stem cell population maintenance”, “rRNA processing”, and “translational initiation”, and 1903 significantly upregulated genes for multilineage differentiation and development based on both the gene ontology (GO) analysis (Fig. 1f) and gene set enrichment analysis (GSEA) (Figs. 1g, S1c, d). Of note, the “ectoderm differentiation” is not enriched for the upregulated genes (Fig. S1c, d), indicating lineage skewing towards mesendoderm differentiation of DDX18 KD hESCs (Fig. S1e). When compared to the transcriptomes of the three germ layers29, the transcriptome of DDX18 KD cells clustered with those of endoderm and mesoderm but not ectoderm cells, further highlighting the skewed mesendoderm differentiation of DDX18 KD cells (Fig. S1f). Among the upregulated genes enriched with the “embryonic organ development” (Fig. 1g) are HOXA-D clusters (e.g., HOXA in Fig. 1h), key developmental regulators that are not expressed but held in bivalent chromatin domains in pluripotent stem cells to be resolved and activated upon differentiation30,31,32.
a Western blot analysis of pluripotency and lineage-specific transcription factors in control (shLuc) and DDX18 knockdown (KD) cells with two independent shRNAs (shDDX18#1 and shDDX18#3). β-Actin was used as the loading control. The experiment was repeated independently 3 times with similar results. b Quantitative RT–PCR analysis of pluripotency genes and differentiation genes. Bars represent the average of three independent experiments. T is also known as Brachyury. c Flow cytometric analysis of the pluripotency-related surface markers SSEA-4 and TRA-1-60. Data represent three replicates of two independent shRNAs. d Alkaline phosphatase (AP) staining of control and DDX18 KD cells. Biological replicates were performed using two shRNAs targeting DDX18. e Volcano plot showing dysregulated genes in DDX18 KD cells by RNA-seq. The green dots denote downregulated genes in DDX18 KD samples (log2 [fold change] <−1, p < 0.05); red dots denote upregulated genes in DDX18 KD samples (log2 [fold change] >1, p < 0.05). Biological replicates from two hESC lines were analyzed. The p-value was calculated with edgeR. f Gene Ontology analysis of the downregulated and upregulated genes in DDX18 KD cells relative to control cells. The most significantly enriched biological processes GO terms with p values are plotted. The p values were calculated using the DAVID bioinformatics resource system. g Geneset enrichment analysis (GSEA) for the terms “Cytoplasmic translation”, “Ribosome biogenesis”, and “Embryonic organ development” in DDX18 KD hESCs. P value was calculated using the GSEA web tool. h IGV track showing HOXA cluster genes are upregulated upon DDX18 KD. i SUnSET assay measuring relative puromycin levels incorporated into proteins during translation in DDX18 KD and controls human ESCs, indicated by the anti-puromycin western blot. The Coomassie blue staining gel and the Vinculin blot are the loading controls. WT and mutant DDX18 protein levels under KD and rescue conditions are also indicated. The experiment was repeated independently twice with similar results. Source data are provided as a Source Data file.
The prevalent association of translation-related GO and biological processes with downregulated genes upon DDX18 KD prompted us to investigate the potential translational defects of DDX18 KD hESCs. First, a SUnSET assay to measure the newly synthesized peptides33 revealed a marked decrease of total protein synthesis in DDX18 KD relative to control cells, a global translational defect that can be rescued with wild-type DDX18 (WT) but not the RNA helicase inactive DQAD mutant34,35 (Fig. 1i). Next, SILAC-MS (stable isotope labeling by amino acids coupled with mass spectrometry) quantitative proteomics (Supplementary Data 1) identified 879 downregulated and 319 upregulated proteins with the enrichment of translation-related and differentiation-related GO terms, respectively, in DDX18 KD relative to control KD hESCs (Fig. S1g, h). By integrating the RNA-seq and SILAC data (Fig. S1i), we found that, while many genes/proteins are positively regulated at both transcriptional and translational levels (Groups I and VI), there are more genes/proteins regulated at translational levels without significant transcriptional changes (Groups II and V). Only a small number of genes/proteins exhibit anti-correlation between transcription and translation (Groups III and IV) (Fig. S1i; Supplementary Data 2). Furthermore, consistent with our previous findings in mouse ESCs (mESCs)19, our DDX18 ChIP-seq experiment in hESCs revealed that no significant DDX18 binding signals could be detected in other genomic regions beyond the rDNA loci (Fig. S5f), excluding the direct transcriptional effects of DDX18 on the pluripotency and development gene expression programs except for the rDNA gene expression.
Together, these results establish the roles of DDX18 and its RNA helicase activity in maintaining hESCs through its nucleolar functions in RiBi and translational control, a finding in line with emerging evidence linking translational control and protein synthesis to ESC maintenance and embryonic development20,36,37.
DDX18 undergoes phase separation in vitro and in vivo
To understand how DDX18 may exert its nucleolar functions in regulating RiBi and translation for hESC pluripotency, we first performed 3D structured illumination microscopy (3D-SIM) and confirmed its sub-nucleolar localization in the NPM1-containing GC layer in hESCs (Fig. 2a, b). Nucleolus assembly is driven by liquid-liquid phase separation (LLPS), facilitated by nucleolar proteins harboring intrinsically disordered regions (IDRs) that mediate multivalent interactions to assist macromolecular assembly24,38,39. Employing PONDER VSL2 webtool40 for protein disorder predictions we identified two IDRs at DDX18’s N-terminal and C-terminal ends, respectively (Fig. S2a). Phase separation properties of DDX18 under or near physiological salt concentration (150 mM NaCl) were first verified in vitro using DDX18-GFP recombinant protein (Fig. 2c, conditions A’, B’, and C’), which does not require the molecular crowding reagent polyethylene glycol (PEG). DDX18 droplets can be disrupted by higher salt concentrations (Fig. 2c, d, condition series from D’ to F’); however, this disruption could be overcome by increasing the protein concentration (Fig. 2d). The LLPS nature of DDX18 in vivo was also confirmed in hESCs expressing DDX18-GFP at the endogenous DDX18 locus (Fig. 2e) by time-lapse microscopy, revealing that DDX18-GFP undergoes fusion and fission within minutes (Fig. 2f) comparable with the nucleolus fusion speed in vivo11,38, and by fluorescence recovery after photobleaching (FRAP) with the average time for DDX18 to fully recover after the photobleaching around 104 ± 8 s (Figs. 2g, S2b), slightly slower than GC protein NPM138.
a Schematic depiction of a eukaryotic cell with representative nuclear and nucleolus structures. FC (fibrillar center), DFC (dense fibrillar component), GC (granular component). (Created in BioRender. Malik, V. (2025) https://BioRender.com/v27l155). b Representative hESC SIM images showing the localization of DDX18 relative to DFC (FBL) and GC (NPM1). The experiment was repeated independently twice with similar results. c Representative images of liquid-like droplets formed with various NaCl concentrations (A’-F’), visualized by GFP emission of DDX18-GFP. DDX18-GFP is 5 µM. The experiment was repeated independently twice with similar results. d Phase diagrams for LLPS of DDX18 under different concentrations of protein and NaCl determined by turbidity assays. DDX18 concentrations spans from 0.5 µM to 20 µM, and NaCl from 50 mM to 300 mM. Representative images from series A’ to F’ are shown in (c). e Schematic diagram for the DDX18 C-terminal GFP fusion reporter knock-in allele. f Time-lapse images showing the fusion and fission of DDX18-GFP within the nucleolus. g Quantification of FRAP data for DDX18-GFP signal. The bleaching event occurs at t = 0 s. DDX18-GFP exhibits fast dynamics with nearly complete recovery (>90%) on a timescale of τ = 104 (pm) 8 s. Data represent three independent experiments and are presented as mean values (pm) SD. h Representative images of liquid-like droplets formed by fusion proteins containing 10 µM DDX18 full-length (FL), N-terminal IDR deletion (∆NIDR), C-terminal IDR deletion (∆CIDR), and combined N-terminal/C-terminal IDR deletion (∆IDR) detected by GFP (top, in vitro) in 150 mM NaCl, 25 mM Tris-HCl, pH 8.0; anti-Flag immunostaining (middle, in vivo), and the merged images (bottom, in vivo) showing 3xFlag-DDX18 (red), FBL(green) and DAPI (blue) in human ESCs. The experiment was repeated independently twice with similar results. i Phase diagrams for LLPS of DDX18 full-length and truncated mutants (∆NIDR, ∆CIDR, ∆IDR) under various concentrations of proteins (0.5 µM to 20 µM) in the presence of 150 mM NaCl, 25 mM Tris-HCl, pH 8.0 by turbidity assays. Source data are provided as a Source Data file.
DEAD-box helicases are conserved from bacteria to mammals containing the N-terminal domain, helicase core domain, and C-terminal domain41. The interactions implicated in the phase separation include long-range electrostatic effects, prion-like property-mediated interactions, and aromatic amino acid-mediated π-π interactions42. To explore which domains and which types of molecular interactions confer DDX18 functions in phase separation, we analyzed the fraction of charged residues and net charge per residues (NCPR) (Fig. S2c, top track), presence of prion-like domain (PLD) (Fig. S2c, middle track), and propensity for π-π contacts (Fig. S2c, bottom track). We found neither PLD nor π-π contacts were involved in DDX18 phase separation. Instead, NCPR analysis identified positively charged basic tracts (blue) highly enriched within NIDR and CIDR of DDX18 (Fig. S2c). Furthermore, we found that both NIDR and CIDR are required for DDX18 phase separation: lacking either domain (ΔNIDR or ΔCIDR) or both (ΔIDR) greatly reduced the droplet formation in vitro (Fig. 2h, i) and in vivo (Figs. 2h, S2d). When fused with a far-red fluorescent protein (faRFP), NIDR alone, but not CIDR alone, was enough to drive the fusion protein into the nucleolus in vivo (Fig. S2d). Interestingly, we found NIDR harbors a predominant positively charged basic tract consisting of an evolutionarily conserved K-stretch (8 Lysines and 1 Arginine) (Fig. S2e, f), previously reported to be enriched in disordered regions and interact with RNA to drive and regulate cellular condensations43. Accordingly, we found that the K stretch mutant (9 K/R to all A mutation; 9 A) was compromised in its nucleolar localization in vivo (Fig. S2g) and RNA-facilitated formation of the “core-shell” like structure44 compared with WT (Fig. S2h). Of note, the DQAD helicase mutant of DDX18 failed to form any core-shell-like structure (Fig. S2h).
To further corroborate the above findings, we conducted a rescue experiment using wild-type (WT), DQAD, and IDR-deleted versions of DDX18 in DDX18 KD cells, followed by qPCR analysis. Our data demonstrated that only the wild-type DDX18 could fully restore hESC maintenance (Fig. S2i) and pluripotency/lineage gene expression programs (Fig. S2j) during differentiation, thus confirming the essential role of DDX18’s RNA helicase activity and phase separation property in its functionality. Together, these results establish the LLPS nature of DDX18 that is promoted by its RNA helicase activity and the RNA-interacting NIDR, suggesting that DDX18 may contribute to the LLPS of nucleoli in regulating RiBi and protein synthesis for pluripotency of hESCs.
DDX18 binds nucleolar localized RBPs and rRNAs/snoRNAs
To investigate how DDX18 exerts such nucleolar functions through its phase separation properties, we identified DDX18-interacting proteins and RNA targets in hESCs. First, we established a transgenic hESC line expressing 3xFlag-DDX18 for affinity purification of DDX18 and its interacting partners in the hESC nuclear extracts using anti-Flag beads (Fig. S3a). Then, with a stringent cut-off (number of unique peptides ≥4 and ratio of 3xFlag-DDX18/IgG ≥ 4), we identified 140 DDX18-interacting proteins (Supplementary Data 3). GO analysis of these proteins linked them to rRNA processing, translational initiation, and RNA secondary structure unwinding (Fig. 3a). As an important hub in maintaining genome stability, the nucleolus is highly enriched for DNA repair proteins, some of which were identified as DDX18-binding proteins with validated nucleolar localization and functional involvement in RiBi, including PARP145, SMC346, and TRIM2847. However, we cannot exclude the possibility that DDX18 may also play a role in DNA damage repair, a topic worthy of future investigation.
a Summary of the DDX18 interactome showing representative GO terms and corresponding proteins. Nucleolar GC marker proteins NPM1 and NCL are also shown under rRNA processing. b RNA-dependent interactions between DDX18 and NPM1. Western blotting detection of DDX18 and NPM1 protein levels after Flag-IP with and without RNase A treatment. IgG serves as the negative control IP. The experiment was repeated independently twice with similar results. c Representative live imaging of reconstituted BiFC fluorescence. Green fluorescence shows the reconstitution of YFP as an indicator of the protein-protein interaction between DDX18 and NPM1. The experiment was repeated independently twice with similar results. d Relative enrichment of different RNA species identified from DDX18 iCLIP-seq. e HOMER motif analysis of DDX18 iCLIP-seq showing the top consensus sequences. The conserved motif of the C/D box snoRNAs is shown at the top. P value was calculated by HOMER. f GO analysis of biological processes for DDX18 bound mRNAs. Source data are provided as a Source Data file.
Notably, the two well-known nucleolar GC marker proteins, NCL and NPM1, were also detected in the DDX18 interactome. We further confirmed the DDX18-NPM1 interaction by co-immunoprecipitation (co-IP) (Fig. 3b) and bimolecular fluorescence complementation (BiFC)48 (Figs. 3c, S3b), in line with their GC co-localization (Fig. 2b) and supporting a GC-specific function of DDX18. Importantly, the DDX18-NPM1 interaction was attenuated by RNase A treatment in co-IP (Fig. 3b) and BiFC (Fig. 3c) assays, supporting an RNA-facilitated interaction between DDX18 and NPM1.
To identify the RNA targets of DDX18 in hESCs, we performed iCLIP (individual-nucleotide-resolution crosslinking and immunoprecipitation) with the established 3xFlag-DDX18 hESC line (Fig. S3c). DDX18 binds to a diverse set of RNAs, among which rRNAs were most highly represented (Fig. 3d; Supplementary Data 4), while some snoRNAs were also bound by DDX18, consistent with its nucleolus-specific localization (Fig. 2a, b). Upon evaluating our results in relation to iCLIP-seq data for ADAR149, a well-studied nuclear-localized RBP in hESCs, we noted an elevated degree of rRNA and snoRNA binding to DDX18 over ADAR1 (compare Fig. 3d with Fig. S3d). Interestingly, the C/D box snoRNA motif RUGAUGA was ranked among the top three in motif analysis (Fig. 3e), and DDX18 binding to C/D box snoRNAs (SNORDs) was confirmed by CLIP-qPCR, exemplified by a few SNORD members (Figs. S3e, S3f, top panel). We also noticed that DDX18 could bind protein-coding mRNAs (Figs. 3d, S3f, bottom panel; Supplementary Data 5). Of note, mRNA of genes residing in DNA nucleolus-associated domains (NADs) are highly enriched in the nucleolus50, and we found that the DDX18-interacting mRNAs were also enriched for GO terms in RiBi and translation (Fig. 3f). Furthermore, unlike the DFC component FBL51, DDX18 does not bind to 1∼414 nt (5′ ETS-1) region of 47S pre-rRNAs localized in DFC (Fig. S3g), further supporting the nucleolar GC-specific functions of DDX18 in hESCs. These data establish physical and functional interactions between DDX18 and nucleolar RBPs and RNA species that underlie their roles in maintaining the pluripotency of hESCs.
DDX18 maintains nucleolar structural integrity through multiphase liquid miscibility and immiscibility with GC and DFC proteins, respectively
We next investigated how DDX18 may contribute to the structural integrity of the nucleolus, particularly in the context of DFC and GC. GC phase is known for enriching negatively charged proteins such as NPM1 and NCL, and phase separation of NPM1 is enhanced by positively charged peptides52,53. As DDX18 binds to NPM1 both in vitro and in vivo (Figs. 3a–c, S3a, b), we compared the biophysical properties of DDX18 and NPM1 by testing their droplet miscibility. We found that NPM1-RFP cannot form droplets in vitro without the PEG crowding reagent (Fig. 4a), as reported54. Interestingly, adding NPM1-RFP solution without droplets to the preformed DDX18 droplets, we observed a profound “core-shell” like structure containing homogeneous DDX18-GFP and NPM1-RFP mixture in the same compartment, phase-separated from the empty core structures, mimicking the three-layered structure of the nucleolus (Fig. 4b). The hollow-like condensates also underwent shell fusion and cavity fusion (see “Supplementary Movie”). In contrast, neither homogenous droplets nor “core-shell” structures were observed upon mixing preformed DDX18-GFP and RFP-FBL droplets together (Fig. S4a, b), verifying the immiscibility of DDX18 and FBL droplets. The observed differential miscibility of DDX18 with NPM1 and FBL droplets was consistent with NPM1 and FBL being the most representative components of the phase-separated GC and DFC regions within the nucleolus, respectively38. To further probe the potential contribution of DDX18 to nucleolus architecture through its multiphase liquid (im)miscibility, we simultaneously mixed DDX18-GFP, NPM1-CFP, and RFP-FBL recombinant proteins. We observed a co-existing multiphase structure mimicking the regular nucleolus structure (Fig. 4c, top panel). As RNA helicase activity is critical for DDX18 functions (Figs. 1i, S2h–j), we tested how DDX18-bound RNAs (rRNA and snoRNA) may contribute to forming this particular multiphase structure. Intriguingly, the introduction of snoRNA or rRNA leads to a significant increase in the proportion of structures resembling the nucleolus (Fig. 4c, middle and bottom panels). This observation suggests a dynamic interplay between these RNA components and the formation of nucleolar-like structures.
a, b Biophysical properties of the mixtures of purified DDX18-GFP (20 μM) and NPM1-RFP (10 μM) recombinant proteins in vitro in a buffer comprised of 150 mM NaCl, 25 mM Tris-HCl, pH 8.0. Adding NPM1-RFP solution without PEG (i.e., no droplets formed) to the preformed DDX18 (a) established droplets with “core-shell” like structure, imaged with RFP, GFP, and merging RFP/GFP (b). The experiment was repeated independently twice with similar results. c Biophysical properties of the mixtures of purified DDX18-GFP (20 μM), NPM1-CFP (10 μM), and RFP-FBL (20 μM) with or without rRNA/snoRNA in a buffer comprised of 150 mM NaCl, 25 mM Tris-HCl, pH 8.0, detected by GFP, CFP, and RFP, respectively. Merged images and quantification of the formation of indicated core-shell structures are shown, where the percentages reflect the proportion of fields of view exhibiting the core-shell structures. The experiment was repeated independently twice with similar results. d Images of normal (shLuc) and disrupted (shDDX18#1) nucleolar condensate organization. FBL labels the dense fibrillar component (DFC); NPM1 labels the granular component (GC). The experiment was repeated independently twice with similar results. e Quantification of immunostaining images from control and KD cells (shDDX18#1 and shDDX18#3) showing the percentage of inclusive DFC localized cells (represented by the top part of panel d) or exclusive DFC localized cells (represented by the bottom part of (d). Quantifications were performed using 152 cells from shLuc, 211 cells from shDDX18#1, and 202 cells from shDDX18#3, based on three technical replicate cultures. Data are presented as mean values (pm) SD. f A proposed model showing that DDX18 coordinates with NPM1 to safeguard nucleolus organization. (Created in BioRender. Malik, V. (2025) https://BioRender.com/v27l155). Source data are provided as a Source Data file.
To further understand its nucleolar functions in maintaining hESC pluripotency in vivo, we addressed how DDX18 depletion may modulate nucleolar architecture by immunostaining and 3D-SIM imaging in control and DDX18 KD hESCs. In hESCs with control KD (shLuc), the DFC marker FBL is completely encompassed (or included) in the NPM1-marked GC as immiscible multiphase droplets (Fig. 4d, top). In contrast, upon DDX18 depletion (shDDX18), the DFC region starts “leaking/escaping” (or excluded) from the GC and forming “nucleolar cap” or “nucleolar segregation” like structures (Fig. 4d, bottom) without significant changes of the core nucleolar protein levels including the GC markers NPM1 and NCL, and the DFC marker FBL (Fig. S4c, d). Quantification of the representative images on such “inclusive/exclusive” structures indicated that more than 40% of the DDX18 KD hESCs, compared to less than 10% of the control KD hESCs, showed mislocalization of FBL (Fig. 4e). Importantly, mislocalization of FBL, coincident with nucleolar enlargement, was also observed more prominently in the mesendoderm lineage than in the ectoderm lineage upon the directed trilineage differentiation of hESCs (Fig. S4e, f).
Our data thus establish a critical role of DDX18 in maintaining hESCs by preserving the structural integrity of the nucleolus through its multiphase liquid miscibility and immiscibility with the key GC and DFC proteins NPM1 and FBL, respectively (Fig. 4f, left). DDX18 depletion in hESCs, under the KD or differentiation condition, results in the mislocalization of DFC as “nucleolar caps” protruding from the GC (Fig. 4f, right) and lineage-specific nucleolar enlargement (Fig. S4e), causing the defects in RiBi and translation and, consequently, the loss of hESC pluripotency with the skewed differentiation towards mesendoderm lineage.
DDX18 restricts perinucleolar heterochromatin organization via its liquid miscibility and immiscibility with NPM1 and HP1 droplets, respectively
Nucleolar structure alteration is reported to cause chromatin reorganization55,56,57. During ESC differentiation, the initially dispersed H3K9me3-marked and HP1-containing heterochromatin foci (i.e., PNH) start accumulating around nucleoli5,20,58,59. Tethering heterochromatin at nucleoli promotes transcriptional activation of differentiation genes, leading to the exit of pluripotency in mESCs20. We thus asked whether the PNH formation could be facilitated by DDX18 KD, leading to differentiation of hESCs as we observed (Fig. 1). Indeed, we found that DDX18 depletion led to the redistribution of dispersed HP1α/β foci, characteristic features of heterochromatin, from nucleoplasm in control cells (shLuc) closer to nucleolus in DDX18 KD cells (shDDX18) (Fig. 5a, b), colocalizing with heterochromatin-associated H3K9me3 (Fig. S5a), indicative of the PNH accumulation. These findings establish the requirement of nucleolus structural integrity for restricting PNH formation and maintaining the pluripotent dispersed heterochromatin state.
a, b DDX18 depletion leads to perinucleolar heterochromatin organization in hESCs. Representative images of immunostaining for nucleolar NPM1, nucleolus-associated heterochromatin markers HP1α and HP1β, nuclear DAPI, and their merge are shown (a). The yellow arrows, labeled from I to II, indicate the direction of ImageJ immunofluorescence signal quantification shown as line charts (b). Black arrows indicate that HP1α and HP1β can dock at or enter the nucleolus when DDX18 is depleted. The experiment was repeated independently twice with similar results. c Images of the mixtures of purified recombinant proteins DDX18-GFP (10 μM) and RFP-HP1α (40 μM) (top) and NPM1-CFP (10 μM) and RFP-HP1α (40 μM) (bottom) in a buffer comprised of 150 mM NaCl, 25 mM Tris-HCl, pH 8.0 and 5% PEG8000, detected by respective reporters. The experiment was repeated independently twice with similar results. d, e Images of the mixtures of purified DDX18-GFP (10 μM), NPM1-CFP (10 μM), and RFP-HP1α (40 μM) in a buffer comprised of 150 mM NaCl, 25 mM Tris-HCl, pH 8.0 and 5% PEG8000, detected by respective reporters under immunofluorescence microscope (d) and the line chart showing the ImageJ quantified intensity of NPM1-CFP, RFP-HP1α and DDX18-GFP fluorescence signals (e) along the direction indicated in (d). The experiment was repeated independently twice with similar results. f, g DDX18 KD results in centromere clustering around the nucleolus. Representative images for nucleolus (NPM1, red), centromeres (CENPA, green), and DNA (DAPI, blue) (f) quantification of the number of nucleolus-localized centromeres under control and DDX18 KD conditions (g) are shown. Each dot in g represents one cell. ***P < 0.001 (The P-values were calculated using a two-sided t-test and indicated). h ChIP-qPCR analysis of NPM1 genomic occupancy at the centromere locus in control (shLuc) and DDX18 KD (shDDX18#1, shDDX18#3) hESCs. Data are mean±SD of three independent experiments. Chromosome Y serves as a negative control for female H9 hESCs. IgG serves as the ChIP negative control. i A proposed model showing that DDX18 restricts centromere from clustering heterochromatin organization around the nucleolus. (Created in BioRender. Malik, V. (2025) https://BioRender.com/v27l155). Source data are provided as a Source Data file.
To explore the mechanisms by which DDX18 restricts PNH formation, we first sought to analyze the (im)miscibility of the droplets formed by GC phase proteins and the heterochromatin protein HP1α, given that HP1α phase separation regulates the liquid droplet-like properties of heterochromatin60,61,62,63. By mixing RFP-HP1α with DDX18-GFP and NPM1-CFP, respectively, we observed that HP1α droplets were miscible with both DDX18 and NPM1 in droplet formation (Fig. 5c). NPM1 contains two evolutionarily conserved HP1α binding motifs (PXXVXL and PXVXL) (Fig. S5b) with reported roles in regulating HP1α oligomerization and promoting the formation of phase-separated droplets of HP1α60, explaining the miscibility between HP1α and NPM1 droplets. The fact that PNH is not accumulated in undifferentiated hESCs despite normal expression levels of heterochromatin HP1 proteins suggests additional factors may act upon HP1α and NPM1 to restrict their liquid miscibility. We thus tested DDX18 action in this context by simultaneously mixing DDX18-GFP, NPM1-CFP, and RFP-HP1α recombinant proteins to explore their multiphase organization behaviors. We found that NPM1 preserves the homogeneous droplet formation with DDX18 (merged image of the cyan droplets in Fig. 5d), consistent with our prior in vitro observation (Fig. 4c). However, NPM1 suddenly lost its liquid miscibility with HP1α when DDX18 was present with HP1α being pushed outside of the DDX18/NPM1 condensates, resulting in the formation of a shell-like structure (Fig. 5d, e). These data indicate that, while maintaining its liquid miscibility with NPM1 droplets for GC integrity, DDX18 prevents the HP1α-NPM1 droplet miscibility, hindering the PNH accumulation at GC (see Discussion).
Next, we addressed why and how the heterochromatin prefers to dock/accumulate at the nucleolus region during ESC differentiation after DDX18 depletion. As multiphase organization could be regulated by interconnected complexes (nodes) of multivalent proteins63,64, we sought to find the connecting node between PNH and the nucleolus. In mESCs, it has been reported that transcriptionally repressive regions near centromeres are often closer to the nucleolus21. Within the human genome, 45S rDNA tandem repeats are localized as rDNA gene arrays at the short arms of five chromosomes (Chr13, 14, 15, 21, and 22) flanked by the centromere and telomere (Fig. S5c). We detected centromere positioning by staining with the centromere markers CENPA and CENPT and observed a significant increase of centromere clustering around the nucleolus upon DDX18 depletion (Figs. 5f, g, S5d, e). Genome-wide occupancy of DDX18 by ChIP-seq revealed that DDX18 mainly binds to the rDNA transcribed regions (Fig. S5f), which is consistent with our previous study in mESCs19. In contrast, it is well established that the heterochromatin proteins HP1α/β bind to centromeres/pericentromeres65 and that the GC protein NPM1 binds to the centromere66,67,68,69. Thus, we hypothesized that centromere and centromere-binding proteins from the nucleolus might provide the connecting node for PNH to dock around the nucleolus. Indeed, we found that DDX18 depletion resulted in increased centromere binding of NPM1 within hESCs (Fig. 5h).
Together, these data support a model whereby DDX18 coordinates with NPM1 to safeguard nucleolar GC integrity and restrict centromere clustering and PNH formation (Fig. 5i, left). When DDX18 is absent or depleted, leading to the nucleolar cap formation (Fig. 4d, e), NPM1 droplets become miscible with HP1 droplets with increased centromere binding, promoting heterochromatinization (Fig. 5i, right), an early event required for ESC differentiation20.
DDX18 depletion results in nuclear reorganization
To address how chromatin may be reorganized at a genome-wide scale contributing to changes in gene expression and, consequently, the differentiation of DDX18 KD hESCs, we applied the Hi-C seq technology to generate chromatin interaction data in control and DDX18 KD cells. Hi-C data allow the partition of the genome into two compartments called A/B compartments, containing relatively active (A) and inactive (B) regions, respectively64. The saddle plots (Fig. S6a) revealed that DDX18 depletion leads to moderate changes in A/A (an increase from 1.369 to 1.419) and B/B (a decrease from 1.346 to 1.307) homotypic interactions, indicating a decrease in chromatin interaction within heterochromatin. In contrast, more minor changes in A/B or B/A heterotypic interaction were observed (Fig. S6a). Further using principal component analysis (PCA) at 100-kb resolution to classify the genome into A and B compartments, we identified regions that changed compartments upon DDX18 KD. We found 3.2% of regions switched from A to B and 3.3% from B to A (Fig. S6b). However, integrating these results with RNA-seq data showed that over 90% of significantly up- or down-regulated genes were not in compartment-switching regions (Fig. S6c). We then analyzed gene expression changes in 100-kb bins based on their PC1 value changes, indicating shifts towards active or inactive states. We found that a small proportion of bins with extreme PC1 changes correlated with significant gene regulation, with only 195 and 17 genes (out of >2000 differentially expressed) in bins showing PC1 changes greater than −2 and +2, respectively (Fig. S6d). Further analysis of upregulated lineage-associated genes showed most were in regions with moderate PC1 changes (−0.5 to +0.5) (Fig. S6e). This weak correlation between PC1 changes and transcription was also observed globally (Fig. S6f). Together, our results suggest that, while locus-specific transcriptional changes resulting from the compartment switch do occur (Fig. S6g), a vast majority of transcriptional changes observed upon DDX18 KD arise from mechanisms independent of compartment organization. This observation is consistent with the fact that DDX18 is an RBP rather than a typical transcription factor or chromatin remodeler and that DDX18 mainly binds to the rDNA transcribed regions in the nucleolus (Fig. S5f). Hence, it is more likely to influence gene expression via its unique nucleolar functions.
A/B compartments are primarily defined by intrachromosomal interactions within each chromosome70. By analyzing the intrachromosomal interaction for all five rDNA-containing chromosomes (Chr13, 14, 15, 21, and 22), we only found two interacting regions (R1 and R2) of Chr21 showing significantly increased contact frequencies after DDX18 KD (Figs. 6a, S6h). Both interacting regions are close to the pericentromere. R1 is mediated by rDNA closer locus (Site I, Chr21: 9.4 MB-10.1 MB) and chromosome arm locus (Site III, Chr21: 24.9 MB-26.5 MB), while R2 is mediated by pericentromeric locus (Site II, Chr21: 14.7 MB-15.1 MB) and chromosome arm locus (Site III, Chr21: 24.9 MB-26.5 MB) (Fig. 6b). Interestingly, we found that Site III is localized in the lamina-associated domain (LAD)71 (Fig. S6i). The increased interactions of R1 and R2 in DDX18 KD relative to control KD samples suggest that DDX18 depletion could affect nucleolus-associated domain (NAD) and LAD exchange (Fig. 6c). Together with our findings that DDX18 depletion impairs nucleolus integrity (Fig. 4) and PNH organization (Fig. 5), we concluded that DDX18 could play a pivotal role in regulating centromere-telomere proximal rDNA-containing chromosome (i.e., Chr21) organization, i.e., heterochromatin formation.
a, b Hi-C interaction heatmaps (a) and schematic diagram (b) depicting increased chromatin interactions at Chr21 region R1 (I↔III) and R2 (II↔III) in DDX18 KD vs. control KD hESCs. c A proposed model. See the main text for a detailed explanation. d Boxplots showing the change in the strength of significant interactions (the numbers of interactions are indicated) upon DDX18 KD categorized by chromosome arms. The edges of the box represent the 25th and 75th percentile, while the line in the middle of the box is the median. The whiskers extend from the box edges to 1.5 times the interquartile range. e Heatmap showing log10(ratio) of mean normalized interchromosomal contacts per 250-kb bin in DDX18 KD vs. control KD hESCs. Acrocentric chromosomes are highlighted in black boxes. f Illustration of the contact domains and chromatin loops for a 3 Mbp region of chromosome 7p containing HOXA cluster genes from Hi-C data with the epigenetic modifications of the HOXA cluster from public datasets shown. Circles on the matrix denote increased interaction frequency of the HOXA cluster genes with nearby genes in DDX18 KD vs. control KD hESCs. g, h Quantification (g) and representative DNA FISH images (h) of the control and DDX18 KD cells for HOXA9-13 alleles (red) compared with nucleolar (NPM1, green) and nucleoplasm (DAPI, blue). The total numbers (n) of nuclei from two independent experiments were scored. i, j Representative confocal microscopy merged images of RNA-FISH detecting HOXA9 (i) and HOXA11 (j) mRNAs in control and DDX18 KD H9 hESCs. Dotted circles indicate nucleoli detected by Nucleolus Bright Green Dye reaction to predominantly nucleolar and cytoplasmic RNAs. k, l Quantification of HOXA9 (k) and HOXA11 (l) mRNAs RNA-FISH signals at nucleolar (NAD) and nuclear periphery (LAD) proximities in control and DDX18 KD hESCs (representative cells are shown in i, j). RNA-FISH signals outside the nuclear periphery were not counted. Equal numbers of shLuc and shDDX18 cells (N = 39 for HOXA9 and N = 30 for HOXA11), randomly selected from three independent experiments, were analyzed for signal quantifications (mean ± SD). All cells were counterstained with DAPI. The red and green arrows indicate the increase and decrease trends, respectively. Source data are provided as a Source Data file.
Since the nucleolus is a hub for interchromosomal interactions, we next examined the contacts between ribosomal DNAs (rDNAs) from nucleolar organizer region (NOR)-bearing chromosomes (Chr13, 14, 15, 21, and 22). The analysis of chromatin interactions in the short arms of NOR-bearing acrocentric chromosomes is challenging due to their repetitive nature and gaps in the human reference genome GRCh38. Utilizing the mHi-C pipeline72, which retains multimapping reads, enabled us to analyze rDNA interactions, revealing improved detection in the repetitive short arms of acrocentric chromosomes 15, 21, and 22 (Fig. S6j). However, gaps in the GRCh38 reference genome (Fig. S6k top) still limited the analysis. Remapping Hi-C data to the updated T2T-CHM13 reference genome73 allowed us to fill many gaps, enhancing the resolution of contact maps in acrocentric chromosomes (Fig. S6k bottom). Focusing on the impact of DDX18 KD, we detected significant changes in chromatin organization, with 31% of interactions shared between control KD and DDX18 KD conditions (Fig. S6l). Next, to specifically study the impact of DDX18 KD on rDNA-containing chromosomal interactions, we initially calculated the change in interaction strength (observed/expected counts) upon DDX18 KD of all shared interactions across the genome and categorized our results by chromosome arm. Interestingly, this analysis showed a preferential weakening of Hi-C interactions, specifically in the short arms of acrocentric chromosomes containing rDNA repeats. In contrast, long arms and non-acrocentric chromosomes showed no significant change (Fig. 6d). Importantly, an increase in interchromosomal contacts among acrocentric chromosomes was observed (Fig. 6e and Fig. 6c model), suggesting that DDX18 plays a crucial role in organizing NORs.
Nuclear reorganization upon DDX18 depletion results in developmental gene expression changes
To understand how nuclear reorganization involving LAD-NAD exchange and NORs may lead to gene expression changes in DDX18-depleted hESCs, we studied the impact of DDX18 KD on the HOXA gene cluster for the following reasons. First, the HOXA gene cluster is in a gene-poor region of Chr7 that, in human cancer cells, is sequestered by the Nup93 sub-complex for HOXA repression at the nuclear periphery (LAD). The derepression of HOXA genes upon Nup93 depletion results in the LAD to nuclear interior (closer to NAD) relocation74,75. Second, the HOXA gene cluster represents the CTCF-mediated chromatin loops and chromatin status changes during embryonic development and disease progression76,77,78. Intriguingly, a published study reported the interchromosomal interaction between the NOR-bearing Chr21 and HOXA gene-residing Chr7 and its resultant t(7;21) translocation in a human leukemia patient79. Third, the HOXA cluster genes were upregulated in our DDX18 KD hESCs (Fig. 1h) coincident with weakened interchromosomal interactions of HOXA-residing Chr7 with Sites II and III in Chr21 (Fig. S6m), suggesting that the DDX18 KD effect on the 3D conformation of those NOR-bearing chromosomes (e.g., Chr21) could indirectly affect chromatin organization (e.g., Chr7) and developmental gene expression (e.g., HOXA genes) (see Fig. 6c model).
To substantiate our model whereby DDX18 depletion could cause the NAD and LAD exchange and HOXA gene derepression (Fig. 6c), we performed the following studies. First, we performed additional Hi-C data analysis on chromatin interactions involving the HOXA locus and revealed that, despite the lack of a major compartment switch in the DDX18 KD hESCs (Fig. S6a–c), loops involving the HOXA locus within Chr7 appeared strengthened in the DDX18 KD hESCs (Fig. 6f). In addition, DDX18 KD hESCs showed reduced interaction frequencies within the HOXA gene cluster (Fig. S6n, o, left panels) and increased the external interactions between the HOXA cluster and the nearby genes (e.g., HOTTIP) (Fig. S6n, o, right panels). Furthermore, DNA FISH analysis revealed that HOXA9-13 loci are relocated away from the nucleolus (NAD) (i.e., reduced FISH signals) towards the nuclear periphery (LAD) (i.e., increased FISH signals) in DDX18 KD relative to control KD hESCs (Fig. 6g, h). Importantly, RNA FISH analysis also detected the corresponding upregulation of HOXA transcripts, exemplified by increased RNA FISH signals per nucleus of HOXA9 (Fig. 6i, k) and HOXA11 (Fig. 6j, l) at the nuclear periphery (LAD), while their NAD signals were slightly and appreciably reduced. These findings demonstrate that DDX18 depletion modulates the heterochromatin via the LAD and NAD exchanges associated with the NOR-bearing chromosomes, resulting in nuclear reorganization and derepression of developmental genes (e.g., Chr7 reorganization at the HOXA locus; Fig. 6c model).
Nucleolus-specific NoCasDrop restructures the genome and controls gene expression
We have thus far established the roles of DDX18 in maintaining the intrinsic nucleolar architecture (Fig. 4f) and extrinsic heterochromatin organization (Fig. 5i) for pluripotency and lineage commitment/differentiation of hESCs. However, it remains an open question whether the extrinsic mechanism, i.e., centromere clustering and PNH accumulation, is the cause or consequence of hESC differentiation. Therefore, inspired by CasDrop, a recently developed method successfully applied in mammalian cells to activate gene expression80, we created NoCasDrop (Nucleolus specific CasDrop) to address whether tethering heterochromatin towards nucleolus would impair hESC pluripotency.
To this end, we took advantage of the nucleolus-specific phase separation by fusing full-length (FL) DDX18 and its individual domains (NIDR, HD, and CIDR) with the dCas9-mCherry (Figs. 7a, S7a). We found that both FL and NIDR of DDX18 fusion, but not the other domains, could recruit dCas9-mCherry protein to the nucleolus (Fig. S7a). However, FL, but not NIDR, could activate the pre-rRNA transcription (Fig. S7b). Therefore, to avoid any indirect effects resulting from rRNA overexpression, we used NIDR NoCasDrop for our further study (Fig. 7a). To overcome the challenge of manipulating a whole trunk of chromosome fragment by a single dCas9, we targeted tandem repeats in centromere and satellite DNA by infecting the NIDR NoCasDrop stable cell line with a human centromere-specific alpha-satellite sgRNA (sgAlpha) and a non-specific control sgRNA (sgNS), both of which have been functionally validated in human cell lines previously81. This allows multiple copies of dCas9 to coordinately target the same gene locus by the same sgRNA leading to a combined/concerted action in controlling the chromosome localization (Fig. 7a). Then, we performed immunostaining after 48 h of sgRNA infection and found that the centromere (Fig. 7b, c) and heterochromatin maker HP1α (Fig. 7d, e) could be tethered to the perinucleolar region by NoCasDrop and α-satellite specific sgRNA. More importantly, we found the derepression of developmental genes (e.g., HOXA cluster genes) by this NoCasDrop-mediated perinucleolar centromere clustering (Fig. 7f). In contrast, pluripotency genes and RiBi genes were minimally affected within the tested time window (Fig. S7c), indicating the specificity of NoCasDrop and corroborating the Hi-C data. These results also suggest that the derepression of HOXA cluster genes is an early event leading to the loss of pluripotency in DDX18 KD hESCs.
a Schematic diagram of NoCasDrop (Nucleolus CasDrop). Repetitive loci such as centromeres can be tethered by α-satellite specific sgRNA and dCas9-mCherry-NIDR (DDX18) to nucleolus hub in vivo. (Created in BioRender. Malik, V. (2025) https://BioRender.com/v27l155). b, c Alpha (α)-satellite sgRNAs (sgAlpha) and NoCasDrop recruit centromere to the nucleolus. Representative images for nucleolus (NPM1, red), centromeres (CENPA, green), and nucleus (DAPI, blue), as well as merged images (b) and quantification of the number of centromere foci around nucleolus from more than 50 cells (c) are shown. Each dot in c represents one cell. ***P < 0.001 (The P-value was calculated using a two-sided t-test). sgNS acts as a non-specific sgRNA control. d, e NoCasDrop with sgAlpha recruits perinucleolar heterochromatin (HP1α, magenta) at the nucleolus (NPM1, green) in hESCs. Representative images of immunostaining for nucleolar NPM1, nucleolus-associated heterochromatin markers HP1α and HP1β, nuclear DAPI, and their merge are shown (d). The yellow arrows, labeled from I to II, indicate the direction of ImageJ immunofluorescence signal quantification, as shown in line charts (e). The experiment was repeated independently twice with similar results. f qRT-PCR analysis of HOXA cluster genes upon sgAlpha and NoCasDrop treatment. sgNS acts as a non-specific control. g, h A summary model. Under the pluripotent state, DDX18 is highly expressed. It maintains the nucleolar structural integrity while restricting perinucleolar heterochromatin (PNH) formation through its strong liquid miscibility with the core GC factor NPM1 and relatively weaker miscibility with heterochromatin protein HP1 and immiscibility with the core DFC factor FBL (g). When DDX18 is depleted (e.g., through genetic manipulation) or downregulated (e.g., during hESC differentiation), NPM1 increases its binding to the centromere and acquires the liquid miscibility with HP1, resulting in the escape of DFC from the center of the nucleolus and the formation of “nucleolar caps” with accumulated PNH and altered nuclear organization. Consequently, intrachromosomal and interchromosomal reorganization leads to the derepression of developmental genes (e.g., HOXA) (h). (Created in BioRender. Malik, V. (2025) https://BioRender.com/v27l155). Source data are provided as a Source Data file.
These data reinforce a major function of DDX18 in restricting centromere clustering around the nucleolus for the repression of developmental genes (e.g., HOXA cluster genes) in undifferentiated hESCs and provide a molecular tool to tether centromere clustering at the perinucleolar region for cell fate manipulation.
- SEO Powered Content & PR Distribution. Get Amplified Today.
- PlatoData.Network Vertical Generative Ai. Empower Yourself. Access Here.
- PlatoAiStream. Web3 Intelligence. Knowledge Amplified. Access Here.
- PlatoESG. Carbon, CleanTech, Energy, Environment, Solar, Waste Management. Access Here.
- PlatoHealth. Biotech and Clinical Trials Intelligence. Access Here.
- Source: https://www.nature.com/articles/s41467-024-55054-8