Selective RNA sequestration in biomolecular condensates directs cell fate transitions

P-body purification

To profile P-body contents, we adapted a fluorescence-activated sorting method^29,42 using a green fluorescent protein (GFP)-LSM14A expression construct in HEK293T cells (Fig. 1a). LSM14A is an established protein component of P-bodies²⁹, and GFP-LSM14A puncta colocalized with the P-body marker EDC4 (ref. ⁴³) (Fig. 1b). After cell lysis, intact GFP-LSM14A particles were detectable by microscopy and were readily isolated using fluorescence-activated particle sorting (FAPS) (Fig. 1c and Extended Data Fig. 1a). By contrast, particles were undetectable in cells expressing a cytoplasmic GFP control (Fig. 1c and Extended Data Fig. 1a). To corroborate these data, we depleted DDX6 in GFP-LSM14A HEK293T cells to disrupt P-bodies^19,21,28,32. Both short hairpin RNA (shRNA)-mediated suppression of DDX6 and CRISPR knockout (KO) resulted in the complete loss of GFP⁺ particles (Fig. 1d,e and Extended Data Fig. 1b,c). To confirm the presence of RNA in GFP-LSM14A particles, we stained lysates from GFP-LSM14A-expressing cells with SYTOX Blue, an RNA-binding fluorescent dye. We observed RNA in GFP⁺ particles but not in GFP⁻ particles (Extended Data Fig. 1d). These findings confirm the robust labeling and isolation of P-bodies using GFP-LSM14A.

We characterized RNA from purified P-bodies and corresponding cytosolic fractions using Smart-seq⁴⁴ (hereafter P-body-seq). We detected 3,994 mRNAs enriched in P-bodies of HEK293T cells relative to the cytosol (Fig. 1f). Gene ontology (GO) analysis revealed that P-body-enriched mRNAs encoded regulators of RNA processing, transcription, chromatin organization and cell cycle (Fig. 1g), while cytosolic mRNAs were involved in housekeeping functions such as metabolic processes and structural components (Fig. 1h). To ensure that our analyses were not biased by polyA-directed library preparations, we performed snapTotal-seq⁴⁵, an alternative low-input library preparation that uses random primers rather than oligo(dT). snapTotal-seq results were highly consistent with Smart-seq data (~70% overlap; Extended Data Fig. 1e). Moreover, we compared our dataset to P-body-enriched transcripts previously reported in HEK293T cells²⁹ and found that 3,038 mRNAs (76%) were shared between the datasets (Extended Data Fig. 1f). To confirm P-body-seq results using an orthogonal approach, we employed single-molecule fluorescence in situ hybridization (FISH) (smFISH) in conjunction with immunofluorescence (IF)⁴⁶ and verified the localization of POLK and TET2 within P-bodies (Fig. 1i,j and Extended Data Fig. 1g). In line with previous reports^26,38, approximately 70% of total POLK and TET2 mRNA was localized to P-bodies determined using either smFISH or P-body-seq, demonstrating substantial transcript sequestration in P-bodies.

P-bodies have been proposed as sites for RNA decay^20,21,47,48; however, P-body enrichment showed poor correlation with transcript half-life measurements in HEK293T cells⁴⁹, consistent with previous reports²⁹ (Extended Data Fig. 1h). Read distribution analysis also showed no evidence of increased truncated transcripts in P-bodies relative to the cytoplasm (Fig. 1k), in line with previous reports^26,29,38. In addition, we observed poor correlation between mean polyA tail length⁵⁰ and P-body enrichment (r = −0.028 in Smart-seq data and r = −0.075 in snapTotal-seq data) (Fig. 1l), indicating that P-body-localized RNAs are intact and are not preferentially deadenylated relative to cytoplasmic transcripts.

We next asked whether properties intrinsic to RNAs correlated with P-body enrichment. Transcript length had no appreciable relationship with P-body localization, but transcripts with high AU content were enriched among P-body-targeted mRNAs (Extended Data Fig. 1i). Notably, AU-rich sequences are associated with inefficient translation^28,51,52. Accordingly, analysis of published ribosome profiling data obtained in HEK293T cells⁵⁰ revealed reduced translation efficiency for P-body-associated transcripts (Fig. 1m). These data suggest that transcripts sequestered in P-bodies are translationally repressed compared to cytoplasmic-enriched mRNAs.

P-body contents are cell type specific

P-body-associated transcripts can reenter translation following genetic or environmental perturbation^{19,25,29,51,53,54}, suggesting that their sequestration may influence developmental fate decisions. Yet, defining P-body function in complex biological processes has been challenging, in part because sequencing-based approaches to define P-body contents have been limited to nonvertebrates or transformed cell lines^25,29,55. We thus applied P-body-seq to profile transcripts in various human cell types across developmental stages. We stably integrated GFP-LSM14A at the AAVS1 locus in human ES cells and cultured them under naive and primed conditions, which model the pre- and post-implantation epiblast, respectively⁵⁶. We also induced differentiation of the same human pluripotent stem cell line into progenitors of all three germ layers (mesoderm, endoderm and neural progenitors) and ultimately into neurons (Fig. 2a). IF confirmed robust differentiation into the expected lineages (80% efficiency or greater; Extended Data Fig. 2a), which was supported by the expression of lineage-specific genes (Extended Data Fig. 2b). We detected EDC4⁺ P-bodies in each cell type, albeit in varying amounts (Extended Data Fig. 2c,d), suggesting differences in RNA sequestration between cell types.

To profile P-body contents, we sorted GFP-LSM14A⁺ particles from each cell type. Particle size and fluorescence varied between cell types (Fig. 2b), likely reflecting differential expression of GFP-LSM14A from the AAVS1 locus, but, in all cases, we isolated sufficient RNA for P-body-seq. Supporting our observations in HEK293T cells, analysis of half-life data in human pluripotent stem cells⁵⁷ suggested that RNA in P-bodies is not preferentially degraded compared to cytoplasmic RNA (Extended Data Fig. 2e). We observed diverse RNA biotypes in P-bodies, including protein-coding mRNA and lincRNA (Extended Data Fig. 2f). We note, however, that RNA sequestration in P-bodies is not merely a reflection of expression level, as the expression of P-body-enriched transcripts was distributed similarly relative to that of cytoplasmic and unenriched transcripts (Extended Data Fig. 2g). We likewise identified repetitive elements in P-bodies, although we did not detect marked differences in their abundance between P-bodies and the cytoplasm (Extended Data Fig. 2h,i). Based on principal-component analysis, we observed unique mRNA profiles for P-bodies from each cell type, suggesting that distinct cell identities have unique transcripts stored in P-bodies (Fig. 2c). Principal-component analysis further suggested a stepwise progression from naive ES cells toward mesoderm and/or endoderm lineages on one path and the ectoderm lineage on another (Fig. 2c). Notably, we found that P-body contents did not necessarily cluster most closely with the cytoplasm from the same cell type (Fig. 2c). For example, P-body samples from neurons cluster most closely with cytoplasmic samples from neural progenitors and P-body contents of mesoderm progenitors cluster with the cytoplasm of primed ES cells (Fig. 2c). These data suggest that P-bodies sequester transcripts from stem and progenitor cells to suppress their protein-level expression during differentiation.

P-body contents reflect preceding human developmental stages

We grouped transcripts based on their enrichment across cell types (Fig. 2d and Extended Data Fig. 2j). A fraction of transcripts was enriched in the P-bodies of all cell types, and GO categories linked to DNA damage and cell cycle progression were common across cell types (Fig. 2e). By contrast, transcripts enriched in the P-bodies of single cell types encoded fate-instructive factors, such as LEUTX and DUXA in naive ES cells (Fig. 2f). Both LEUTX and DUXA are associated with totipotency and the eight-cell (8C) embryo^58,59, the developmental stage preceding naive pluripotency. To explore whether P-bodies sequester transcripts important for stem and progenitor cell maintenance during differentiation, we identified P-body-enriched transcripts related to cell fate and examined their localization in differentiated cell types and corresponding developmental precursors (Fig. 2g). P-bodies of differentiated cell types contained transcripts associated with the preceding developmental state, while these transcripts were largely cytoplasmic or unenriched in precursor cells (Fig. 2g and Discussion). smFISH analysis of OCT4 (POU5F1) in human primed ES cells and endoderm progenitors confirmed these findings, with OCT4 transcripts transitioning from primarily cytoplasmic localization in primed ES cells to P-bodies during endoderm differentiation (Extended Data Fig. 3a,b). We also observed a significant overlap between transcripts that were more expressed in precursor cells and P-body enriched in differentiated cells (Fig. 2h–j). Finally, we conducted gene set enrichment analysis (GSEA) using differential gene expression data from P-body and cytoplasmic fractions for each cell type (Fig. 2k,l and Extended Data Fig. 3c). In neurons, we observed robust enrichment of a transcriptional signature characteristic of neural progenitor cells in the P-body fraction (Fig. 2k). Consistent with these findings, P-body dissolution impaired neuronal specification from neural progenitors^28,35,36, providing functional evidence that P-bodies sequester genes related to stem and progenitor cell self-renewal to facilitate differentiation. Conversely, a neuronal transcriptional signature was enriched in the cytoplasm of neurons, where these transcripts are expected to be available to translational machinery (Fig. 2k).

Next, we assessed the functional requirement of P-bodies in endoderm differentiation using a CRISPR interference (CRISPRi)-based approach to deplete DEAD-box helicase 6 (DDX6)²⁸ and dissolve P-bodies¹⁹ (Extended Data Fig. 3d). DDX6 depletion significantly impaired differentiation toward AFP⁺ hepatocytes (Extended Data Fig. 3e) and suppressed the induction of other hepatic differentiation markers such as ALB, G6PC and CYP3A7 compared to control cells (Extended Data Fig. 3f). These results suggest that P-bodies are essential for differentiation to hepatocytes.

To test whether P-body-associated transcripts eventually decline after differentiation is completed, we extended neuronal differentiation to 20 d (Extended Data Fig. 3g). Analysis of P-body contents in mature neurons showed substantial overlap (~50%) with transcripts sequestered from neurons at day 7 of differentiation (Extended Data Fig. 3h); however, prolonged culture led to progressive depletion of neural progenitor-associated transcripts from P-bodies (Extended Data Fig. 3i). Similar patterns were observed in mature endoderm progenitors, which showed decreased enrichment for primed ES cell expression profiles compared to newly differentiated endoderm progenitors (Extended Data Fig. 3j,k).

We next examined P-body-associated transcripts across pluripotent states. In primed ES cells, P-bodies showed enrichment of naive ES cell-specific gene signatures (Extended Data Fig. 3c), consistent with observations that P-body dissolution promotes primed-to-naive state conversion²⁸. Similarly, P-bodies in naive ES cells were enriched for gene expression signatures characteristic of the preceding developmental stage, including zygotic genome activation (ZGA), the 8C embryo and 8C-like cells (8CLC)^58,59 (Fig. 2l and Extended Data Fig. 3l). While naive human ES cells typically express low levels of ZGA- and 8CLC-related transcripts⁵⁸, our data suggest that these transcripts are sequestered into P-bodies to prevent their translation and inhibit inappropriate reversion to an earlier developmental state. Conversely, naive pluripotency transcripts were enriched in the cytoplasm relative to P-bodies (Fig. 2l), likely reflecting active expression of genes supporting the naive cell fate. These findings suggest that RNA condensates in ES cells selectively sequester totipotency-associated factors. Altogether, these data indicate that P-body contents do not simply reflect the gene expression profiles of a given cell type but are instead enriched for transcripts characteristic of the preceding developmental stage.

RNA sequestration is conserved across vertebrates

Next, we investigated whether P-body contents and functions are conserved across vertebrate species. We first considered mouse naive and primed ES cells⁶⁰. Similar to their human counterparts, naive mouse ES cells resemble the in vivo pre-implantation mouse epiblast, while the primed state represents post-implantation epiblast cells⁶¹. Conversion from naive to primed mouse ES cells was efficient (Fig. 3a,b and Extended Data Fig. 4a), with both cell types harboring similar numbers of P-bodies (Fig. 3c).

P-body-seq analyses of mouse naive and primed ES cells (Fig. 3d) revealed that a significant fraction of P-body-enriched transcripts exhibited cell type-specific localization (54%) (Fig. 3e and Extended Data Fig. 4b), reinforcing our observations in human pluripotent stem cells. GO analysis revealed that transcripts enriched in P-bodies of both naive and primed ES cells were associated with cell cycle progression and RNA processing, similar to human datasets (Fig. 3e). Transcripts exclusively enriched in P-bodies of naive and primed ES cells were associated with chromatin remodeling and growth factor response, respectively (Fig. 3e).

Direct cross-species comparison revealed that ~50% of P-body-enriched transcripts were shared between mice and humans in both naive and primed cell states (Extended Data Fig. 4c,d).

To test whether this mechanism is conserved in phylogenetically distant vertebrates, we transduced chicken ES cells with a GFP-LSM14A lentivirus. Chicken ES cells are derived at the blastoderm stage and contribute to chimeric embryos⁶². In line with our findings in human and mouse cells, we detected GFP-LSM14A particles in chicken ES cells (Extended Data Fig. 4e), which we purified by FAPS. After P-body-seq, we found that mRNAs enriched in P-bodies of chicken ES cells exhibited extensive overlap with the P-body transcriptomes of mice and humans (Extended Data Fig. 4f). Comparative analysis revealed similar transcript features among all three species, particularly in the preferential sequestration of AU-rich mRNA in P-bodies, regardless of transcript length (Extended Data Fig. 4g). Thus, RNA sequestration in P-bodies is conserved across vertebrates.

Further analysis using the odds ratio statistic identified a significant positive association of chicken ES cells with both mouse and human ES cells, with the strongest association observed between mouse and human samples (Extended Data Fig. 4h). GO analysis of P-body-associated transcripts shared between chicken ES cells and mouse and human naive and primed ES cells revealed enrichment in categories related to RNA processing, transcription, chromatin organization and cell cycle (Extended Data Fig. 4i,j). These data indicate that transcripts sequestered in chicken ES cell P-bodies encode important regulatory factors, consistent with our findings in human and mouse ES cells.

Disrupting P-bodies induces translation of totipotency proteins in mouse ES cells

In human naive pluripotent stem cells, we found that P-bodies sequester transcripts related to the 8C stage embryo (Fig. 2l and Extended Data Fig. 3l). In mice, the equivalent developmental stage occurs in the two-cell embryo, although rare cells in naive mouse ES cell cultures, known as 2C cells, transiently express genes characteristic of the two-cell state⁶³. To determine whether transcripts enriched in the P-bodies of mouse naive ES cells comprise a signature characteristic of totipotent cells, we performed GSEA. Consistent with our findings in human naive ES cells, we observed a significant enrichment of 2C-associated transcripts among P-body-enriched mRNAs, regardless of library preparation method (Fig. 3f and Extended Data Fig. 5a). smFISH analysis confirmed that the naive pluripotency markers Dppa2 and Zfp42 were predominantly cytoplasmic in naive ES cells but primarily P-body enriched in primed ES cells (Extended Data Fig. 5b–e). This observation further indicates that P-body-based regulation of developmental processes is conserved between species.

To test the functional consequence of sequestration of 2C transcripts, we employed CRISPR–Cas9 to KO Ddx6 and dissolve P-bodies in mouse ES cells harboring the 2C-specific reporters MERVL-tdTomato and Zscan4c-GFP^64,65 (Fig. 3g, schematic). Ddx6 depletion significantly increased signal from both reporters (Fig. 3h,i) without altering cell proliferation (Extended Data Fig. 5f), suggesting that disrupting P-bodies facilitates the otherwise rare conversion of naive ES cells to the 2C state. RNA sequencing (RNA-seq) of MERVL and Zscan4 double-positive cells confirmed increased expression of genes (Obox3, Zscan4b, Zscan4d, Zscan4e and Zfp352) and transposable elements (MERVL, MT2B1, MLT2E) characteristic of the 2C state following Ddx6 suppression (Fig. 3j,k), suggesting that P-body dissolution enhances cell fate conversion. KO of other essential P-body factors, encoded by Eif4enif1 and Lsm14a³², similarly led to P-body loss and activation of the MERVL and Zscan4 reporters (Fig. 3h,i), supporting our hypothesis that P-bodies sequester fate-instructive 2C transcripts that are transiently suppressed and enter translation upon P-body disruption.

To explore this possibility further, we assessed the integrity of P-body-associated transcripts in ES cells, including those encoding 2C regulators. Analysis of both snapTotal-seq and Smart-seq data revealed over 70% overlap in P-body-associated transcripts (Extended Data Fig. 5g) between methods. Further comparison between P-body enrichment and mRNA half-lives in ES cells based on 4-thiouridine labeling⁶⁶ demonstrated minimal correlation between mRNA stability and P-body enrichment (r = −0.069) (Extended Data Fig. 5h). We found no appreciable evidence of truncated mRNAs or preferential deadenylation⁶⁷ of P-body-associated RNAs compared to cytosolic fractions (r = −0.092) (Extended Data Fig. 5i,j), consistent with previous reports^26,29,38. These findings suggest that, in ES cells, similar to our observations in HEK293T cells, P-body-associated transcripts are not preferentially degraded compared to cytosolic transcripts. We then examined the translation rate of RNAs within P-bodies of mouse ES cells^68,69. We observed a strong, inverse correlation between P-body enrichment and translation efficiency, suggesting that RNAs localized to P-bodies in naive and primed ES cells are characterized by poor translation efficiency (Fig. 4a). These results indicate that P-body-associated RNAs in ES cells are intact and translationally suppressed.

To directly assess whether P-body-enriched mRNAs engage with translational machinery in stem cells after P-body dissolution, we generated a targeted DDX6 degron system in mouse naive ES cells (Ddx6-FKBP12^F36V) (Fig. 4b) and performed polysome profiling after acute DDX6 loss. Addition of degradation tag-13 (dTAG-13) facilitated DDX6 protein degradation and eliminated detectable P-bodies within 6 h (Fig. 4c–e and Extended Data Fig. 6a–c). Consistent with our model, we observed a significant increase in the ribosome occupancy of P-body-associated mRNAs following DDX6 suppression compared to total cytosolic RNAs (P < 2.2⁻¹⁶; Fig. 4f,g). GO analysis confirmed that transcripts exhibiting enhanced ribosome occupancy were associated with stem cell maintenance (Extended Data Fig. 6d). Given the observation that disrupting P-bodies in mouse naive ES cells facilitated conversion to the 2C state, we next asked whether 2C-related transcripts demonstrated higher ribosome occupancy after DDX6 degradation. Indeed, 2C-associated mRNAs enriched in P-bodies exhibited increased ribosome occupancy after P-body dissolution (Fig. 4h,i). These data provide further evidence that mRNAs sequestered in P-bodies engage translational machinery upon their release from P-bodies in ES cells, which in turn, alters cell fate.

To measure protein-level changes, we performed large-scale, quantitative proteomics in Ddx6 degron ES cells after dTAG-13 treatment for 1 and 3 d. P-body dissolution caused significant upregulation of proteins encoded by P-body-associated mRNAs (Fig. 4j), including regulators of 2C-like cells (Fig. 4k). Elevated levels of these proteins facilitated the transition of ES cells to a 2C-like state, as shown by the increased abundance of totipotency factors such as ZSCAN4C, ZSCAN4D, ZSCAN4F, USP17LE and EIF1 (ref. ⁶⁴) (Fig. 4l). These findings demonstrate that disrupting RNA sequestration in P-bodies promotes the translation of proteins important for cell fate transitions.

In line with previous studies, a fraction of P-body-enriched RNAs underwent degradation after acute loss of DDX6 (Extended Data Fig. 6e). These downregulated mRNAs were linked to differentiation processes (Extended Data Fig. 6f), perhaps reflecting a separate mechanism for protecting transcripts needed for differentiation until appropriate signals or conditions are achieved during development⁷⁰.

miRNAs direct cell type-specific sequestration of mRNAs into P-bodies

Our data suggest cell type-specific sequestration of mRNA into P-bodies, yet the underlying mechanism remained unclear. We first considered RNA modifications, focusing on N⁶-methyladenosine (m⁶A), the most abundant modification in eukaryotic mRNA⁷¹. We used a human ES cell model that permits doxycycline-inducible ablation of the m⁶A methyltransferase METTL3 (TET-OFF METTL3)⁷² (Extended Data Fig. 7a). Eight days of doxycycline treatment suppressed METTL3 expression (Extended Data Fig. 7b) and reduced global m⁶A levels (Extended Data Fig. 7c), while pluripotency genes remained unaffected (Extended Data Fig. 7d). IF for EDC4⁺ puncta revealed no change in P-body numbers upon METTL3 deletion (Extended Data Fig. 7e). We profiled P-body contents and found no significant differences in METTL3-KO cells compared to the control (85% overlap; Extended Data Fig. 7f; compare white and gray circles). Moreover, only a small fraction of P-body-enriched RNA was m⁶A methylated, and enrichment remained largely unchanged after METTL3 KO (Extended Data Fig. 7f). These results indicate that m⁶A methylation does not primarily direct RNA sequestration into P-bodies in human ES cells, consistent with findings on stress granules in mouse ES cells⁷³.

We next explored the potential involvement of RBPs in RNA sequestration within P-bodies by mapping the P-body enrichment of transcripts targeted by 53 RBPs using published crosslinking immunoprecipitation followed by sequencing (CLIP–seq) data⁷⁴ (Fig. 5a). Targets of RBPs that localize to RNA granules, including PUM2 and HNRNPU^19,75, were enriched in P-bodies; however, these proteins are widely expressed and unlikely to confer cell type specificity. De novo motif analysis of P-body-enriched transcripts predicted recognition by multiple RBPs that were conserved across cell types, with no evident cell type-specific patterns (Extended Data Fig. 7g,h). GO analysis of the identified RBP targets revealed enrichment for pathways consistent with known P-body functions, including 3′ untranslated region (3′-UTR) processing, ribonucleoprotein granule assembly and mRNA stabilization (Extended Data Fig. 7i). These data suggest that RBPs are unlikely to drive cell type-specific RNA sequestration.

Further inspection of the CLIP–seq data revealed enrichment for targets of the Argonaute protein family (AGO1, AGO2, AGO3 and AGO4) within P-bodies (Fig. 5a). AGO proteins are essential components of the RNA-induced silencing complex and cooperate with miRNAs to suppress mRNA translation^76,77. Previous studies have shown localization of AGO proteins in P-bodies and suggested a connection between certain miRNAs and mRNA recruitment to ribonucleoprotein granules in nonmammalian cells and transformed cell lines^23,78,79,80. Moreover, miRNAs are expressed in a cell type-specific manner and play a prominent role in cell fate transitions^81,82.

To test the role of miRNAs in directing RNA sequestration of developmental transcripts into P-bodies, we profiled small RNA populations in both P-body and cytoplasmic fractions across mouse and human pluripotent cells and neural progenitors using small noncoding RNA-seq⁸³ (Extended Data Fig. 8a). Uniform manifold approximation and projection plot analysis revealed clustering of naive and primed mouse ES cell samples according to their cellular origin (Fig. 5b). Subsets of miRNAs were selectively enriched in P-bodies of pluripotent cells as well as progenitor cells (Fig. 5c and Extended Data Fig. 8b–d), including miR-300 and miR-let-7, which are known regulators of stem cell potency^84,85. Targets of miRNAs enriched in P-bodies were also sequestered within these RNA condensates (Fig. 5d and Extended Data Fig. 8e–g) and translationally repressed (Fig. 5d). Further analysis⁸⁶ revealed that P-body-targeted transcripts were enriched for specific miRNA binding sites that were absent in cytoplasmic transcripts (Extended Data Figs. 8h,i and 9a,b).

We plotted P-body enrichment for targets of miR-let-7, miR-300 and miR-381 across cell types. In both naive and primed ES cells, we observed increased enrichment of their targets in P-bodies (Fig. 5e,f). By contrast, we found no enrichment for target transcripts in mouse embryonic fibroblast (MEF) P-bodies (Fig. 5e,f). Together, our data suggest that miRNA-targeted genes are sequestered into P-bodies in a cell type-specific manner.

Disrupting miRNA function prevents RNA sequestration in P-bodies

Previous studies demonstrated the importance of AGO2 in the sequestration of let-7 targets into P-bodies in transformed cells⁷⁸. To investigate the involvement of AGO2 in global RNA sequestration, we generated Ago2-KO mouse ES cells⁸⁷ (Extended Data Fig. 9c). Ago2-KO cells had similar numbers of P-bodies as wild-type (WT) cells (Extended Data Fig. 9d), although P-bodies appeared modestly smaller (Extended Data Fig. 9d). We then sorted P-bodies from Ago2-KO ES cells and compared their transcriptome to WT counterparts. We found that Ago2 KO significantly reduced miRNA-targeted mRNA levels in P-bodies (Fig. 5g).

Next, we investigated whether modulation of polyA site usage (that is, alternative polyadenylation (APA)) influenced RNA sequestration into P-bodies. APA generates RNA isoforms that commonly differ in the length of their 3′ UTR⁸⁸. Because miRNA target sequences are frequently found in the 3′ UTR, we reasoned that forcing proximal polyadenylation (that is, shorter 3′ UTRs) would relieve miRNA-based targeting of transcripts into P-bodies. To test this, we suppressed expression of Nudt21, encoding a component of the CFIm complex that facilitates distal polyadenylation^5,89, in MEFs during reprogramming to pluripotency (Fig. 5h). We chose this context because a matched dataset of APA changes is available⁵, permitting direct comparison between polyA site usage and P-body enrichment. Nudt21 knockdown (KD) led to increased reprogramming efficiency (Extended Data Fig. 9e) and expression of pluripotency-related genes⁵ (Extended Data Fig. 9f). Similar to Ago2 KO, P-body numbers were largely unaffected by Nudt21 suppression (Extended Data Fig. 9g). P-body-seq from uninduced MEFs as well as reprogramming intermediates with and without Nudt21 silencing revealed that, in MEFs and day 3 control reprogramming cells, 3′ UTRs of P-body-targeted RNAs were longer than cytoplasmic counterparts, suggesting that regulatory sequences within 3′ UTRs mediate P-body enrichment (Fig. 5i). After Nudt21 suppression, we observed a global reduction in 3′ UTR length and the 3′ UTRs of transcripts enriched in P-bodies had similar lengths relative to those of cytoplasmic transcripts (Fig. 5i). We next used published data to compare changes in polyadenylation site usage to changes in transcript localization between P-bodies and the cytoplasm after Nudt21 KD⁵. Notably, transcripts enriched in P-bodies of control cells but absent from P-bodies after Nudt21 KD (that is, genes that relocated away from P-bodies) had shorter 3′ UTRs after Nudt21 suppression (Fig. 5j). Analysis of previously published polyA site usage data⁵ revealed that these transcripts contained significantly fewer miRNA target sites after Nudt21 KD (Extended Data Fig. 9h). These results suggest that the loss of miRNA binding sites prevents mRNA localization to P-bodies.

Disrupting miRNAs relocalizes target transcripts and influences cell identity

We next examined whether manipulating specific miRNAs could impact RNA localization and cell identity by depleting miR-300 in mouse ES cells using an antisense inhibitor (Fig. 6a). Following 48 h of anti-miR-300 treatment, we observed a significant reduction of miR-300 targets in P-bodies (Fig. 6b).

To test whether releasing miR-300 targets from P-bodies was sufficient to induce a change in cell fate (Fig. 6a), we examined the expression of 2C-related genes. We observed upregulation of 2C-related transcripts, including Zscan4 and MERVL, upon miR-300 suppression, suggesting that miR-300 restricts reversion of pluripotent cells to an earlier developmental fate (Fig. 6c). Notably, these transcripts are not direct targets of miR-300, suggesting that their upregulation results from a change in cell fate, rather than regulation from the miRNA itself. These findings indicate that RNA sequestration in P-bodies can be manipulated to change cell identity.

We next tested whether mRNAs could be directed into P-bodies through the addition of miRNA target sequences. We generated reporter constructs based on the pluripotency master regulator encoded by Nanog, which included a 3′ UTR with either six WT or six mutant let-7 sequences (let-7^WT and let-7^Mut)⁹⁰ and 24 bacteriophage MS2 stem loops (Extended Data Fig. 9i). These stem loops recruit a tagged version of the MS2 coat protein (JF₅₄₆ HaloTag-NLS-MCP) to permit direct, subcellular visualization of Nanog mRNA. We note that the nuclear localization signal on MCP effectively clears free MCP from the cytoplasm, strongly increasing the signal-to-noise ratio for the assay⁹¹. After transfection into Nanog-KO ES cells, Nanog-let-7^WT-MS2 transcripts colocalized with the P-body marker EDC4 (Fig. 6d and Extended Data Fig. 9j), suggesting that let-7 target sequences were sufficient to direct Nanog mRNA into P-bodies. By contrast, Nanog-let-7^Mut-MS2 transcripts failed to accumulate in P-bodies (Fig. 6d and Extended Data Fig. 9j). Both constructs produced similar steady-state RNA levels (Fig. 6e), excluding the possibility that the observed localization patterns arose from differences in RNA abundance. NANOG protein levels, however, were lower in cells expressing Nanog-let-7^WT-MS2, supporting a nondegradative, posttranscriptional regulatory mechanism (Fig. 6f).

Finally, we investigated whether miRNA-dependent recruitment of RNAs into P-bodies could be leveraged to alter cell identity. Under naive conditions, Nanog-KO cells self-renew; however, when switched to serum–LIF ES culture medium, NANOG helps to sustain self-renewal. We therefore transitioned Nanog-KO cells expressing either Nanog-let-7^WT-MS2 or Nanog-let-7^Mut-MS2 from naive to standard culture conditions. Cells expressing Nanog-MS2-let-7^WT had substantially lower self-renewal capacity than Nanog-let-7^Mut-MS2-expressing cells, as evidenced by increased colony number and size in clonal assays (Fig. 6g–i). These data highlight the crucial role of miRNA-mediated RNA sequestration in regulating cell potency.

Manipulating P-body-mediated RNA sequestration to direct cell fate

We next sought to manipulate P-bodies to facilitate otherwise inefficient cell fate conversions from human pluripotent stem cells to rare, clinically relevant cell types. We first focused on human totipotent-like cells, which have been identified in vitro at a frequency of 0.1% in naive pluripotent cell culture⁵⁸. P-bodies in naive human pluripotent stem cells sequester transcripts associated with the totipotent 8C state (Figs. 2l and 6j), likely preventing their translation and restricting naive ES cell plasticity. To test whether P-body dissolution could overcome this restriction, we generated human naive ES cells carrying a totipotency-specific reporter, TPRX1-GFP⁵⁹ (Fig. 6k). DDX6 suppression increased the proportion of TPRX1-GFP⁺ cells by 100-fold (~0.1% to ~10%) over previous reports⁵⁸ (Fig. 6k,l). As further validation, we used CRISPRi to deplete DDX6 (ref. ²⁸), which activated H3Y1, a DUX4 target gene expressed in the 8C embryo⁹²(Fig. 6m). Proliferation rates were unchanged between DDX6-KD and control samples (Extended Data Fig. 9k), suggesting that disrupting P-bodies increased the rate of cell fate conversion rather than cell expansion. Sorted TPRX1⁺ cells displayed increased expression of genes (for example, DUXA, ZSCAN5B, ZSCAN4) and transposable elements (for example, LTR12, MLT2A1) (Fig. 6n,o) associated with totipotency and the 8C state, while downregulating primed pluripotency markers such as POU5F1 and MER51B⁹³, consistent with enhanced cell fate conversion. These findings demonstrate that P-bodies sequester totipotency-associated transcripts to suppress their expression and suggest that manipulating P-bodies could enable the induction and expansion of rare human cell types in vitro.

We next explored whether modulating P-bodies could enhance the conversion from human ES cells to PGCLCs, which are difficult to derive in vitro, share transcriptional features with totipotent stem cells^94,95 and are highly relevant for reproductive medicine. Our data indicated that P-bodies in primed human ES cells sequester transcripts associated with a primordial germ cell signature⁹⁶ (Fig. 6p), including FXR1, ING2, EIF4G3 and MEIOC (Extended Data Fig. 9l). We introduced shRNAs targeting DDX6 into primed human pluripotent cells carrying primordial germ cell (PGC)-specific reporters, TFAP2C-GFP and PRDM1 (BLIMP1)-Tomato⁹⁶ (Fig. 6q). We induced germ cell fate and achieved a maximum conversion efficiency of less than 8% in standard conditions; however, DDX6 suppression consistently increased the formation of TFAP2C-and-BLIMP1 double-positive PGCLCs to greater than 24% (Fig. 6r). Furthermore, DDX6-depleted TFAP2C⁺BLIMP1⁺ PGCLCs exhibited elevated expression of PGC markers, including TFAP2C, PRDM1 and DPPA3 (Fig. 6s), without significant changes in cell proliferation (Extended Data Fig. 9m). These findings demonstrate that P-body dissolution facilitates the efficient programming of primed human ES cells toward the germ cell lineage. Our work further establishes a framework for using condensate biology to expand clinically relevant stem cell populations, highlighting P-body modulation as a strategy for cell fate engineering.