A DEAD-box helicase drives the partitioning of a pro-differentiation NAB protein into nuclear foci

How cells regulate gene expression in a precise spatiotemporal manner during organismal development is a fundamental question in biology. Although the role of transcriptional condensates in gene regulation has been established, little is known about the function and regulation of these molecular assemblies in the context of animal development and physiology. Here we show that the evolutionarily conserved DEAD-box helicase DDX-23 controls cell fate in Caenorhabditis elegans by binding to and facilitating the condensation of MAB-10, the C. elegans homolog of mammalian NGFI-A-binding (NAB) protein. MAB-10 is a transcriptional cofactor that functions with the early growth response (EGR) protein LIN-29 to regulate the transcription of genes required for exiting the cell cycle, terminal differentiation, and the larval-to-adult transition. We suggest that DEAD-box helicase proteins function more generally during animal development to control the condensation of NAB proteins important in cell identity and that this mechanism is evolutionarily conserved. In mammals, such a mechanism might underlie terminal cell differentiation and when dysregulated might promote cancerous growth.

During animal development, gene regulation must be temporally precise for proper cell-fate decisions to occur.The evolutionarily conserved heterochronic gene pathway regulates multiple aspects of developmental timing and governs multiple cell-fate decisions.Heterochronic genes were first discovered in C. elegans from genetic screens for mutants with temporal transformations in the cell fates of a stem cell-like population of hypodermal seam cells 1 .The C. elegans heterochronic genes execute a temporally regulated larval developmental program that drives a precise succession of events, culminating in the terminal differentiation of the seam cells and the onset of adulthood [2][3][4][5] .The mammalian homologs of these genes play critical roles in various aspects of physiology and development, including metabolism, stem-cell regulation, and fertility [6][7][8][9][10] .The dysregulation of heterochronic genes has been implicated in multiple cancers [6][7][8][9][10][11][12][13] , including chronic myeloid leukemia (CML) 12 and hepatocellular carcinoma (HCC) 14 .
During C. elegans development, the heterochronic gene lin-28 is expressed during the early larval stages and its protein product, LIN-28, inhibits the processing of the let-7 microRNA (Fig. 1a) 1,4,13,15 .LIN-28 levels drop, and in the fourth and final larval stage, let-7 microRNA levels are upregulated, silencing LIN-41 and thereby relieving lin-29a repression by LIN-41 [16][17][18] .LIN-28 also acts independently of let-7 to downregulate lin-29b levels via the hbl-1 gene during the first three larval stages 18 .During the fourth larval stage, these coordinated LIN-28 activities no longer inhibit LIN-29, which then drives exit from the (stem) cell cycle, terminal differentiation and the larval-to-adult transition 16,19 .LIN-29 is an early growth response (EGR) protein that acts in part with the NGFI-A-binding protein (NAB) transcriptional co-factor MAB-10 both to promote the expression of genes that drive terminal hypodermal seam cell differentiation and repress the expression of larval-specific hypodermal genes and to drive the onset of adulthood 20 .This process is evolutionarily conserved.For example, in mammals EGR proteins interact with NAB proteins to regulate terminal differentiation in multiple lineages 21,22 .EGR proteins have also been implicated in the regulation of luteinizing hormone LHβ expression in the pituitary gland, a key hormone involved in the onset of puberty 8,23 .Furthermore, EGR1, the mammalian homolog of LIN-29, is a barrier to the reprogramming of somatic cells into human induced pluripotent stem cells (iPSCs) 7 , highlighting the key involvement of this gene in controlling stem-cell fates.Despite the importance of the heterochronic pathway, mechanisms that regulate the transcriptional activities of the terminal effectors LIN-29 (EGR) and MAB-10 (NAB) remain largely unknown.
To better understand mechanisms that control the activity of the C. elegans developmental timing NAB protein MAB-10, we analyzed MAB-10 genetically and biochemically.Here we report that the formation of MAB-10 nuclear foci is facilitated by the DEAD-box helicase protein DDX-23, which binds to the MAB-10 protein and functions as a regulator of stem-cell fate.We propose that MAB-10 protein condensates control the transcriptional landscape in terminally differentiated seam cells.We further propose that the interaction between DDX-23 and MAB-10 is evolutionarily conserved, that a similar mechanism regulates mammalian stem-cell development and the onset of puberty, and that misregulation of this process can promote cancer progression.

MAB-10 protein forms nuclear foci in vivo
To better define the role of MAB-10 during C. elegans development, we characterized the MAB-10 expression pattern and MAB-10 protein dynamics.To examine MAB-10 expression and localization under native regulatory control, we endogenously tagged the C-terminus of the mab-10 genomic locus with GFP using the CRISPR-Cas9 genome editing technology.The MAB-10::GFP animals did not display any of the phenotypic abnormalities characteristic of mab-10(lf) mutants, indicating that this GFP-tagged mab-10 locus maintains mab-10 function.MAB-10::GFP expression turned on in the nuclei of two classes of epithelial-like cells-hypodermal cells and seam cells-in the early L4stage.As the animals completed the larval-to-adult transition and became young adults, MAB-10::GFP expression increased and became localized throughout nuclei as well as in nuclear foci.In 1-day-old adults, most of the MAB-10::GFP signal was localized to nuclear foci (Fig. 1b).The tissue expression patterns as well as the subcellular localization of MAB-10::GFP were in agreement with previous work from our laboratory studying a transgenic strain expressing MAB-10::GFP in a mab-10(lf) background 20 .
Protein self-interaction is a common feature of proteins found in nuclear bodies 24,25 .Using a yeast two-hybrid spot assay, we showed that MAB-10 proteins can self-interact (Fig. 1c), suggesting that MAB-10 self-association might facilitate MAB-10 localization to nuclear foci.To examine the dynamic properties of the MAB-10 nuclear foci, we performed fluorescence recovery after photobleaching (FRAP) studies and time-lapse imaging of MAB-10::GFP foci in vivo using 1-day-old adults (Fig. 1d).We found that MAB-10::GFP foci recovered fluorescence more slowly after photobleaching than did diffuse MAB-10::GFP in the hypodermal nuclei in which the fluorescence was distributed throughout the nucleus (Fig. 1d, e).The half-maximal fluorescence recovery time, a measure of the MAB-10 protein exchange rate between the photobleached spot and the surrounding environment, was substantially longer for foci (t 1/2 of ~18.6 s) than for the diffuse MAB-10::GFP signal in the hypodermal nuclei (t 1/2 of ~1.2 s) (Fig. 1f).The reduced mobility of MAB-10 proteins in nuclear foci as revealed by FRAP data suggests that diffusion barriers affect kinetic properties of the molecules within these structures.

DDX-23 is a regulator of stem-cell fate in C. elegans hypodermal seam cells
While lin-29 mutants fail at all of the events that characterize the larvalto-adult transition, mab-10 mutants fail at only a subset of these events 20 (Supplementary Fig. 1a), suggesting that there are factors that act with or in parallel to MAB-10 in controlling terminal seam cell differentiation and the onset of adulthood.To identify such factors, we performed a mutagenesis screen using a reporter for the adult-specific col-19 collagen gene, a downstream target gene of the heterochronic pathway that is highly expressed in the adult hypodermis 3,26 .In wildtype adults the col-19::GFP fluorescent reporter maIs105 is upregulated, whereas in lin-29 mutant adults this reporter is not expressed.In mab-10 mutant adults col-19::GFP expression is reduced relative to that in wild-type animals (Supplementary Fig. 1b).We screened mab-10(lf) mutants for second-site mutations that further reduced their low col-19 expression level (Fig. 2a).From this screen, we isolated two causal mutations (n5703 and n5705) that proved to be distinct alleles of the gene ddx-23 (Fig. 2b and Supplementary Fig. 1c).Each of these ddx-23 mutations alone caused reduction of col-19::GFP expression in adult hypodermal cells in a wild-type background (Fig. 2c and Supplementary Fig. 1d).Expressing a wild-type copy of the ddx-23 gene (using ddx-23(+) transgenes) in worms with ddx-23 mutations restored col-19 expression (Fig. 2b, c).We note that the strain with the mutation ddx-23(n5705) carried two different DDX-23 missense mutations, S365N and I383F.We recreated these two mutations individually using CRISPR-Cas9, examined the phenotypes of the two single-mutant strains (Supplementary Fig. 2a-d), and concluded that the DDX-23[I383F] mutation was responsible for the visible phenotypic abnormalities of the original screen isolate (Fig. 2c, g and Supplementary Fig. 2b-d).For simplicity, we refer to this mutation as n5705.These observations established that ddx-23 loss-of-function reduces col-19 expression levels, and that ddx-23, like lin-29, is required for col-19 upregulation in the adult.
The ddx-23 gene encodes a member of the evolutionarily conserved DEAD-box helicase protein family (Supplementary Fig. 1e).DEAD-box proteins are generally involved in RNA biogenesis and in humans the DEAD-box protein is a component of the spliceosomal U5 snRNP complex.In C. elegans, DDX-23 has been shown to regulate tissue differentiation 27 and support microRNA biogenesis 28 .Null mutations of ddx-23 cause embryonic lethality 28 .The two mutations we isolated with phenotypic consequences are missense mutations (G293E and I383F) of amino acid residues that are highly conserved among DDX-23 homologs (Fig. 2d, e) and likely cause a partial loss of the ddx-23 gene function.We examined DDX-23 expression by generating a transgenic strain that contains the translational reporter n6092 [tagRFP-T::ddx-23], which tags the endogenous ddx-23 gene locus with tagRFP-T.tagRFP-T::DDX-23 was expressed at all stages of development and broadly in the nuclei of multiple tissues throughout the animal, including the hypodermal cells and seam cells (Supplementary Fig. 2e).
To ascertain if human DDX23 functions similarly to C. elegans DDX-23, we asked if the expression of a transgene encoding human DDX23 (codon-optimized for C. elegans expression), ceDDX23, could rescue the low col-19 expression phenotype of ddx-23 mutants.Since Fig. 1 | MAB-10 protein forms dynamic nuclear foci.a Known gene regulatory interactions in the heterochronic pathway.LIN-29 (red) and MAB-10 (blue) act together on a subset of LIN-29 target genes to downregulate larval-specific genes and turn on adult-related genes.b Differential interference contrast (DIC) and confocal GFP fluorescence images of hypodermal cells in early L4, young adult, and 1-day-old adult wild-type animals expressing the translational reporter n5909 [mab-10::gfp], which tags the endogenous mab-10 gene locus with gfp.Images are representative of ten animals.Scale bar, 40 μm.c Yeast two-hybrid spot assay showing the self-interaction of MAB-10 protein (upper right spot) in quadruple drop-out plates (-His/-Ade/-Trp/-Leu) containing the competitive inhibitor of histidine synthesis 3-AT.d Representative confocal fluorescent images of individual frames from in vivo fluorescence recovery after photobleaching (FRAP) time-lapse studies of two MAB-10 nuclear foci ("Focus") and two diffuse MAB-10::GFP signals ("Diffuse") in hypodermal nuclei of 1-day-old wild-type adults expressing the reporter n5909 [mab-10::gfp].Images are representative of n = 13 (for "Focus") and n = 14 (for "Diffuse") independent experiments examining the MAB-10::GFP signal.Dotted red circle, nuclear circumference; white circle, focal area of laser bleaching and region of measurement; scale bars, 2 μm.e FRAP curves quantifying observations from the experiments shown in (c) (two MAB-10 nuclear foci ("Focus", red and pink) and two diffuse MAB-10::GFP signals ("Diffuse", blue and gray) in hypodermal nuclei).Fluorescence intensity measurements in the area of laser bleaching, normalized by considering the pre-bleach fluorescence intensity to be 1 and postbleach intensity to be 0, are shown.t = 0, time measuring pre-bleaching intensity.f Half-maximal recovery times from FRAP data for 13 experiments examining MAB-10::GFP in foci ("Focus") and 14 experiments examining MAB-10::GFP diffuse signal ("Diffuse").Error bars, mean value +/− SEM.****P = 1.278 × 10 −6 (two-sided t test).Source data for (e, f) are provided as a Source Data file.
the brood size of ddx-23 mutants are small and transgenic lines were difficult to establish, we scored F1 progeny from the injections as well as those transgenic line that we were able to obtain.All of the 20 F1 progeny (i.e., positive for the co-injection marker) obtained from the injection of ddx-23 mutants with ceDDX23 had full or partially rescued col-19::GFP expression, and the transgenic lines established from this experiment also partially rescued the low col-19::GFP expression in ddx-23 mutants (Fig. 2f and Supplementary Fig. 1f).This result indicates that the C. elegans ddx-23 and human DDX23 genes likely function similarly and act in similar molecular genetic pathways, highlighting the evolutionarily conserved nature of ddx-23 in stem-cell fate determination.

ddx-23
Because col-19 encodes an adult-specific collagen and low col-19 expression in ddx-23 mutants (Fig. 2c) suggests a defect in the larval-toadult transition, we asked if partial loss-of-function ddx-23 mutants are defective in other aspects of the onset of adulthood.We first examined the formation of the adult lateral alae, distinct longitudinal cuticular structures, using electron microscopy.ddx-23 mutants had partially disrupted alae (Supplementary Fig. 2f), consistent with the previous finding that the disruption of col-19 function leads to defects in adultspecific alae structure 29 .Furthermore, like mab-10 mutants, ddx-23 mutants had seam cells that underwent extra divisions, and a ddx-23 mutation enhanced the supernumerary-seam cell phenotype of mab-10 mutants (Fig. 2g).In addition, in ddx-23 mutant adults some seam cells failed to undergo the terminal fusion characteristic of the larvalto-adult transition of wild-type animals (Supplementary Fig. 2g).Like ddx-23 single mutants, mab-10 ddx-23 double mutants also showed a defect in seam cell fusion.mab-10 does not function in this terminal fusion process 20 .We conclude that ddx-23 plays multiple roles in the heterochronic pathway and the larval-to-adult transition, working with or in parallel to mab-10 to control seam cell exit from the cell cycle while also functioning independently of mab-10 in regulating seam cell fusion and adult alae formation (Fig. 2h and Supplementary Fig. 1a).

DDX-23 controls MAB-10 nuclear foci formation
To examine how DDX-23 and MAB-10 function together to control seam cell exit from the cell cycle, we first tested the localization of DDX-23 and MAB-10 in vivo by using a transgenic strain that coexpresses MAB-10::GFP and tagRFP-T::DDX-23, both of which are fluorescently tagged at their endogenous locus.We observed that both MAB-10::GFP and tagRFP-T::DDX-23 are expressed in the nuclei of hypodermal cells and seam cells and that they colocalize to nuclear foci (Fig. 3a).We confirmed that the C. elegans MAB-10 protein can physically interact with DDX-23 in yeast two-hybrid spot assays.We observed interactions between MAB-10 and DDX-23 (Fig. 3b).We further examined the interaction between DDX-23 and MAB-10 in vivo in C. elegans using a split GFP (spGFP) system 30,31 (Fig. 3c).We tagged DDX-23 with the N-terminal fragment of GFP (spGFPN) and MAB-10 with the C-terminal portion of GFP (spGFPC) and drove the expression of both by the seam cell promoter P ceh-16 .GFP fluorescent signal should be detected only when the proteins containing the two spGFP fragments physically interact with each other.We found that DDX-23 interacted with MAB-10 in vivo in living C. elegans (Fig. 3d); no GFP signal was observed when an empty spGFPC control tag was used.This spGFP experiment also allowed us to identify the subcellular locations where DDX-23 and MAB-10 proteins interact.We observed diffuse spGFP signal in the seam cell nuclei along with distinct nuclear foci with concentrated spGFP fluorescence within these nuclei (Fig. 3d, inset).We then asked if these nuclear foci at which DDX-23 and MAB-10 interact are the same sites at which MAB-10 condensates form.We generated a transgenic reporter strain that expresses an mCherry translational reporter for MAB-10 (P mab-10 ::mab-10::mCherry) as well as the spGFP constructs that we used to observe the interaction between DDX-23 and MAB-10 (DDX-23-spGFPN and MAB-10-spGFPC).We showed by co-localization of the mCherry and GFP signals that the MAB-10 nuclear foci are indeed the sites of interaction between DDX-23 and MAB-10 (Fig. 3e).We note that the addition of the mCherry translational reporter for MAB-10 (P mab-10 ::mab-10::mCherry) to the spGFP experiment reduced the brightness and distinct nature of the DDX-23-MAB-10 spGFP foci; the overexpressed MAB-10::mCherry from the extrachromosomal array might have competed with MAB-10::spGFPC for the binding to DDX-23::spGFPN.Together, these data support the conclusion that MAB-10 and DDX-23 proteins interact at nuclear foci.
Both DDX-23 and MAB-10 proteins contain large intrinsically disordered regions (IDRs) (Supplementary Fig. 3a).Interactions between IDRs can act as a driving force for protein condensation into subnuclear bodies 32 .We tested if DDX-23 and MAB-10 can form homotypic interactions in vitro and found that both purified recombinant DDX-23-mNeonGreen fusion protein (Supplementary Fig. 3b-e) and MAB-10-mScarlet I fusion protein (Supplementary Fig. 3f-h) showed homotypic interactions.DDX-23 formed homotypic interactions readily, while MAB-10 was less efficient at forming MAB-10 spots in an homotypic assay as indicated by the quantification of fluorescent spots (Supplementary Fig. 3e, h).
Since our in vivo data showed that DDX-23 and MAB-10 proteins interact at nuclear foci (Fig. 3c-e), we next asked whether the DDX-23 protein influences the incorporation of MAB-10 protein into these nuclear structures.We assessed the in vitro formation of MAB-10-mScarlet I protein condensates in the presence or absence of a DDX-23-mNeonGreen fusion protein.The MAB-10 protein condensed more efficiently (~3.5-fold,P < 0.001) in the presence of DDX-23 (Fig. 3f-i and Supplementary Fig. 3i), indicating that DDX-23 can facilitate the condensation of MAB-10 protein.As shown above, the DDX-23 missense mutation I383F causes a defect in seam cell fate (Fig. 2 and Supplementary Fig. 2).We asked if this mutation can influence MAB-10 foci formation by testing the ability of MAB-10 to be incorporated into MAB-10 structures in a heterotypic assay with DDX-23[I383F] protein.The efficiency of MAB-10 incorporation into MAB-10 structures in vitro was reduced by ~60% (P = 0.0024) in the presence of mutant DDX-23[I383F] protein as compared to wild-type protein (Fig. 3g-i).We next asked how the IDR of DDX-23 might influence MAB-10 condensate formation.Specifically, using heterotypic assays of MAB-10 and DDX-23 proteins, we observed a 76% reduction (P < 0.01) in the number of MAB-10 structures formed in assays involving DDX-23 lacking its IDR (DDX-23(ΔIDR), Fig. 3g) compared to the number formed in assays performed with wild-type DDX-23 (Fig. 3h, i).These results suggest that in addition to the DEAD-box domain the lowcomplexity IDR of DDX-23 is also involved in driving the condensation of MAB-10.
We examined the ability of DDX-23 to facilitate the formation of MAB-10 foci in vivo by comparing MAB-10::GFP foci in wild-type and ddx-23 mutants.We found that for the n5703 and n5705 alleles of ddx-23, the number of animals with MAB-10::GFP foci in adult hypodermal nuclei was reduced by 60% (P = 0.0002) and 76% (P < 0.0001), respectively (n = 30 for each genotype) (Fig. 3j, k LIN-29 and MAB-10 colocalize in repressive condensates and silence larval genes, including hedgehog-related genes Since MAB-10 is a co-factor for the transcription factor LIN-29 and coregulates genes involved in terminal differentiation, we examined the relationship between LIN-29 expression and the timing of the appearance of MAB-10 foci.We generated and examined a transgenic strain that contains the translational reporter n5908 [lin-29::gfp], which tags the endogenous lin-29 locus with GFP, and an mCherry translational reporter for MAB-10 (P mab-10 ::mab-10::mCherry).We found that LIN-29 and MAB-10 proteins colocalized at nuclear foci in the hypodermal seam cells of adult animals (Fig. 4a).We next asked whether LIN-29 foci formation is dependent on MAB-10 or DDX-23 proteins.We examined the expression of LIN-29::GFP in mab-10, ddx-23, or mab-10; ddx-23 double mutants and showed that LIN-29 foci were formed in the adult hypoderm in a MAB-10-and DDX-23-independent manner (Fig. 4b).
We next examined the roles of LIN-29, MAB-10 and DDX-23 in regulating gene expression.Since MAB-10 functions as a transcriptional co-factor for LIN-29 to regulate a subset of LIN-29 target genes and our data showed that adult-specific MAB-10 nuclear foci formation was in part facilitated by DDX-23, we performed RNA sequencing (RNA-Seq) studies of lin-29, mab-10, and ddx-23 mutant adults and compared their gene expression profiles to that of wild-type adults.We found that the majority of genes that were misregulated in all three mutants were upregulated (Fig. 4c), showing a major role for LIN-29, MAB-10, and DDX-23 proteins in transcriptional repression.We observed a significant overlap among the genes that are upregulated in lin-29, mab-10, and ddx-23 mutant adult animals (Supplementary Fig. 4a).This overlap was absent among the genes that were downregulated in lin-29, mab-10, and ddx-23 mutant adult animals.These findings suggest that LIN-29, MAB-10, and DDX-23 repress a common set of target genes.We performed protein-domain enrichment analysis (Fig. 4d) and gene-enrichment analysis (Supplementary Fig. 4c) to identify categories of genes co-regulated by these three proteins.We observed that genes involved in structural constituents of the cuticle and the Patched (hedgehog-like) family of genes are enriched among the shared transcriptionally repressed targets of LIN-29, MAB-10 and DDX-23 (Fig. 4d).For example, there are many larval-specific collagen genes normally expressed highly in L4-stage larvae and downregulated in adults; these genes were upregulated in lin-29, mab-10, and ddx-23 mutant adults compared to wild-type adults (Supplementary Fig. 4d).This observation indicates that the LIN-29, MAB-10, and DDX-23 proteins are critical in establishing the normal adult gene-expression profile and that they act by transcriptionally repressing larvalrelated genes.
We also found several hedgehog-related genes and genes that encode putative hedgehog receptors to be upregulated in lin-29, mab-10 and ddx-23 mutants (Fig. 4d and Supplementary Fig. 4e).These genes are the evolutionary homologs of Hedgehog (Hh) and its receptor Patched (Ptc), which control various developmental processes in species ranging from Drosophila to humans 33 and have been implicated in the regulation of mammalian stem cell maintenance and proliferation 34,35 .Hedgehog-and Patched-related genes are known to be involved in C. elegans molting 36 but have not been previously linked to the heterochronic pathway, which includes lin-29, mab-10 and now ddx-23.
Our data suggest that LIN-29 and MAB-10 act in the same transcriptional condensate to co-repress a subset of larval-specific genes.While DDX-23 functions to concentrate MAB-10 into these foci, partitioning of LIN-29 into these foci is not dependent on MAB-10 or DDX-23.We suggest that DDX-23 binds to and facilitates the partitioning of MAB-10 protein into LIN-29 transcriptional condensates in adult hypodermal nuclei, thereby enabling MAB-10 to function as a transcriptional co-factor for LIN-29 and co-repress their target genes, which include larval genes and hedgehog-related genes (Fig. 4e).

Discussion
Recent studies of transcriptional condensates have revealed new possible mechanisms of transcriptional control.There is growing evidence that transcriptional condensates are involved in both the activation [37][38][39][40][41][42][43][44] and the repression [45][46][47][48] of gene expression.However, little is known about how the formation of such condensates is controlled.In this study, we report that in vivo the C. elegans DEAD-box helicase protein DDX-23 regulates stem-cell fate at least in part by binding to and enhancing the partitioning of MAB-10 proteins into condensates, which in turn repress LIN-29/MAB-10 co-regulated target genes-including known larval-specific genes and Hedgehog-related genes-and thereby drive terminal cell differentiation and the onset of adulthood (Fig. 4e).Partial loss-of-function mutations in ddx-23 reduce MAB-10 partitioning into condensates and upregulate larval-specific genes and Hedgehog-related genes, resulting in a failure of the stem cell-like seam cells to terminally differentiate and transition to adulthood (Fig. 4e).Previous studies of C. elegans, mice and human cell lines have shown that suppression of LIN-28 levels promotes the expression of LIN-29 (EGR) and MAB-10 (NAB) proteins, both let-7 microRNAdependently via LIN-41 (mammalian homolog TRIM71) [16][17][18] and let-7-independently via HBL-1 (mammalian homolog IKAROS) 18,49-51 and GFI1 52 (Fig. 4f).The ddx-23 gene was previously found to play a role in the primary microRNA processing of let-7 microRNA 28 .Here we describe a gene regulatory mechanism that regulates the activities of MAB-10 (NAB) and LIN-29 (EGR): DDX-23 drives the condensation of MAB-10 (NAB) protein into nuclear focal structures to spatiotemporally organize the nucleus and transcriptionally repress LIN-29 (EGR) target larval genes, including hedgehog-related genes.We propose that this process is evolutionarily conserved and that a comparable molecular genetic mechanism-in which the partitioning of NAB proteins by a DEAD-box helicase drives repression of proliferation-related and Hedgehog-related genes by EGR-controls mammalian stem cell development and the onset of puberty (Fig. 4f).In mammals, the DEAD-box helicase DDX5 protein 53 and EGR1 protein 7 act as a barrier to somatic cell reprogramming, and the DEAD-box protein Ddx20/DP103 represses transcriptional activation by Egr2 54 .We propose that DEADbox proteins function to control EGR transcriptional activity by facilitating the condensation of its co-factor NAB, thereby coordinating molecular genetic programs that control cell differentiation.
The highly conserved LIN28/let-7 axis has previously been found to regulate Hedgehog signaling in multiple contexts, e.g., during retina regeneration in zebrafish 55 and in a mouse model for embryonal tumors with multilayered rosettes (ETMRs), a tumor-type characterized by LIN-28A overexpression 56 .LIN28 is implicated in regulating the onset of puberty in mice and humans 9,[57][58][59][60][61] , while Hedgehog signaling controls cell proliferation and cell-type determination in the pituitary gland of mice by regulating the luteinizing hormone LHβ expression, a key hormone involved in the onset of puberty [62][63][64][65] .Hence in mammals, multiple components of the evolutionarily conserved heterochronic pathway have been implicated in the onset of puberty, while in C. elegans this pathway similarly regulates the larval-to-adult transition.Our findings suggest a missing link between the LIN28/let-7 axis and Hedgehog signaling in regulating this process: LIN28 signaling promotes the expression of EGR and NAB proteins, which in turn leads to the repression of Hedgehog genes.We propose that this repression is achieved by the condensation of NAB proteins to form a repressive condensate, a process driven in part by a DEAD-box helicase protein and that the repression of Hedgehog-related genes subsequently induces the expression of the luteinizing hormone LHβ and leads to cell differentiation and the onset of puberty (Fig. 4f).
Our findings also have implications for the understanding of molecular mechanisms of carcinogenesis.The overexpression of human LIN28 inhibits let-7 microRNA gene activity and thereby promotes cancerous growth 12,66 , possibly by augmenting downstream Shh Hedgehog signaling 56 .We speculate that this regulation is driven by the downregulation of EGR, which acts as a tumor suppressor and is commonly mutated or suppressed in multiple human cancers 67,68 and that this downregulation leads to the upregulation of EGR target genes (including Hedgehog genes) that are normally silenced to regulate cell differentiation and the onset of adulthood.In healthy cells, DEAD-box helicase proteins ensure repression of EGR target genes essential to cell proliferation by facilitating the formation of MAB-10 (NAB) repressive transcriptional condensates; failure to form such condensates could lead to abnormal cell proliferation.Indeed, DDX23 mutations and deletions have been detected in several human cancers 69 , suggesting a tumor suppressor role for this gene.Based on our findings, we propose as a mechanism that these DDX23 genetic defects result in a failure in the formation of DDX23-NAB transcriptional condensates and in the consequent upregulation of EGRrepressed targets-which include genes involved in Hedgehog signaling-leading to the abnormal cell proliferation that drives cancerous growth.

C. elegans strains and transgenes
All C. elegans strains were cultured as described previously 70 .We used the N2 Bristol strain as the reference wild-type strain, and the polymorphic Hawaiian strain MT22656 for genetic mapping and SNP analysis 71 .All strains were maintained at 20 °C.Details about the strains used in this study can be found in Supplementary Table 1.

Molecular cloning and transgenic strain construction
To endogenously tag the mab-10 genomic locus at its C-terminus with GFP (transgene n5909[mab-10::gfp]) using CRISPR-Cas9 72 , we generated a mab-10::gfp homologous repair template for Cas9mediated gfp knock-in.We PCR-amplified a ~3 kb genomic region centered on the mab-10 stop codon from N2 genomic DNA and cloned the resulting fragment into the pCR-Blunt vector using the ZeroBlunt TOPO Cloning Kit (Thermo Fisher).Using the Gibson assembly kit (New England Biolabs), we inserted a GA-linker and gfp in-frame between the last amino acid codon of the mab-10 gene and the stop codon, resulting in a homologous repair template with ~1.5 kb homology arms.To avoid cleavage of the repair template by Cas9, we mutated the PAM motif of the Cas9 target site in the repair template.We also generated a Cas9-sgRNA plasmid to cleave upstream of the mab-10 stop codon by inserting a targeting sequence (5'-GAAAAAGTAGCAATACAACA-3') into the pDD162 Cas9-sgRNA plasmid 72 by site-directed mutagenesis.Forward primer 5'-GAAAAAGTAGCAATACAACAGTTTTAGAGCTAGAAATAGCAAGT-3' and reverse primer 5'-CAAGACATCTCGCAATAGG-3' were used.We co-injected the homologous repair template, the mab-10 Cas9-sgRNA plasmid, and an unc-22 sgRNA co-CRISPR plasmid into wildtype N2 worms.Twitching progeny (indicative of successful unc-22 co-CRISPR integration) were selected and genotyped for integration of gfp into the mab-10 locus.Finally, the strain was backcrossed to wild-type N2 animals to remove the unc-22 mutation.
To generate the ddx-23 rescuing transgene nEx2848, the region spanning 5 kb upstream of the ddx-23 gene (P ddx-23 ), ddx-23 genomic DNA, and the 3'UTR were PCR-amplified using forward primer 5'-GG TACAACCATGTTCAAATCAGCATCC-3' and reverse primer 5'-CG TGAGTTTGGTTCTGGAGCTT-3'.This PCR fragment was cloned into a pUC19 vector and a Tag-RFP-T sequence was inserted immediately before the ddx-23 stop codon to generate the ddx-23 rescuing transgene nEx2958.P ddx-23 ::ceDDX23::ddx-23 3'UTR (nEx2971 and nEx2972) was generated by HiFi DNA Assembly (New England Biolabs) of P ddx-23 , wild-type human DDX23 cDNA codon-optimized for expression in C. elegans and containing 3 synthetic introns generated using gBlocks Gene Fragments (Integrated DNA Technologies), and the ddx-23 3'UTR in a pUC19 vector.For the split-GFP protein-protein in vivo interaction assays, all cloning was performed using the HiFi DNA Assembly system.The N-terminal GFP (spGFPN) plasmid was cloned by inserting the ddx-23 cDNA and spGFPN fragment between a 2.5 kb ceh-16 promoter and the unc-54 3'UTR.The C-terminal GFP (spGFPC) plasmid was cloned by inserting the mab-10 cDNA (or an ATG start codon sequence for the "empty" control) and the spGFPC fragment between a 2.5 kb ceh-16 promoter and the unc-54 3'UTR.Complete plasmid sequences of all plasmids are available from the authors upon request.
Transgenic strains were generated by germline transformation as described 74 .All transgenic constructs were injected at 1-50 ng/ml.

Microscopy and image analysis
All C. elegans micrographs excluding electron micrographs are of lateral views of hermaphrodite animals.Nomarski DIC and epifluorescence micrographs were obtained using a Zeiss Axio Imager Z2 compound microscope and Zen Blue (Zeiss) software.Confocal microscopy was performed using the Zeiss LSM 800 instrument, and images were obtained using the Zen Blue 2.0 (Zeiss) software.The resulting images were prepared using ImageJ (National Institutes of Health) and Adobe Illustrator CS4 softwares.Image acquisition settings were calibrated to minimize the number of saturated pixels and were kept constant throughout each experiment.Graphs and indicated statistical analyses were performed using GraphPad Prism 6 software.For electron micrographs, animals were assayed by picking L4 males to plates, fixing 24 h later, and performing electron microscopic analysis, as previously described.

Fluorescence recovery after photobleaching (FRAP)
We immobilized 1-day-old adult animals with 10 mM levamisole (an acetylcholine agonist that induces muscle contraction) on 2% agarose pads for confocal imaging using the Zeiss LSM 800.We imaged hypodermal cells that were proximal to the cover slip.We identified two imaging regions of interest (ROIs) for each FRAP experiment (one for FRAP measurements, one for an imaging-associated bleaching control).We manually defined the FRAP ROI as the region around a given MAB-10::GFP focus.As a control, we performed FRAP studies of the diffuse MAB-10::GFP signal in hypodermal cells.To achieve complete bleaching of GFP in the FRAP ROI, bleaching was conducted with 25 rounds of 100% laser induction with a 488 nm laser.Hypodermal cells used to perform FRAP studies of diffuse MAB-10::GFP signals ("Diffuse") were then tracked for at least 35 s, while hypodermal cells used to perform FRAP studies of MAB-10::GFP foci ("Focus") were tracked for at least 80 s.Data were analyzed using ImageJ, measuring the fluorescence intensity for all frames and fitting the FRAP curve by an increasing exponential.

Mutagenesis screen for enhancers of mab-10 and mapping of ddx-23
To screen for enhancers of the mab-10 seam cell defect, we mutagenized mab-10(tm2497) mutants with ethyl methanesulfonate (EMS) as described previously 70 .The starting strain contained the col-19::gfp (maIs105) transgene, which shows a reduced expression level in mab-10(lf) adults relative to that of wild-type adults and served as a reporter for the heterochronic pathway.We used a Nikon SMZ18 fluorescence dissecting microscope to screen F2 progeny of mutagenized hermaphrodites for second-site mutations that further reduced the low hypodermal col-19 expression levels of mab-10(tm2497) mutants.Screen isolates were backcrossed to an otherwise wild-type strain containing the maIs105 reporter.ddx-23(n5703) and ddx-23(n5705) were identified by crossing the mutant isolate with a Hawaiian strain containing the maIs105 reporter, isolating the F2 progeny with low hypodermal col-19 expression and mapping the mutation using singlenucleotide polymorphisms 71 and whole-genome sequencing.Transgenic rescue experiments demonstrated that these ddx-23 mutations were the causative mutations, as described in the text.

Seam cell count
The number of seam cells was quantified in 1-day-old adults, using either the maIs105[col-19::gfp] or wIs78[ajm-1::gfp + scm::gfp] transgenes.DIC and epifluorescence optics using a Zeiss 63x objective lens on a Zeiss Axio Imager Z2 compound microscope were used.The number of seam cells was scored by counting GFP-positive seam cell nuclei along the length of each animal.
Yeast two-hybrid binding assay ddx-23 and mab-10 cDNAs were cloned into pGADT7 and pGBKT7 plasmids (Takara), and the constructs were co-transformed into yeast strain Y2HGold (Takara).Co-transformants were selected on SD plates containing minimal supplements without leucine and tryptophan (SD/-Leu/-Trp) for 2 days at 30 °C.To test for protein interaction, co-transformants were picked and grown overnight in SD/-Leu/-Trp broth.Cultures were diluted to a similar optical density at 600 nm, and 3 μl of each culture was spotted onto SD plates containing minimal supplements without tryptophan, leucine, histidine, and adenine (SD/-Leu/-Trp/-His/-Ade) and containing 10 mM of the competitive inhibitor of histidine synthesis 3-AT, and cultured for 2 days at 30 °C.

Recombinant protein expression and protein purification
HiFi DNA Assembly into a pDEST17 (Thermo Fisher) backbone was used to create all bacterial expression vectors.For the recombinant DDX-23 expression vector, the ddx-23 cDNA sequence followed by mNeonGreen and a 6xHis tag at its 3' end was inserted into pDEST17.Site-directed mutagenesis of this plasmid was performed to create the DDX-23 I383F mutant and the DDX-23(ΔIDR) mutant (removing amino acid residues 2-265).For the recombinant MAB-10 expression vector, mab-10 cDNA followed by mScarlet I and a 6xHis tag was inserted into pDEST17.Vectors expressing C-terminally 6xHis-tagged mNeonGreen and mScarlet I were created as "empty" controls.
For protein expression, plasmids were transformed into Shuffle T7 competent E. coli cells (New England Biolabs).A fresh bacterial colony was inoculated into LB medium containing ampicillin and grown overnight at 30 °C.Cells were diluted 1:100 in 1 liter of LB medium containing ampicillin and grown at 30 °C until the OD 600 reached between 0.5 and 0.6.IPTG was added to 0.5 mM, followed by incubation at 16 °C for 18 h.Cells were collected and stored frozen at −80 °C.

In vitro protein interaction assay
As described in the figure legends, recombinant proteins were added to solutions at varying concentrations with 200-220 mM final NaCl with or without 10% PEG-8000 as crowding agent in the reaction buffer (50 mM Tris-HCl pH 7.4, 10% glycerol, 1 mM DTT).The protein solution was loaded onto glass slides within a silicone isolator (Sigma) and imaged using a Zeiss LSM 800.For all experiments, at least three independent mixtures were analyzed.For better visualization, micrographs were color-inverted using ImageJ for Fig. 3f, h.For quantification of fluorescent spots, we performed Particle Analysis using ImageJ with a size threshold of 1 μm 2 .

Gene expression analyses
For RNA-Seq experiments, total RNA from three biological replicates of age-synchronized young adult hermaphrodites (200 in total, picked manually) of wild-type, lin-29(n836), mab-10(tm2497), and ddx-23(n5705) genotypes were prepared using a BeadBug microtube homogenizer (Sigma) and 0.5-mm zirconium beads (Sigma).RNA was extracted using the RNeasy Mini kit (QIAGEN) according to the manufacturer's instructions.RNA libraries were prepared for sequencing using standard Illumina protocols and the NeoPrep system.RNA-Seq was performed using the Illumina NextSeq500 platform, and data were analyzed using standard protocols 75 .Reads were mapped against Caenorhabditis elegans ws258 reference sequences using STAR/2.5.3a, and gene expression was quantified using RSEM v. 1.3.0.Differential gene expression analysis was performed using DESeq2.An adjusted P value cutoff of <0.05 and a fold change of more than 2 was used to identify differentially expressed genes between the wild-type and mutants.There were 102 genes-including larval-specific collagen genes and several hedgehog-related (grd) genes and putative hedgehog receptor (ptr) genes-that were upregulated in all three mutants (lin-29, mab-10, and ddx-23 mutants) relative to wild-type animals.We performed gene ontology enrichment analysis and protein domain enrichment analysis using DAVID functional annotation tools 76,77 .

Statistical analysis
Unpaired, two-tailed t tests were used to compare half-maximal recovery times from FRAP studies of MAB-10::GFP in foci and diffuse signals, numbers of seam cells that undergo extra seam cell division between strains, normalized gene expression abundance log 2 (fpkm+1) of multiple genes between strains, the number of homotypic droplets of control mNeonGreen alone to that of purified DDX-23-mNeon-Green, the number of homotypic droplets of control mScarlet I alone to that of purified MAB-10-mScarlet I, and the number of MAB-10 droplets in mixed heterotypic assays with control, wild-type DDX-23, mutant DDX-23[I383F] or DDX-23(ΔIDR) proteins.Two-tailed t-tests were also used to compare MAB-10 total nuclear intensity in wild-type and ddx-23(n5705) animals, and maIs105 transgene levels in wild-type, mab-10, ddx-23, mab-10; ddx-23 and ddx-23(+) rescue animals.Fisher's exact test was used to compare the number of animals that have MAB-10::GFP foci in adult hypodermal nuclei in wild-type versus ddx-23(n5703) or ddx-23(n5705) mutants.Gene ontology enrichment analysis was performed according to DAVID functional annotation tools 76,77 ; Fisher's exact test was used to determine P values, and False Discovery Rates were calculated using the Benjamini-Hochberg procedure to correct for multiple testing.Statistical tests were performed using GraphPad Prism 6 software.Biological replicates were performed using separate populations of animals.

Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Fig. 4 |
Fig. 4 | DDX-23-driven formation of MAB-10 (NAB) nuclear foci might control terminal differentiation and the onset of adulthood by transcriptionally repressing LIN-29 (EGR) target hedgehog and larval-specific genes.a Representative confocal fluorescent images of five animals expressing n5908[lin-29::gfp] (green channel), which fluorescently tags lin-29 at its endogenous locus, and nEx3004[MAB-10::mCherry] (red channel).Inset: Hypodermal seam cell nuclei.Scale bar, 40 μm.b DIC and confocal fluorescent images of hypodermal cells in 1day-old wild-type, mab-10(tm2497), ddx-23(n5705) and mab-10(tm2497); ddx-23(n5705) adults expressing the reporter n5908[lin-29::gfp].Images are representative of 5-8 animals per genotype.Scale bar, 40 μm.c RNA-Seq analysis of lin-29, mab-10, and ddx-23 mutant adults.The number of genes upregulated (black)and downregulated (gray) relative to wild-type adults are shown for the indicated genotypes.d Protein domain enrichment analysis of genes that are upregulated in lin-29, mab-10, and ddx-23 mutant adults, as identified by RNA-Seq analysis.Statistical tests were performed according to DAVID functional annotation tools76,77 , where Fisher's exact test was used to determine P values and false discovery rates were calculated by the Benjamini-Hochberg procedure to correct for multiple testing.e A proposed model for how DDX-23-driven formation of MAB-10 nuclear foci controls LIN-29/MAB-10 co-regulated target gene repression in the adult hypoderm.DDX-23 binds to and enhances the partitioning of MAB-10 proteins into repressive foci, causing silencing of LIN-29/MAB-10 co-repressed target genes-including larval-specific and Hedgehog-related genes-and thereby defining adult cell identity (left panel).Loss-of-function mutations in ddx-23 reduce MAB-10 partitioning into LIN-29 repressive foci and upregulate LIN-29/MAB-10 co-repressed genes including larval-specific and Hedgehog-related genes, resulting in a failure to maintain adult cell identity (right panel).Color key for proteins are as follows: DDX-23 (purple), MAB-10 (blue), LIN-29 (red), chromatin factor (yellow).f Models for how the heterochronic pathway regulates cell differentiation and the juvenile-toadult transition in C. elegans (upper panel) and mammals (lower panel).For example, LIN28 signaling in mammals (LIN-28 signaling in C. elegans) coordinates the expression of EGR (C.elegans LIN-29) and NAB (C.elegans MAB-10) proteins, both let-7 microRNA-dependently via TRIM71 (mammalian homolog of the C. elegans LIN-41) and let-7-independently via IKAROS (mammalian homolog of the C. elegans HBL-1) and GFI1.The formation of nuclear foci that contain NAB (MAB-10) proteins is facilitated by a DEAD-box helicase protein (DDX-23) and function to transcriptionally repress EGR (LIN-29) target genes including proliferation-related genes and hedgehog-related genes.The repression of proliferation-related genes leads to cellular differentiation, while Hedgehog signaling leads to luteinizing hormone LHβ expression and the onset of puberty.DDX-23 was previously described to play a role in the primary microRNA processing of let-7 microRNA 28 (dotted lines), possibly providing a feedback regulatory loop in this pathway.Source data for (c) is provided as a Source Data file.