Cellular quiescence uncouples the proteome from the transcriptome in neural stem cells

TL;DR

Quiescent neural stem cells retain mRNAs in the nucleus, causing a mismatch between RNA and protein levels, which helps protect the cells.

Contribution

Discovery of a conserved mechanism in quiescent neural stem cells where mRNAs are retained in the nucleus to repress translation.

Findings

Quiescence causes transcripts from over 2000 genes to accumulate in the nucleus, leading to protein downregulation.

GA-rich mRNAs relocalize to nuclear speckles enriched with SR-proteins, promoting their retention and translational repression.

The mechanism is conserved from Drosophila to mammals and distinct from TOR-dependent translational repression.

Abstract

Quiescence is a cellular state defined by reversible cell-cycle arrest and diminished biosynthesis, particularly of nucleic acids and proteins. These features protect stem cells from proliferation-induced mutations, self-renewal exhaustion, and environmental insults. Despite relevance to development, tissue homeostasis and cancer, we lack understanding about many aspects of quiescence regulation and unique molecular markers for this state. Here, we employ Drosophila and mammalian neural stem cells to reveal that a mechanism for inhibiting translation in quiescence is selective nuclear enrichment of transcripts from more than 2000 genes, resulting in uncoupling between transcriptome and proteome. Three-quarters of these transcripts become increasingly nuclear as quiescence deepens, and nuclear bias predicts protein downregulation for the large majority of targets. We find that a large fraction of nuclear-biased transcripts present GA-rich multivalency and relocalise to nuclear speckles with increased SR-protein enrichment, which we propose promotes their nuclear retention. Finally, our evidence for differing degrees of transcript processing in steady-state quiescence suggests regulated sequential deployment of factors towards cell-cycle reentry. In brief, we present a previously unappreciated layer of post-transcriptional control of quiescence.

Quiescent neural stem cells inhibit protein synthesis by selectively retaining many mRNAs in the nucleus. GA-rich multivalent transcripts accumulate in SR-protein–enriched nuclear speckles as quiescence deepens, reducing their cytoplasmic translation and producing a marked uncoupling of transcriptome and proteome.

Quiescence induces significant nucleocytoplasmic bias changes in transcripts from >2,000 genes, with ~75% becoming increasingly nuclear as NSCs enter deeper quiescence.Nuclear or cytoplasmic mRNA bias predicts protein down- or up-regulation with >80% concordance, explaining transcriptome–proteome uncoupling in qNSCs.Nuclear-biased mRNAs are specifically enriched for GA-rich multivalency and relocalise to SSRM2-positive nuclear speckles.Speckles in qNSCs show increased SR-protein signal, and GA-multivalent transcripts (e.g., EIF3A, MAP1B) preferentially accumulate there.The pathway is conserved from Drosophila to mammals and is distinct from TOR-dependent translational repression.

Quiescence induces significant nucleocytoplasmic bias changes in transcripts from >2,000 genes, with ~75% becoming increasingly nuclear as NSCs enter deeper quiescence.

Nuclear or cytoplasmic mRNA bias predicts protein down- or up-regulation with >80% concordance, explaining transcriptome–proteome uncoupling in qNSCs.

Nuclear-biased mRNAs are specifically enriched for GA-rich multivalency and relocalise to SSRM2-positive nuclear speckles.

Speckles in qNSCs show increased SR-protein signal, and GA-multivalent transcripts (e.g., EIF3A, MAP1B) preferentially accumulate there.

The pathway is conserved from Drosophila to mammals and is distinct from TOR-dependent translational repression.

GA-rich multivalent mRNAs accumulate in nuclear speckles to repress translation during NSC quiescence, a pathway conserved from Drosophila to mammals.

Figures

Downregulation of novel *Drosophila* protein Snx induces anachronical qNSCs.(A) In wild-types (WT), qNSCs are present in “early” (newly hatched) but not “late” (wandering third-instar stage) larvae, display a cytoplasmic fibre (arrowheads) and downregulate Mira and Dpn. *NSC* > *GFP* was *grh-GAL4, UAS-mCD8::GFP*. (B) Late larval NSCs of the 2V327 mutant or *CG14712*^*RNAi*^ (GL00466) have a morphology reminiscent of qNSCs and lack PH3. MARCM clones (see Methods) were by labelled by NLS::GFP and *NSC* > *CG14712*^*RNAi*^ was *nab-GAL4, UAS-CG14712*^*GL00466*^. All scale bars: 10 µm. (C) Sequen

Proteome changes as NSCs transition between active and deeper quiescence states.(A) PCA plot. (B) TMT spectrometry protein intensity changes (log_2_ difference, arbitrary units) from day 0 (aNSCs); mean and s.d. indicated for each time-point with *n* = 1 biological replicate for each of 10 time points. (C) Control proteins plotted on the background of all. (D) Most upregulated (5, green) and downregulated (10, magenta) biological process GO terms in 21d-qNSCs. (E) Change of specified categories of proteins plotted on the background of all. Source data are available online for this figure.

Downregulation of novel *Drosophila* protein Snx induces anachronical qNSCs.(A) In wild-types (WT), qNSCs are present in “early” (newly hatched) but not “late” (wandering third-instar stage) larvae, display a cytoplasmic fibre (arrowheads) and downregulate Mira and Dpn. *NSC* > *GFP* was *grh-GAL4, UAS-mCD8::GFP*. (B) Late larval NSCs of the 2V327 mutant or *CG14712*^*RNAi*^ (GL00466) have a morphology reminiscent of qNSCs and lack PH3. MARCM clones (see Methods) were by labelled by NLS::GFP and *NSC* > *CG14712*^*RNAi*^ was *nab-GAL4, UAS-CG14712*^*GL00466*^. All scale bars: 10 µm. (C) Sequen

Nuclear-biased transcripts in qNSCs have overrepresentation of GA-rich multivalent mRNAs and are enriched in nuclear speckles.(A) Average fraction of transcripts containing five common types of mRNA multivalency (which can coexist in the same transcript) plotted for nuclear biasing, cytoplasmic biasing, or mRNAs not changing subcellular bias from n = 4 biological replicates (fracRNA-seq data). (B) Co-staining of *oligo(dT)* with the nuclear speckle markers Srrm2 and phosphor-SR in aNSCs versus qNSCs of increasing quiescence depth. All scale bars: 10 µm. (C) Quantification of *oligo(dT)* and ph

Proteome changes as NSCs transition between active and deeper quiescence states.(A) PCA plot. (B) TMT spectrometry protein intensity changes (log_2_ difference, arbitrary units) from day 0 (aNSCs); mean and s.d. indicated for each time-point with *n* = 1 biological replicate for each of 10 time points. (C) Control proteins plotted on the background of all. (D) Most upregulated (5, green) and downregulated (10, magenta) biological process GO terms in 21d-qNSCs. (E) Change of specified categories of proteins plotted on the background of all. Source data are available online for this figure.

Nuclear retention of transcripts increases as NSCs transition between active and deeper quiescence states.(A) PCA plot. (B) Z-score at 0, 3 and 10 d in BMP4 for marker transcripts for each subcellular compartment. Nuclear transcripts: *nuclear paraspeckle assembly 1* (*Neat1*), *metastasis-associated lung adenocarcinoma transcript 1* (*Malat1*), *plasmacytoma variant translocation 1* (*Pvt1*). Cytoplasmic transcripts: *small subunit ribosomal protein 14* (*Rps14*), *glyceraldehyde-3-phosphate hydrogenase* (*Gapdh*), *7S RNA 1*(*Rn7s1*). (C) Z-scores at 0, 3 and 10 d in BMP4 for the transcripts

Nuclear-biased transcripts in qNSCs have overrepresentation of GA-rich multivalent mRNAs and are enriched in nuclear speckles.(A) Average fraction of transcripts containing five common types of mRNA multivalency (which can coexist in the same transcript) plotted for nuclear biasing, cytoplasmic biasing, or mRNAs not changing subcellular bias from n = 4 biological replicates (fracRNA-seq data). (B) Co-staining of *oligo(dT)* with the nuclear speckle markers Srrm2 and phosphor-SR in aNSCs versus qNSCs of increasing quiescence depth. All scale bars: 10 µm. (C) Quantification of *oligo(dT)* and ph

Full text

Introduction

The importance of stem cell quiescence has gained recognition, with acknowledgement of its centrality to tissue development, homeostasis and repair (often reliant on stem cell reactivation); cancer (quiescent cancer stem cells evading cytostatic and immune therapies, and able to reconstitute the original disease); and, in the mammalian adult central nervous system, learning, memory, and mood regulation (following reactivation of hippocampal neural stem cells (NSCs) and generation of new neurons and glia) (Ma et al, 2009; de Morree and Rando, 2023).

Challenges of studying quiescence include absence of positive markers, rendering quiescent cells hard to identify; as well as its reversibility and heterogeneity, it being a continuum of states between near-active (“shallow”) and profound quiescence (dormancy), with depth defined by reactivation speed (Ma et al, 2009; Rodgers et al, 2014; Llorens-Bobadilla et al, 2015; Kwon et al, 2017; de Morree and Rando, 2023). Nonetheless, we now appreciate that across species and cell types diverse signals converge on the Target of Rapamycin (TOR) pathway towards reversibly pausing the cell cycle and diminishing the very resource-intensive cellular process of protein translation (Paik et al, 2009; Renault et al, 2009; Chell and Brand, 2010; Sousa-Nunes et al, 2011; Paliouras et al, 2012; Valcourt et al, 2012); and that it is tightly controlled not only at the transcriptional level (including its overall lowering), but by several post-transcriptional mechanisms, and even post-translationally (Urbán et al, 2016; Marescal and Cheeseman, 2020; de Morree and Rando, 2023).

Drosophila NSCs are a powerful model with which to study quiescence (Ding et al, 2020). Among ~100 pairs of central brain NSCs and hundreds in the ventral nerve cord, all but five pairs undergo quiescence during ~24–48 h intervening between embryonic and postembryonic neurogenesis (Truman and Bate, 1988). Here, we show that downregulation of nucleocytoplasmic transport factors in Drosophila NSCs is sufficient to induce quiescent-like features, that Drosophila and mouse NSCs have a higher nucleocytoplasmic ratio of polyadenylated (poly(A)) RNA in quiescence, and that nuclear biasing of poly(A) RNA is also found in quiescent human stem cell types. We demonstrate that nuclear biasing of mRNAs is strongly associated (>80% cases) with decreased levels of corresponding proteins in mouse quiescent NSCs (qNSCs) and propose that this mechanism constitutes an additional layer of post-transcriptional regulation of quiescence that enables fast protein upregulation in response to appropriate cues. Globally, increased nucleocytoplasmic ratio of mRNA in quiescence uncouples the proteome from the transcriptome and constitutes a previously unknown hallmark of this state. Finally, we show that nuclear biasing was attributable to functionally related transcripts, many also related at the level of primary structure, enriched in GA multivalency specifically. Multivalent mRNA regions are sequences enriched in certain nucleotide repeats (reinforced by conserved codon biases) that encode intrinsically disordered protein domains recognised by specific RNA-binding proteins (RBPs), and likely to form condensates (Faraway et al, 2025). We find that GA-rich multivalent transcripts relocate from the nucleoplasm to nuclear speckles (ribonucleoprotein condensates; Chaturvedi and Belmont, 2024; Faraway et al, 2025) in consort with serine/arginine (SR)-rich proteins, and propose that this contributes to nuclear retention of mRNAs in qNSCs.

Results

Downregulation of novel protein in Drosophila NSCs induces anachronic quiescence

In addition to paused proliferation, Drosophila quiescent and active NSCs (aNSCs) differ also in soma diameter (~4 μm in early larvae, when quiescent; ~10–12 μm in late larvae, when active (Chell and Brand, 2010)); expression levels of NSC markers (downregulating the cortical protein Miranda (Mira) and the HES-family transcriptional repressor Deadpan (Dpn) below detectability in many qNSCs with some antibodies (this study)); and morphology (presenting a cytoplasmic extension/fibre of unknown function only when quiescent (Truman and Bate, 1988)) (Fig. 1A,D). Following larval hatching and feeding, qNSCs reactivate in a stereotypical spatiotemporal pattern (Truman and Bate, 1988), with soma enlarging, markers becoming detectable in all NSCs, and fibres passed on to transit-amplifying daughters (Bostock et al, 2020) (Figs. 1A,D and  EV1). At late larval stages, all NSCs are active. In a forward-genetic ethyl methane-sulphonate (EMS) screen, we recovered a Drosophila mutant (2V327) in which late larval NSCs, normally active, were cell-cycle arrested whilst displaying a fibre, features of qNSCs (Figs. 1A,B and  EV1A,B). The phenotype was seen in whole homozygous animals as well as homozygous clones induced in early larvae (Fig. 1B), demonstrating cell-autonomy and derivation from mitotic cells (basis of labelled clone generation; Lee and Luo, 1999). We hypothesised that the 2V327 mutation led to anachronic quiescence reentry of NSCs.Figure 1. Downregulation of novel Drosophila protein Snx induces anachronical qNSCs.(A) In wild-types (WT), qNSCs are present in “early” (newly hatched) but not “late” (wandering third-instar stage) larvae, display a cytoplasmic fibre (arrowheads) and downregulate Mira and Dpn. NSC > GFP was grh-GAL4, UAS-mCD8::GFP. (B) Late larval NSCs of the 2V327 mutant or CG14712^RNAi^ (GL00466) have a morphology reminiscent of qNSCs and lack PH3. MARCM clones (see Methods) were by labelled by NLS::GFP and NSC > CG14712^RNAi^ was nab-GAL4, UAS-CG14712^GL00466^. All scale bars: 10 µm. (C) Sequencing of 2V327/WT reveals a premature STOP codon in the mutant at aminoacid position 254 of Snx, whose structural motifs are indicated. (D) Quantification of cells per brain lobe for markers and in genotypes indicated; RNAis for snx were induced with nab-GAL4. Key of histograms with multiple datasets: The total height of the histograms represents the number of detectable Dpn cells; within these, the total height of the black-containing portion represents the number of detectable Mira cells; within those, if cells had fibres or PH3 (mutually exclusive categories in more than 99% cases) the black bar was hatched with magenta or cyan, respectively. Histograms represent mean and error bars, s.e.m. from n ≥5 animals per genotype. Student’s t tests were performed for the total number of detectable Dpn cells; phenotype averages were: WT early 50.8; WT late (wandering third-instar stage) 96.8; snx^2V327^ late 37.9; snx^2V327^/Df late 29.8; snx^RNAi^ (18347GD) late 81.6; snx^RNAi^ (GL00466) late 47.7. Student’s t test significance is indicated for the total number of Dpn^+^ cells, **P < 0.01, ****P < 0.0001. Exact P values (available on PRISM only when p > 0.0001) were snx^RNAi^ (18347GD) late vs WT late: 0.0035; and vs snx^RNAi^ (GL00466) late: 0.0001. ANOVA test from all variables: ****P < 0.0001. (E) Schematic of experimental design to transiently induce RNAi expression in larval NSCs via temperature shifts: embryos were reared at 25 °C and newly hatched larvae were then placed at either 18 °C (green line), for no RNAi induction, or at 31 °C (red bar), for maximal RNAi induction; some of the latter were placed back at 18 °C to stop RNAi induction at the times indicated; B = before or R = after Recovery ‘R’ period (recovery from RNAi-induced phenotypes); ALH, after larval hatching; beyond 96 h ALH larvae were “late” and all animals were dissected at white prepupal stage. RNAis (UAS-snx^GL00466^ and UAS-Cherry^BL35785^) were induced with nab-GAL4. Quantifications of categories as per key in (D). Histograms represent mean and error bars s.e.m. from n = 10 animals per condition. Student’s t tests were performed for the total number of detectable Dpn cells; phenotype averages were: mCherry^RNAi^ 18 °C 94.1; snx^RNAi^ 18 °C 97.6; 48 h mCherry^RNAi^ B 95.3; 48 h snx^RNAi^ B 85.2; 48 h snx^RNAi^ R 95.9; 72 h mCherry^RNAi^ B 97.5; 72 h snx^RNAi^ B 84.55; 72 h snx^RNAi^ R 94.95; mCherry^RNAi^ 31 °C 94.1; snx^RNAi^ 31 °C 46.9. Student’s t test significance is indicated for the total number of Dpn^+^ cells, *P < 0.05, ****P < 0.0001. Exact P values (available on PRISM only when P > 0.0001) were mCherry^RNAi^ 18 °C vs snx^RNAi^ 18 °C: 0.9404; 48 h snx^RNAi^ R vs 48 h mCherry^RNAi^ B: 0.6278; 72 h snx^RNAi^ R vs 72 h mCherry^RNAi^ B: 0.0326. ANOVA test from all variables: ****P < 0.0001 except for the number of Dpn^+^ NBs between 48 h mCherry^RNAi^ B, 48 h snx^RNAi^ B, and 48 h snx^RNAi^ R, which does not change significantly. Source data are available online for this figure.

Deficiency mapping exposed a small genomic region responsible for the 2V327 phenotype, with hemizygote animals recapitulating that of homozygotes (Figs. 1B and  EV1C). Amongst seven protein-coding genes within the candidate region, RNA interference (RNAi) for only one, CG14712, phenocopied the 2V327 mutant (Fig. 1B). Genomic sequencing of CG14712 exons in 2V327 heterozygous animals uncovered a premature STOP codon, consistent with its disruption causing the phenotype (Fig. 1C). We named this previously uncharacterised gene snorlax (snx). Snx is predicted to be a nucleoporin (Nup), i.e., constituent protein of the nuclear pore complex (NPC), of the Phenylalanine-Glycine (FG) repeat category (Flybase.org; Fig. 1C).

Homozygous and hemizygous snx^2V327^ animals were both lethal as undersized late larvae, and quantification of NSC features showed comparable phenotypes (Fig. 1D). 2V327 is thus a strong loss-of-function allele, likely a null. Appreciably fewer than the customary ~100 cells per central brain lobe were detectable with Mira and Dpn antibodies in snx^2V327^ late larvae (Fig. 1D), but staining for the apoptotic marker Death Caspase-1 (Dcp-1) was negative in the NSCs examined (Fig. EV1D). Overall, reemergence of a fibre accompanying cell-cycle arrest, along with downregulation of Mira and Dpn, is consistent with late larval quiescence reentry in snx mutant and knockdown NSCs.

Reversibility is a defining feature of quiescence (de Morree and Rando, 2023). We were unable to rescue the snx^2V327^ phenotype, unsurprisingly given the strict stoichiometric requirements for Nups, with overexpression phenotypes such as impaired cell growth or viability, possibly due to dominant-negative effects (Davis and Fink, 1990; Wozniak et al, 1994; Boer et al, 1998; Boeglin et al, 2016). Nonetheless, to test reversibility of the snx loss-of-function phenotype, we transiently induced snx^RNAi^ in NSCs followed by a period of recovery, and analysed animals before, B, and after recovery, R (Fig. 1E). Cells undetectable by anti-Mira or anti-Dpn before recovery reemerged as visible with these following the recovery period, demonstrating that they had neither died nor differentiated. Furthermore, recovery also decreased the number of NSCs displaying fibres and increased the mitosis index (reported by expression of phospho-histone H3, PH3), indicating a shift of qNSCs to aNSCs (Fig. 1E). Like for snx, transient knockdown in NSCs of another gene encoding two Nups, Nup98-96 (Fontoura et al, 1999), also induced reversible downregulation of Dpn, Mira and PH3 accompanied by a fibre (Fig. EV1E). We concluded that, by the criteria enumerated above, snx^RNAi^ and Nup98-96^RNAi^ NSCs reenter quiescence anachronically.

Downregulation of many nucleocytoplasmic transport factors in Drosophila NSCs induces quiescence features

The NPC is the evolutionarily conserved gateway for bidirectional transport between nucleus and cytoplasm in eukaryotic cells (Cautain et al, 2005). It is assembled from multiple copies of ~30 distinct Nups (Dataset EV1), a third of which contain repeated FG motifs that form high-specificity low-affinity interactions with cargo complexes (Aitchison and Rout, 2012; Breuer and Ohkura, 2015). We wondered whether qNSC might also be induced by downregulation of other Nups, spread across structural classes (Dataset EV1). NSC-specific knockdown of 17 out of 27 further Nups induced features of quiescence with at least one RNAi, 12 of which with more than one (Fig. EV2A). Negative outcomes could be due to the target having no role in quiescence regulation, ineffective RNAi (although some appeared effective in other contexts; Breuer and Ohkura, 2015; Boeynaems et al, 2016; Jahanshahi et al, 2016), long Nup protein half-life (Savas et al, 2012), or maternal contribution (Luschnig et al, 2004). The fact that knockdown of multiple Nups among all structural classes induced features of quiescence suggested that disruption of their function at the NPC caused the phenotype. Nonetheless, some Nups have functions beyond nucleocytoplasmic transport, such as transcriptional and microtubule regulation (Jühlen and Fahrenkrog, 2018). To verify that nucleocytoplasmic transport perturbation could underlie qNSC induction, we knocked down other transport components. Whilst small proteins (<40 kDa) can passively diffuse across the NPC, efficient distribution of larger ones between nucleus and cytoplasm depends on active transport fuelled by the small guanosine nucleoside triphosphate (GTP) hydrolysing enzyme Ran. This employs another set of evolutionarily conserved proteins, karyopherins (Kimura et al, 2017; Mackmull et al, 2017) (Dataset EV2), which bind nuclear localisation and/or export signals in protein cargo to facilitate their translocation across the NPC (Aitchison and Rout, 2012). Knockdown of Ran, its GTPase-activating protein (RanGAP) and guanine nucleotide exchange factor (RanGEF), or of a small subset of karyopherins, also induced qNSC features (Fig. EV2B). Particularly, knockdown of Exportin-1 and −2, Importin-ß, Tnpo and Tnpo-SR led to strong phenotypes with at least two RNAis (Fig. EV2B). In all, downregulation of many, yet specific, nucleocytoplasmic transport mediators in Drosophila NSCs induces quiescence features.

Poly(A) RNA becomes more biased to the nucleus in Drosophila qNSCs

Karyopherins mediate the transport of functionally related proteins (Kimura et al, 2017; Mackmull et al, 2017). Biological process gene ontology (GO) analysis of known cargo for the few karyopherins whose knockdown induced qNSCs (Kimura et al, 2017; Mackmull et al, 2017) (Fig. EV2B) led us to hypothesise that messenger RNA (mRNA) metabolism might be altered in qNSCs versus aNSCs. We also noticed that Nups with analogous phenotype (Fig. EV2A) included those of the so-called mRNA export platforms (Kim et al, 2018). mRNA accounts for most poly(A) RNA, and we considered that there might be an altered distribution of poly(A) RNA between aNSCs and qNSCs. In situ hybridisation with a fluorescently-labelled oligo(dT) probe reported visibly lower poly(A) RNA in Drosophila qNSCs than aNSCs, as expected from diminished transcription in quiescence (Fig. 2A). We reproducibly found relatively large discrete poly(A) RNA puncta within nuclei of deeply quiescent NSCs (newly hatched larvae), and in permanently active mushroom body NSCs most poly(A) was cytoplasmic (Fig. 2A). By quantifying poly(A) fluorescence in nuclear and cytoplasmic compartments of NSCs in the two states, we discovered that the nuclear/cytoplasmic ratio of poly(A) RNA was higher in qNSCs than in the mushroom body aNSCs (Fig. 2A,B). We also determined the nuclear/cytoplasmic poly(A) RNA ratio in the same NSC (across specimens) as it reactivated, and found the ratio to decrease during reactivation (Fig. 2A,B). We concluded that nuclear biasing of poly(A) RNA is a trait of Drosophila qNSCs.Figure 2Drosophila and mouse qNSCs accumulate nuclear poly(A) RNA relative to aNSC.(A) Permanently active (mushroom body, MB) NSCs (labelled with nab-GAL4) are larger and contain more poly(A) RNA than qNSCs. Arrowheads: poly(A) RNA accumulations within the nucleus of deeply quiescent NSC (0–4 h ALH). As reactivation progressed (single lineage labelled with grh-GAL4 identified across time and specimen), the same qNSC gradually increased poly(A) RNA and its distribution shifted towards the cytoplasm (split magenta channel with hatched NSC outlines and nuclei outlined in blue). (B) Quantifications from individual cells, such as those depicted in (A). Histograms represent mean, and error bars s.e.m. from n ≥5 animals per stage; values normalised to aNSC (MB) average were: 0–4 h ALH VNC 16.9; 24–29 h ALH VNC 6.9; 29–34 h ALH VNC 5.7. Mann–Whitney test vs MB P values: **P \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\le \,$$\end{document} 0.001. Exact P values were 0–4 h ALH VNC: 0.0047; 24–29 h ALH VNC: 0.0048; 29–34 h ALH VNC: 0.0047. (C) Adult mouse hippocampal NSC cultures; quiescence induced by BMP4. Insets are single-channel high magnification of single cells (arrowheads in low magnification) with the nucleus outlined in blue. (D) Quantification from specimens such as those depicted in (C) was pooled from n = 5 biological replicates (cells cultured on different days and/or from different animals or passages). Histograms represent mean and error bars s.e.m.; values normalised to aNSCs in each of ‘nuc’, ‘cyto’ or ‘nuc/cyto’ averages were: qNSC^nuc^ 0.88; qNSC^cyto^ 0.48; qNSC^nuc/cyto^ 1.83. Mann–Whitney test aNSC vs qNSC: **P < 0.001, ****P < 0.0001. Exact P values are only available on PRISM when P > 0.0001. (E) Adult hippocampal sections from transgenic Nestin-EGFP mice where NSCs were identified by Nestin::GFP, and active and quiescent states identified by the presence/absence of mKi67 (filled and open arrowheads, respectively). (F) Quantification from specimens such as those depicted in (E) from n = 1 animal. Histograms represent mean and error bars s.e.m.; value normalised to aNSCs average was: qNSC 1.64. Mann–Whitney test: *P < 0.05. Exact P value was 0.0387. (G) Quantification from in vivo adult mouse hippocampal NSCs identified by Gfap and Sox2 staining, with active and quiescent identified by the presence/absence of Mcm2 from n = 3 animals. Histograms represent mean and error bars s.e.m.; value normalised to aNSCs; average was: qNSC 1.54. Mann–Whitney test: ****P < 0.0001. In the histograms, green indicates a significantly higher average than the control (grey), while red indicates a significantly lower one. Source data are available online for this figure.

Nuclear biasing of poly(A) RNA is a hallmark of quiescence

Nuclear retention of transcripts would be an efficient way for quiescent cells to reduce protein synthesis whilst remaining able to quickly resume it in response to reactivation cues. We thus reasoned that this might be a widespread mechanism of quiescence. We examined primary adult mouse hippocampal NSCs, which we previously showed become quiescent after 3 days in bone morphogenetic protein 4 (BMP4), a known niche component (Blomfield et al, 2019). Synchronicity of quiescence induction in culture allowed us to determine relative levels of nuclear and cytoplasmic poly(A) RNA across many cells in an equivalent state. Both nuclear and cytoplasmic poly(A) signals decreased in qNSCs compared to aNSCs, as expected by diminished transcriptional activity, but cytoplasmic levels consistently decreased more, resulting in increased nuclear/cytoplasmic poly(A) ratio (Fig. 2C,D). Inspection of hippocampal NSCs in vivo also showed increased nuclear/cytoplasmic poly(A) ratio in quiescent versus active cells (Fig. 2E–G). Separation between the two conditions was less marked in the tissue, given the continuity of states within the niche and no means of determining how long each cell had been quiescent or active for, but the magnitude of relative poly(A) RNA nuclear bias was remarkably similar to that measured in vitro. Surveying of active and quiescent human blood marrow hematopoietic stem and progenitor cells returned the same finding (Fig. EV3). We concluded that nuclear biasing of poly(A) RNA is an evolutionarily conserved hallmark of quiescence across species and cell types, in physiological and diseased contexts.

Most nucleocytoplasmic transport factors, mRNA export, and splicing factors are downregulated in qNSCs

We next took advantage of the mouse synchronous monoculture system to perform proteome comparison between mouse hippocampal aNSCs and qNSCs, both as a global overview, not reported before, and to enquire about relative levels of nucleocytoplasmic transport factors, mRNA export machinery, and splicing factors (since only fully spliced mRNAs are export-competent; Aitchison and Rout, 2012). We prepared protein extracts from NSCs on different days post-BMP4 addition (0 days being aNSCs), having ascertained that longer exposure to BMP4 corresponded to deeper quiescence (Fig. EV4). Longitudinal proteome profiling was carried out by tandem mass tag spectrometry (TMT; Dayon et al, 2008), and principal component analysis (PCA) showed that BMP4 exposure accounted for most protein level changes already within 3 days (Fig. 3A), our standard quiescent condition. In total, 5771 proteins were reliably identified and quantified in all fractions, with over four-fifths changing consistently in either direction, and nearly as many upregulated as downregulated over 21 days of BMP4 exposure (Fig. 3B; Dataset EV1; ProteomeXchange Consortium PXD037487). Validating the data, proteins whose levels are known to be altered in qNSCs behaved as expected in both directions (Blomfield et al, 2019; Morrow et al, 2020) (Fig. 3C; Dataset EV1).Figure 3. Proteome changes as NSCs transition between active and deeper quiescence states.(A) PCA plot. (B) TMT spectrometry protein intensity changes (log_2_ difference, arbitrary units) from day 0 (aNSCs); mean and s.d. indicated for each time-point with n = 1 biological replicate for each of 10 time points. (C) Control proteins plotted on the background of all. (D) Most upregulated (5, green) and downregulated (10, magenta) biological process GO terms in 21d-qNSCs. (E) Change of specified categories of proteins plotted on the background of all. Source data are available online for this figure.

Biological process GO analysis revealed the most downregulated categories of proteins in qNSCs as (m)RNA processing and metabolism, as well as DNA organisation and metabolism, while the most upregulated categories pertained to oxidative phosphorylation and cell adhesion (Fig. 3D; Dataset EV2). We found decreasing levels of most nucleocytoplasmic transport factors when NSCs transitioned from active into deep quiescence (having detected 23 Nups, representing all structural classes and including Nup155 and Tpr, which play key roles in mRNA export (Scott et al, 2019); and 26 karyopherins or Ran exchange factors) (Fig. 3E; Dataset EV1). Most mRNAs are exported from the nucleus to the cytoplasm as messenger ribonucleoprotein complexes (mRNPs) via a non-discriminatory bulk mRNA export pathway that relies on ATP (rather than GTP), karyopherins, and certain Nups. We detected 38 mRNA export factors, of which 37 were downregulated in qNSCs (Fig. 3E; Dataset EV1). Nonetheless, there is evidence for, largely uncharacterised, selective mRNA export pathways (Scott et al, 2019). Only a few nuclear retention factors are known (Wegener and Müller-McNicoll, 2018), and those detected in our proteomic analysis were also downregulated (Dataset EV1). Levels of factors involved in mRNA decay in either compartment (Łabno et al, 2016; Schmid and Jensen, 2018; Puno et al, 2019) were not appreciably altered by quiescence induction, suggesting that none of these were responsible for the observed nuclear-biasing of poly(A) RNA in qNSCs either (Dataset EV1). We detected 228 splicing factors (GO term 0008380), of which 204 were downregulated in qNSCs; intriguingly, 12 were upregulated (Fig. 3E; Dataset EV1).

Nucleocytoplasmic biasing of specific transcripts uncouples proteome and transcriptome in qNSCs

We next wondered what transcripts accounted for the nuclear biasing of poly(A) RNA in qNSCs. To determine this, we fractionated cultured mouse hippocampal aNSCs and qNSCs into nuclear and cytoplasmic compartments (Fig. EV5A) and performed RNA-seq on each (fracRNA-seq). We prepared samples from two quiescence depths, corresponding to 3 and 10 days BMP4 exposure (qNSC-3d and qNSC-10d). PCA showed progressive divergence of BMP4-treated NSCs from aNSCs as a function of time in BMP4, and clustering of samples according to subcellular compartment (Fig. 4A). Transcripts of 30,265 genes were detected in all samples, of which 19,979 (66%) were protein-coding; about 95.5% of reads mapped to protein-coding genes (Dataset EV3; Gene Expression Omnibus GSE162047). For transcripts pertaining to each gene, regardless of isoform, levels, or splicing status, we determined the proportion of exonic reads in the nucleus versus cytoplasm from paired extracts for each sample. Each gene was ascribed a bias-score Z = log_2_ [number of nuclear reads/number of cytoplasmic reads], with Z > 0 indicating nuclear-bias and Z < 0 cytoplasmic-bias (Dataset EV3). Validating the quality of the fractions and analyses, transcripts with known bias for either compartment (Yap et al, 2018) behaved as expected in all samples (Fig. 4B; Dataset EV1).Figure 4. Nuclear retention of transcripts increases as NSCs transition between active and deeper quiescence states.(A) PCA plot. (B) Z-score at 0, 3 and 10 d in BMP4 for marker transcripts for each subcellular compartment. Nuclear transcripts: nuclear paraspeckle assembly 1 (Neat1), metastasis-associated lung adenocarcinoma transcript 1 (Malat1), plasmacytoma variant translocation 1 (Pvt1). Cytoplasmic transcripts: small subunit ribosomal protein 14 (Rps14), glyceraldehyde-3-phosphate hydrogenase (Gapdh), 7S RNA 1(Rn7s1). (C) Z-scores at 0, 3 and 10 d in BMP4 for the transcripts detected across all conditions (omitting eleven data points out of range); white lines: median and quartile boundaries from n = 4 biological replicates for each time point. Number of significantly nuclear biasing (magenta) or cytoplasmic biasing (green) genes from pairwise comparison of each time-point, and overall from likelihood ratio testing across the time-course. (D) Most nuclear-biased (10, magenta) and cytoplasmic-biased (10, green) biological process GO terms in qNSCs relative to aNSCs. Fisher’s exact test. (E) Nuclear- or cytoplasmic-biased transcripts largely predict directionality of protein levels change (shown correlation at 10 versus 0 d BMP4): 80% proteins downregulated corresponding to nuclear-biased transcripts and 77% proteins upregulated corresponding to cytoplasm-biased ones. (F) Protein expression levels (scaled to normalise range of expression between samples) plotted against corresponding whole-cell mRNA expression levels (assembled in silico from fracRNA-seq data and also scaled for the same purpose). Concordance between these was assessed by a Pearson’s correlation test (P values unchanged by scaling). Note that, unlike in aNSCs, in the qNSCs highly expressed mRNAs do not correlate with high protein levels, as concluded from the non-significant P values and the poor correlation coefficient, from which it can be seen that the proteome becomes uncoupled from the transcriptome. (G) Intersection of proteomic and transcriptomic data suggests that downregulation of nucleocytoplasmic transport factors is itself subject to regulation via mRNA nuclear bias even at early stages qNSC induction. Fisher’s exact test. Source data are available online for this figure.

The median bias-score of all transcripts increased as cells shifted from active to increasingly deeper quiescence, consistent with the nuclear biasing observed with the oligo(dT) probe (Fig. 4C). In aNSCs, transcripts from 12,096 genes were significantly biased to either subcellular fraction, 5637 (47%) of which were nuclear-biased (3026 more than fourfold, i.e., |Z | > 2). In qNSCs-3d, transcripts from 12,812 genes were significantly biased, 6601 (52%) of which were nuclear-biased (3470 more than fourfold). In qNSCs-10d, transcripts from 12,227 genes were significantly biased, 7653 (63%) of which were nuclear-biased (4106 more than fourfold). In sum, the proportion of nuclear-biased genes increased with quiescence (Fig. 4C).

We next performed pairwise comparisons of individual genes between each condition and likelihood ratio test of individual genes across the time-course to identify genes that changed overall (Fig. 4C; Dataset EV3). Consistent changes, i.e. in the same direction across the time-course, comprised genes whose products swapped bias in subcellular compartments as well as those whose Z-scores changed significantly even if retaining the same compartment bias. Transcripts of 388 genes changed significantly between aNSCs and qNSCs-3d, of which 247 (64%) became more nuclear (20 more than fourfold); 2584 changed significantly between aNSCs and qNSCs-10d, of which 2409 (93%) became more nuclear (311 more than fourfold); and 590 changed significantly between qNSCs-3d and qNSCs-10d, of which 571 (97%) became more nuclear (128 more than fourfold) (Fig. 4C; Dataset EV3). Transcripts of 2173 genes (86% protein-coding) changed subcellular bias significantly and consistently (i.e., in the same direction across the time-course). Of these, transcripts of 1616 genes (74%) became increasingly nuclear (92% protein-coding). In summary, transcripts for most genes had no significant subcellular bias in any condition nor, if they had, did the direction or magnitude of bias change significantly or consistently between active and quiescent states. Notwithstanding, transcripts of more than 2000 genes, mostly protein-coding, did change subcellular bias significantly, three-quarters of which became increasingly nuclear as NSCs shifted from active into deeper quiescence.

Biological process GO analysis revealed most nuclear-biased categories of transcripts in qNSCs as (m)RNA processing and RNP biogenesis, as well as DNA organisation and metabolism (Fig. 4D; Dataset EV4), in agreement with the downregulated proteome (Fig. 3D; Dataset EV2). Cytoplasmic-biased transcripts in qNSCs were enriched in actin cytoskeleton organisation GO terms, consistent with an enriched “cell adhesion” GO term in proteomics (Fig. 3D; Dataset EV2), and our observations of increased morphological complexity upon quiescence induction (note insets in Fig. 2D and fibres in Drosophila qNSCs). Moreover, we observed >80% concordance between increased nuclear- and cytoplasmic biasing of transcripts and respective protein down- or upregulation in qNSCs (Figs. 4E and  EV5B). This revealed a major impact of subcellular partitioning of mRNAs on the directionality of corresponding protein level changes in qNSCs, so impactful that it results in global transcriptome and proteome uncoupling in qNSCs (Fig. 4F). We concluded that altered nucleocytoplasmic distribution of transcripts in qNSCs contributes greatly to regulation of protein expression and renders often-used single-cell or -nuclei transcriptomics a poor proxy to know the proteome of quiescent cells.

Intersection of our proteomic and transcriptomic data points at downregulation of nucleocytoplasmic transport factors being an early event in qNSC induction, itself subject to regulation via mRNA nuclear bias (Fig. 4G), suggesting that this is a feedforward mechanism underlying shifting of quiescence depth.

Nuclear biasing transcripts are generally more completely spliced in qNSCs than in aNSCs

To assess whether nuclear biasing of transcripts during quiescence could be due to decreased splicing, we analysed intron retention (IR). For each intron of transcripts found in each subcellular fraction, we ascribed an IR-score = [number of intronic reads/(number of intronic reads + number of spliced reads)], with 0 < IR < 1. Most significant IR changes were observed in nuclear fractions as expected (Fig. EV6A,B; Dataset EV5), with those in cytoplasmic fractions an order of magnitude lower (Fig. EV6C,D). In nuclear fractions, 3362 introns corresponding to 1708 genes showed significant IR changes between at least two conditions (Fig. EV6B). Of these, most (65%) corresponded to transcripts that became increasingly nuclear in qNSCs (Fig. EV6B, large pie chart; Dataset EV5). Introns of a single gene usually showed the same directionality of IR change as NSCs shifted from active into deeper quiescence (Dataset EV5). Surprisingly, the vast majority (84%) of significant IR changes in the nucleus consisted of IR decrease (Fig. EV6B, small pie charts). In summary, despite a small and potentially meaningful group of transcripts with increased IR, most nuclear biasing transcripts were in fact more completely spliced. We concluded that nuclear biasing of transcripts in qNSCs is not generally due to decreased splicing.

In agreement with the categories of transcripts that became increasingly nuclear as quiescence deepened, biological process GO terms for the 934 genes encoding nuclear-biased transcripts with decreased IR in nuclear fractions, pertained to (m)RNA processing and chromatin regulation; and those for the 98 nuclear-biased transcripts with increased IR pertained mostly to mitosis, followed by (m)RNA processing (Dataset EV6). This suggests that different functional categories of transcripts are regulated differently and will be ready sooner than others for nuclear export.

A large proportion of nuclear biasing mRNAs in qNSCs have GA-rich multivalency and accumulate in nuclear speckles

The most common multivalent mRNA sequences contain GA, GC, C, CAG, or CUG repeats (Faraway et al, 2025). It was recently discovered that a protein-mRNA feedback mechanism, named interstasis, induces retention of GA-rich mRNAs in nuclear speckles (Faraway et al, 2025). Having noticed a strong similarity between the biological GO terms corresponding to nuclear-biased transcripts and those rich in purine multivalency (Faraway et al, 2025), we enquired into the contribution of multivalent sequences to nuclear-biased transcripts in qNSCs. We found that although transcripts contained various types of multivalency regardless of biasing (or lack thereof), nuclear-biased transcripts were the only group that was specifically enriched in GA multivalency, while displaying lower than average representation of other types of multivalency (Fig. 5A).Figure 5. Nuclear-biased transcripts in qNSCs have overrepresentation of GA-rich multivalent mRNAs and are enriched in nuclear speckles.(A) Average fraction of transcripts containing five common types of mRNA multivalency (which can coexist in the same transcript) plotted for nuclear biasing, cytoplasmic biasing, or mRNAs not changing subcellular bias from n = 4 biological replicates (fracRNA-seq data). (B) Co-staining of oligo(dT) with the nuclear speckle markers Srrm2 and phosphor-SR in aNSCs versus qNSCs of increasing quiescence depth. All scale bars: 10 µm. (C) Quantification of oligo(dT) and phosphor-SR speckle/nucleoplasmic signal from individual cells such as those depicted in (B) from one of n = 2 biological replicates; white lines: median and quartile boundaries. Median values displayed over dot plots. Phenotype averages were normalised to aNSC and were: oligo(dT) in 3d-qNSC 1.43 and in 10d-qNSC 1.71; phospho-SR in 3d-qNSC 1.30 and in 10d-qNSC 1.98. Mann–Whitney test: ****P < 0.0001. ANOVA test for each: ****P < 0.0001. (D) Quantification of Psap, Eif3a and Map1B mRNA speckle/nucleoplasmic signal from individual cells in aNSCs versus qNSCs of increasing quiescence depth (representative images for each transcript can be found in Fig. EV6) from one of n = 2 biological replicates; white lines: median and quartile boundaries. Median values are displayed over dot plots. Phenotype averages were normalised to aNSC and were: Psap in 3d-qNSC 1.62 and in 10d-qNSC 1.26; Eif3a in 3d-qNSC 2.37 and in 10d-qNSC 2.61; Map1b in 3d-qNSC 2.28 and in 10d-qNSC 1.61. Mann–Whitney test, ***P < 0.001, ****P < 0.0001. ANOVA test for each probe: ****P < 0.0001. (E) Model: altered speckle composition in qNSCs, namely with increased SR proteins, turn it into a “sink” for some mRNAs, namely GA-rich multivalent ones, which become nuclear-retained. Source data are available online for this figure.

Nuclear speckles are membraneless nuclear compartments within the interchromatin space made up of protein-RNA condensates, particularly enriched in spliceosome components, including the SR-rich family of proteins (Chaturvedi and Belmont, 2024). Experimental enlargement of speckles has been found to have a dose-dependent effect on nuclear biasing of endogenous GA-rich mRNAs (Faraway et al, 2025). We thus enquired whether nuclear-biased mRNAs in qNSCs were enriched in speckles. Using the Serine/Arginine Repetitive Matrix 2 (SSRM2) protein as a speckle marker (Chaturvedi and Belmont, 2024), we found increased speckle/nucleoplasm localisation of nuclear poly(A) RNA in response to induction and deepening of qNSCs (Fig. 5B,C). SR proteins can bind to GA-multivalent regions and, if hypophosphorylated, relocalise from the nucleoplasm to speckles (Faraway et al, 2025). Using an anti-phospho-SR antibody, we found a significant increase in speckle/nucleoplasm phospho-SR signal with induction and deepening of qNSCs (Fig. 5B,C). We then took as examples 3 transcripts either poor or rich in GA multivalency (Faraway et al, 2025; Prosaposin: PSAP as negative control; Eukaryotic translation initiation factor 3 subunit A: EIF3A, and microtubule-associated protein1B: MAP1B as positive controls) and found that only the positive controls relocalised from the nucleoplasm to the speckle with induction and deepening of qNSCs (Figs. 5D and  EV7A–C). In all, we concluded that nuclear biasing of at least GA-rich multivalent transcripts is regulated by retention in nuclear speckles with increased abundance of SR proteins. This also suggests that the composition of these organelles is altered in qNSCs versus aNSCs.

Discussion

Our work implicates altered nucleocytoplasmic partitioning of mRNA as a major regulator of NSC quiescence that contributes significantly to lower protein synthesis and ultimately leads to the uncoupling of the proteome from the transcriptome. Though we are the first to report this large-scale layer of post-transcriptional gene regulation in quiescence, altered nucleocytoplasmic partitioning of specific factors has been previously found in this state. Studies in human fibroblasts and zebrafish NSCs have reported Exportin-1-dependent changes in microRNA biogenesis and localisation in quiescence, including accumulation of microRNA-9 and Argonaute proteins in qNSC nuclei (Katz et al, 2016; Martinez et al, 2017). In Drosophila, a transient low-level nuclear pulse of the homeobox transcription factor Prospero (Pros; matching that of its adaptor Mira, shown in Fig. EV1) induces NSC quiescence, and RanGEF/Rcc1/Bj1 has been implicated in excluding Pros from aNSC nuclei to allow their self-renewal (Lai and Doe 2014; Joy et al, 2015). The work here presented shows these to be glimpses into extensive nucleocytoplasmic alterations in quiescence.

We started by showing that attenuation of various nucleocytoplasmic transport factors in Drosophila NSCs induces quiescence features, and that most of those factors are downregulated in mouse NSCs in response to a physiological quiescence signal. We then identified mRNA as a category of cargo that is differentially partitioned between nucleus and cytoplasm in active versus quiescent cells across three different organisms, and propose nuclear biasing of mRNA to be a general mechanism of quiescence that diminishes protein translation. Nuclear residence protects mRNAs from viral nucleases and cytoplasmic decay pathways, and has been observed in reaction to stress or changing cellular conditions, such as differentiation, viral infection or oncogenic transformation, with release enabling rapid cellular responses (Wegener and Müller-McNicoll, 2018). This report is the first to implicate it in quiescence.

In NSCs, we demonstrated substantial and significant nuclear biasing of hundreds of transcripts, with evidence for feedforward mechanisms involving mRNAs and encoded proteins. Remarkably, nuclear or cytoplasmic biasing of transcripts respectively predicted down- and upregulation of the vast majority of encoded proteins, evidencing the impact of mRNA partitioning on protein expression in qNSCs, and explaining our finding of uncoupled proteome and transcriptome in qNSCs. This has important implications within and beyond our field. Uncoupling of proteome and transcriptome has likely contributed to the hitherto failure to identify a unique marker for quiescence in general, and qNSCs in particular. Furthermore, our findings should be taken into consideration when using single-cell or single-nuclei RNAseq to identify cell types and states.

We found that most nuclear-biased mRNAs in qNSCs were completely spliced and encode (m)RNA processing regulators, whereas the minority with increased intron-retention mostly encode cell-cycle and mitotic regulators. We postulate that differential splicing levels between distinct transcript groups could underlie the sequential deployment of factors during quiescence entry and, presumably, the reciprocal during quiescence exit. This will likely orchestrate a finely tuned sequence of events whereby gene products relying differentially on transcription, splicing and/or cytoplasmic translocation will need more or less time for deployment. We note that this could even trigger feedforward loops where earlier factors promote the readiness of others to be expressed, thus favouring a prompt reversibility. A recent study reported widespread IR in various quiescent cell types (Yue et al, 2020) but did not report on qNSCs, nor did we find this in our study. It is reasonable to speculate that different cell types adopt distinct molecular strategies towards the same goal of selective nuclear biasing of transcripts as a general mechanism of translation repression in quiescence.

Several mechanisms could explain why most nuclear-biased transcripts are more fully spliced in quiescence (Fig. EV6B). While nuclear-biased transcripts could have been re-imported in the nucleus in their fully spliced form, this is not parsimonious, especially in a cellular context characterised by low biosynthetic requirements like quiescence. Moreover, we did not find the upregulation of specific mRNA import pathways in our data, which rather supports a scenario where bidirectional transport is downregulated in quiescence. We hypothesise that nuclear-biased transcripts have lower intron retention due to their extended residence times in the nucleus, specifically at the speckle (Fig. 5B,C). It is also possible that some nuclear-biased transcripts benefit especially from the few splicing proteins that are upregulated in qNSC (Fig. 3E).

The fact that nucleocytoplasmic compartmentalisation of most transcripts was not significantly altered between aNSCs and qNSCs, and that more nuclear-biased transcripts were not associated with higher IR, argues against general disruption of mRNA export or splicing in these cells. Since the bulk mRNA export pathway is non-discriminatory and its canonical components were only modestly downregulated in quiescence, possibly commensurately to transcriptional decrease, it is more likely that qNSCs selectively regulate (a) discriminatory mRNA nuclear retention and/or export pathway(s) for a few hundred transcripts. The observations here presented, in conjunction with recent data on GA-multivalent mRNA accumulation in nuclear speckles (Faraway et al, 2025), suggest that altered speckle composition in qNSCs might turn it into a “sink” for nuclear biasing of at least some mRNAs, particularly GA-rich multivalent ones (Fig. 5E). Conversely, it is possible that upon quiescence exit the properties of the speckles change so that nuclear-biased mRNAs can be released, exported to the cytoplasm, and translated to drive reactivation. Work by the Ule lab (Faraway et al, 2025) has implicated SR protein phosphorylation by CDC-like kinases (CLKs) as mediating the composition and properties of the speckle. Future work will explore a potential role for nuclear speckles in quiescence regulation, with a particular focus on whether manipulating speckle properties is sufficient to control the transition between a proliferative and quiescent state.

We envision that the post-transcriptional mechanisms here uncovered interplay with other (post)transcriptional and (post)translational ones in a series of positive-feedback loops, whereby initially small changes in protein or cytoplasmic transcript abundance are amplified, underlying the continuum of states between active and deeply quiescent cells. One key question is how the mechanism uncovered here is integrated with the TOR pathway towards a concerted cellular programme of quiescence entry and exit. Obvious direct connections we looked into were whether TOR kinases regulate SR protein phosphorylation directly (lower TOR leading to SR hyperphosphorylation and consequent enrichment in speckles) or if TOR pathway components were to be GA-multivalency-rich. We found no evidence for these scenarios, and future work will be needed to reveal at what level(s) the two pathways combine coherently.

Methods

Reagents and tools tableReagent/resourceReference or sourceIdentifier or catalogue no. Experimental models

Drosophila melanogaster :

FRT82B BDRC5748 TM3,Sb/TM6B,Tb BDRC3720 Df(3R)ED5516 BDRC8968 Df(3R)Exel8153 BDRC7963 Df(3R)Exel6276 BDRC7743 Df(3R)Exel6161 BDRC7640 Df(3R)ED5514 BDRC8957 Df(3R)ED5518 BDRC9084 Df(3R)ED5559 BDRC8920 Df(3R)Exel8154 BDRC7961 Df(3R)Exel7309 BDRC7960 y,w,hs-FLP[1.22]; tub-GAL4,UAS-NLS::GFP::6xmyc; FRT82B,tubP-GAL80[LL3]/(TM6B) Shaw et al, 2018 FRT82B,2V327/TM6BTb This study UAS-Dcr2 BDRC24646 nab-GAL4 DGRC104533UAS-mCD8::GFP (Drosophila melanogaster)BDRC5137 w;;tub-GAL80[ts] BDRC7108 grh-GAL4 Chell and Brand, 2010 UAS-mira::3xGFP Sousa-Nunes et al, 2009 UAS-mCherry[RNAi] BDRC35785 UAS-CG14712[RNAi] BDRC35622 UAS-CG14712[RNAi] NIGHMJ21299 UAS-Nup107[RNAi] VDRC110759 KK UAS-Nup107[RNAi] VDRC22407 GD UAS-Nup205[RNAi] VDRC38608 GD UAS-Nup205[RNAi] VDRC38610 GD UAS-Nup205[RNAi] NIG11943R-1 UAS-Sec13[RNAi] VDRC110428 KK UAS-Sec13[RNAi] NIG6773R-3 UAS-Sec13[RNAi] NIG6773R-1 UAS-Sec13[RNAi] BDRC32468 UAS-Sec13[RNAi] VDRC50367 GD UAS-Nup154[RNAi] NIG4579R-2 UAS-Nup154[RNAi] VDRC21878 GD UAS-Nup154[RNAi] VDRC106136 KK UAS-Nup154[RNAi] BDRC34710 UAS-Nup154[RNAi] NIG4579R-3 UAS-Nup44A[RNAi] BDRC38357 UAS-Nup44A[RNAi] BDRC32942 UAS-Nup44A[RNAi] NIG8722R-1 UAS-Nup44A[RNAi] NIG8722R-3 UAS-Nup44A[RNAi] VDRC40717 GD UAS-Nup75[RNAi] VDRC27495 GD UAS-Nup75[RNAi] NIG5733R-1 UAS-Nup75[RNAi] BDRC28315 UAS-Nup93-1[RNAi] VDRC100315 KK UAS-Nup93-1[RNAi] VDRC16189 GD UAS-Nup93-1[RNAi] BDRC34090 UAS-Nup93-1[RNAi] BDRC31196 UAS-Nup93-1[RNAi] BDRC33908 UAS-Nup133[RNAi] BDRC58290 UAS-Nup133[RNAi] VDRC110194 KK UAS-Nup160[RNAi] VDRC21937 GD UAS-Nup160[RNAi] NIG4738R-2 UAS-Nup160[RNAi] VDRC109318 KK UAS-Nup160[RNAi] BDRC32391 UAS-Nup160[RNAi] NIG4738R-3 UAS-Nup43[RNAi] NIG7671R-1 UAS-Nup43[RNAi] NIG7671R-4 UAS-Nup43[RNAi] VDRC33645 GD UAS-Nup43[RNAi] VDRC108595 KK UAS-Nup43[RNAi] NIGHMJ21643 UAS-Nup37[RNAi] VDRC16342 GD UAS-Nup37[RNAi] VDRC109814 KK UAS-Nup37[RNAi] NIG11875R-2 UAS-Nup37[RNAi] BDRC62328 UAS-Nup188[RNAi] VDRC102650 KK UAS-Nup188[RNAi] NIG8771R-3 UAS-Nup188[RNAi] NIG8771R-4 UAS-Nup188[RNAi] VDRC36023 GD UAS-Nup188[RNAi] NIGHMJ22993 UAS-Nup35[RNAi] NIG6540R-1 UAS-Nup35[RNAi] NIG6540R-4 UAS-Nup93-2[RNAi] VDRC22552 GD UAS-Nup93-2[RNAi] BDRC51758 UAS-Ndc1[RNAi] VDRC3408 GD UAS-Ndc1[RNAi] VDRC101264 KK UAS-mbo[RNAi] VDRC47693 GD UAS-mbo[RNAi] VDRC47692 GD UAS-mbo[RNAi] NIGNIG 6819R-3 UAS-Nup358[RNAi] BDRC34967 UAS-Nup358[RNAi] VDRC38581 GD UAS-Nup358[RNAi] VDRC38583 GD UAS-Nup358[RNAi] NIG11856R-1 UAS-Nup214[RNAi] BDRC33897 UAS-Nup214[RNAi] VDRC41964 GD UAS-Nup214[RNAi] VDRC330104 UAS-CG16892[RNAi] VDRC23844 GD UAS-CG16892[RNAi] BDRC62286 UAS-CG16892[RNAi] VDRC101415 KK UAS-Gle1[RNAi] BDRC106646 UAS-Gle1[RNAi] BDRC31641 UAS-Gle1[RNAi] BDRC24918 UAS-Gle1[RNAi] NIG14749R-1 UAS-Gle1[RNAi] NIG14749R-1 UAS-Rae1[RNAi] VDRC29302 GD UAS-Rae1[RNAi] VDRC29303 GD UAS-Rae1[RNAi] BDRC32882 UAS-Rae1[RNAi] BDRC57832 UAS-Rae1[RNAi] NIG9862R-2 UAS-Rae1[RNAi] NIG9862R-3 UAS-Rae1[RNAi] VDRC101338 KK UAS-Nup98-96[RNAi] VDRC109279 KK UAS-Nup98-96[RNAi] NIG10198R-1 UAS-Nup98-96[RNAi] VDRC31198 GD UAS-Nup98-96[RNAi] VDRC31199 GD UAS-Nup153[RNAi] VDRC47155 GD UAS-Nup153[RNAi] BDRC32837 UAS-Nup50[RNAi] VDRC100564 KK UAS-Nup50[RNAi] VDRC20824 GD UAS-Mtor[RNAi] VDRC110218 KK UAS-Mtor[RNAi] VDRC24265 GD UAS-Mtor[RNAi] NIG8274R-2 UAS-Mtor[RNAi] BDRC32941 UAS-Nup58[RNAi] VDRC40773 GD UAS-Nup58[RNAi] BDRC60110 UAS-Nup58[RNAi] VDRC108016 KK UAS-Nup54[RNAi] VDRC42153 GD UAS-Nup54[RNAi] VDRC42154 GD UAS-Nup54[RNAi] VDRC103724 KK UAS-Nup54[RNAi] BDRC57426 UAS-Nup62[RNAi] BDRC35695 UAS-Nup62[RNAi] VDRC44808 GD UAS-Nup62[RNAi] NIG6251R-1 UAS-Nup62[RNAi] NIG6251R-2 UAS-Nup62[RNAi] BDRC52927 UAS-Nup62[RNAi] VDRC44806 GD UAS-Nup62[RNAi] VDRC100588 KK UAS-ran[RNAi] NIG1404R-5 UAS-ran[RNAi] BDRC44587 UAS-ran[RNAi] VDRC104417 KK UAS-ran[RNAi] VDRC24835 GD UAS-ran[RNAi] BDRC42482 UAS-ran[RNAi] BDRC31392 UAS-ranGAP[RNAi] BDRC29565 UAS-ranGAP[RNAi] VDRC30568 GD UAS-ranGAP[RNAi] VDRC108264 KK UAS-Rcc1[RNAi] VDRC110321 KK UAS-Rcc1[RNAi] VDRC38389 GD UAS-Rcc1[RNAi] NIG10480R-2 UAS-Rcc1[RNAi] NIG10480R-1 UAS-RanBP3[RNAi] VDRC104432 KK UAS-RanBP3[RNAi] VDRC38363 GD UAS-RanBP3[RNAi] NIG10225R-4 UAS-emb[RNAi] NIG13387R-1 UAS-emb[RNAi] NIG13387R-4 UAS-emb[RNAi] BDRC31353 UAS-emb[RNAi] VDRC103767 KK UAS-emb[RNAi] BDRC34021 UAS-emb[RNAi] VDRC3347 GD UAS-cse[RNAi] VDRC12648 GD UAS-cse[RNAi] VDRC110215 KK UAS-RanBP21[RNAi] VDRC31706 GD UAS-ebo[RNAi] VDRC34737 GD UAS-ebo[RNAi] BDRC32347 UAS-RanBP16[RNAi] BDRC32347 UAS-RanBP16[RNAi] VDRC34737 GD UAS-cdm[RNAi] BDRC44551 UAS-cdm[RNAi] VDRC40436 GD UAS-Ket[RNAi] VDRC107622 KK UAS-Ket[RNAi] VDRC22348 GD UAS-Ket[RNAi] BDRC44576 UAS-Ket[RNAi] NIG2637R-3 UAS-Ket[RNAi] BDRC31242 UAS-Ket[RNAi] BDRC41845 UAS-Tnpo[RNAi] VDRC105181 KK UAS-Tnpo[RNAi] VDRC30066 GD UAS-Tnpo[RNAi] VDRC6544 GD UAS-Tnpo[RNAi] VDRC6543 GD UAS-Tnpo[RNAi] BDRC27546 UAS-Tnpo[RNAi] BDRC61230 UAS-CG8219[RNAi] VDRC24245 GD UAS-Tnpo-SR[RNAi] BDRC25988 UAS-Tnpo-SR[RNAi] VDRC33571 GD UAS-CG10950[RNAi] VDRC41460 GD UAS-CG10950[RNAi] VDRC41462 GD UAS-CG32164[RNAi] VDRC109183 KK UAS-CG32164[RNAi] NIG32164R-1 UAS-CG32164[RNAi] NIG32164R-2 UAS-CG32164[RNAi] VDRC34422 GD UAS-CG32165[RNAi] VDRC49307 GD UAS-CG32165[RNAi] VDRC49306 GD UAS-CG32165[RNAi] VDRC109561 KK UAS-Karyβ3[RNAi] VDRC105602 KK UAS-msk[RNAi] BDRC33626 UAS-msk[RNAi] BDRC35598 UAS-msk[RNAi] BDRC27572 UAS-msk[RNAi] BDRC34998 UAS-msk[RNAi] VDRC108415 KK UAS-msk[RNAi] VDRC38963 GD UAS-RanBP9[RNAi] VDRC27384 GD UAS-RanBP9[RNAi] VDRC110236 KK UAS-RanBP9[RNAi] BDRC33004 UAS-RanBP9[RNAi] BDRC33005 UAS-RanBPM[RNAi] VDRC45981 GD UAS-RanBP11[RNAi] BDRC55142 UAS-RanBP11[RNAi] VDRC110496 KK UAS-RanBP11[RNAi] VDRC44731 GD UAS-Pen[RNAi] BDRC27692 UAS-Pen[RNAi] NIG4799R-3 UAS-Pen[RNAi] BDRC43142 UAS-Pen[RNAi] NIG4799R-1 UAS-Pen[RNAi] VDRC34265 GD UAS-Pen[RNAi] VDRC34266 GD UAS-Pen[RNAi] VDRC102627 KK UAS-Kap-α3[RNAi] VDRC36104 GD UAS-Kap-α3[RNAi] VDRC36103 GD UAS-Kap-α3[RNAi] VDRC106249 KK UAS-Kap-α3[RNAi] BDRC27535 UAS-Kap-α4[RNAi] VDRC27266 GD UAS-Kap-α4[RNAi] VDRC108143 KK UAS-Kap-α4[RNAi] NIG10478R-1 UAS-Kap-α4[RNAi] NIG10478R-4 UAS-Kap-α1[RNAi] VDRC28921 GD UAS-Kap-α1[RNAi] VDRC28920 GD UAS-Kap-α1[RNAi] NIG8548R-2 UAS-Kap-α1[RNAi] NIG8548R-3 UAS-Kap-α1[RNAi] VDRC108741 KK Mus musculus: Tg(Nes-EGFP)33Enik (Nestin::GFP)Mignone et al, 20045523870Primary adult hippocampal wildtype neural stem cell line no. 5Blomfield et al, 2019Homo sapiens:Umbilical cord blood HSPCsGrey et al, 2020 Antibodies Mouse anti-MiraOhshiro et al, 2000Guineapig anti-DpnShaw et al, 2018Rabbit anti-Phospho-Histone H3Upstate Biotechnology06-570Mouse anti-GFPThermo Fisher ScientificA11120Rabbit anti-GFPThermo Fisher ScientificA6455Chicken anti-GFPThermo Fisher ScientificA10262Chicken anti-GFPAbcamab13970Rabbit anti-Dcp-1Cell Signaling Technology95785Mouse anti-Ki67BD Biosciences550609Rat anti-GfapInvitrogen13-0300Rat anti-Sox2eBioscience14-9811-82Rabbit anti-Mcm2Cell Signaling4007SMouse anti-phosphoSRSigmaMABE50Rabbit anti-Srrm2Thermo Fisher ScientificPA5-66827Anti-Human Lineage CocktailBD Biosciences340546Anti-CD34BD Biosciences8G12Anti-CD38eBioscienceHB7Anti-CD45BiolegendHI30hCD45RAeBioscienceHI100Anti-CD90BD Biosciences5E10Anti-CD49BD BiosciencesGoH3 Oligonucleotides and other sequence-based reagents Cy3-Oligo-dT(50)Gene Link26-4322-02HCR probes for Psap, Eif3a, Map1bFaraway et al, 2025m45S_forward primerCGGTGGTGTGTCGTTCCCm45S_reverse primerGCGTCTCGTCTCGTCTCACTm7SL_forward primerGGAGTTCTGGGCTGTAGTGCm7SL_reverse primerATCAGCACGGGAGTTTTGACmGAPDH_forward primerCCTGACCTGCCGTCTAGAAAmGAPDH_reverse primerCCCTGTTGCTGTAGCCAAATCG14712_forward primerCACGCTTCTAGTTGGCTCGCTCCG14712_reverse primerCTTTGTCGGCTGCTTCTCCTGC Chemicals, enzymes and other chemicals 4% methanol-free formaldehydePolysciences18814DMEM)/F-12 with GlutamaxThermo Fischer Scientific3133109N-2 MAX supplementR&D SystemsAR009Penicillin–streptomycinThermo Fischer Scientific15140LamininSigmaL2020HeparinSigmaH3393-50KURecombinant murine FGF-2PeproTech450-33Recombinant murine EGFPeproTech315-09Recombinant murine BMP4R&D Systems5020-BP-0104% paraformaldehydeThermo Fisher ScientificJ61899.AKFicoll-PaqueGE HealthcareGE17-1440-02poly-L-lysine hydrobromideSigmaP1524Silicone rubber compoundRS692-542Vectashield + DAPIVector LaboratoriesH-2000Vectashield - DAPIVector LaboratoriesH-1000DAPISigmaD9542Aqua-polymountPolysciences1860620DEPC-treated waterThermo Fisher ScientificAM9906yeast tRNAThermo Fisher ScientificAM7119Bovine Serum AlbuminGibco15260-037Dextran SulphateMillipore3407181deionized FormamideSevern Biotech LTD30-63-05Pierce IP Lysis bufferThermo Fisher Scientific87788HALT^TM^ Protease inhibitor cocktailThermo Fisher Scientific87786HALT^TM^ Phosphatase inhibitor cocktailThermo Fisher Scientific78420EDTAThermo Fisher Scientific87788TMT 10-plex Isobaric Label ReagentsThermo Fisher Scientific90110Turbo DNaseAmbionAM2238SuperScript IVThermo Fisher Scientific18090200qPCR BIO SyGreen Master MixPCR BiosystemsPB20.16Kapa Dual-Indexed AdaptorsKAPA BiosystemsKK8720 Software Fiji/ImageJ v2.16.0Schindelin et al, 2012Prism v10.1.1GraphPadMaxQuant v1.6.12.0Cox and Mann, 2008Perseus v1.4.0.2Tyanova et al, 2016REVIGO v??Supek et al, 2011Cutadapt v1.9.1Martin, 2011RSEM v1.3.0Li and Dewey, 2011STAR v2.5.2Dobin et al, 2013R v3.6.1R Core Team, 2025DESeq2 v 1.24.0Love et al, 2014dplyr v1.0.6Wickham et al, 2025ggplot2 v3.2.1Wickham, 2016ClusterProfiler v3.12.0Yu et al, 2012 Other EasySep^TM^ Human Progenitor Cell Enrichment KitStemcell Technologies19356Click-iT EdU Imaging KitThermo Fisher ScientificC10340Pierce BCA Protein Assay kitThermoFischer Scientific23225iST-NHS sample preparation kitPreOmicsP.O.00026PARIS kitThermo Fisher ScientificAM1921KAPA mRNA Hyper-Prep KitKAPA BiosystemsKK8581Aria cytometerBD Biosciences510 confocal microscopeZeiss800 confocal microscopeZeiss880 confocal microscopeZeissSP5 confocal microscopeLeicaSP8 confocal microscopeLeicaSoRa spinning disk microscopeNikonOrbitrap Fusion Tribrid mass spectrometerThermo Fisher ScientificLight Cycler®96 Real-Time PCR SystemRoche2200 TapeStation InstrumentAgilent TechnologiesHiseq 4000Illumina

Drosophila melanogaster genetics, rearing and staging

Flies were mutagenized and reared as published (Slack et al, 2006; Sousa-Nunes et al, 2009): Isogenic FRT82B males were fed with 26 mM EMS in 1% (w/v) aqueous solution of sucrose for 18 h, allowed to recover in our standard cornmeal food (8% (w/v) glucose, 2% (w/v) cornmeal, 5% (w/v) baker’s yeast, 0.8% (w/v) agar in water) for a day, then mated with w;;TM3,Sb/TM6B,Tb virgin females. Stocks were established carrying mutations balanced over TM6B. Mosaic analysis with a repressible cell marker (MARCM) clones (Lee and Luo, 1999) were induced as described (Sousa-Nunes et al, 2009): The MARCM stock y,w,hs-FLP[1.22]; tub-GAL4,UAS-NLS::GFP::6xmyc; FRT82B,tubP-GAL80[LL3]/(TM6B) (gift from G. Struhl) was crossed to the 2V327 mutant (carrying FRT82B) (day 0) and females left to lay for 24 h at 25 °C. Lay tubes were heat-shocked for 1.5 h at 37 °C on both days 2 and 3 to induce mitotic recombination events. Wandering larvae of the desired genotype (hs-FLP[1.22]/+; tub-GAL4, UAS-NLS::GFP::6xmyc/+; FRT82B,tubP-GAL80[LL3]/ FRT82B,2V327) were collected on days 6 and 7, and their CNSs dissected and processed for immunofluorescence. RNAi was performed in conjunction with UAS-Dcr2 (Dietzl et al, 2007) and stocks (identifiers indicated in figure panels and/or legends as well as accompanying Reagents and Tools Table) were obtained from the Transgenic RNAi Project at Harvard Medical School, Vienna Drosophila Resource Centre (VDRC), the Japanese National Institute of Genetics (NIG) Fly, the Department of Drosophila Genomics and Genetic Resources (DGRC) at the Kyoto Stock Centre, and the Bloomington Drosophila Resource Centre (BDRC); deficiency (Parks et al, 2004; Ryder et al, 2004) and balancer stocks, UAS-mCD8::GFP, UAS-Dcr2 and w;;tub-GAL80[ts] were also obtained from the BDRC. The following additional strains were used: w^1118^ as “WT” control; grh-GAL4 (Chell and Brand, 2010) and UAS-mira::3xGFP (Sousa-Nunes et al, 2009) to visualise a subset of NSCs; NP3537-GAL4 (NIG), a GAL4 insertion into the nab gene locus, for pan-NSC RNAi. For larval genotyping, lethal chromosomes were reestablished over balancer chromosomes marked by Dfd-YFP. For strict larval staging experiments, crosses were performed in cages with grape juice plates (25% (v/v) grape juice, 1.25% (w/v) sucrose, 2.5% (w/v) agar) supplemented with live yeast paste. Larvae hatched within 2 h at 25 °C were transferred to our standard cornmeal food and placed at the desired temperature. Early larvae were newly hatched, and late larvae were at the wandering stage. Data from males or females of the same genotype were pooled without distinction.

Drosophila tissue collection

Larval CNSs were dissected in 100 mM Na_2_HPO_4_/NaH_2_PO_4_ (PBS) freshly before all stainings. Late larval CNSs were stained in microfuge tubes, whereas early larval CNSs were L1 and L2 tissues were immobilised on poly-L-lysine-coated slides (prepared in-house) instead; all were fixed in 4% methanol-free formaldehyde (Polysciences 18814) in PBS for 15–20 min at room temperature (RT; ~22 °C).

Mouse genotypes and rearing

WT/RYFP mice were used for the derivation of primary NSC cultures. Nestin-GFP mice (Mignone et al, 2004) (*n *= 4) were used for in vivo NSC analysis. Mice were housed in standard cages under a 12 h light/dark cycle, with ad libitum access to food and water. All care and procedures were performed in accordance with the guidelines of the Francis Crick Institute, national guidelines, and laws (UK Home Office project licence PPL PB04755CC).

Mouse tissue/cell collection and cell culture

Primary NSCs were prepared as published (Blomfield et al, 2019): 7–8 week old mice were sacrificed, and dentate gyri dissected. Cultures were amplified as neurospheres for two passages before dissociation to adherent cultures of NSCs. NSCs were then propagated in Dulbecco’s modified Eagle’s medium (DMEM)/F-12 with Glutamax (Thermo Fischer Scientific 31331093) + N-2 MAX supplement (R&D Systems AR009) + penicillin–streptomycin (Thermo Fischer Scientific 15140) + 2 mg/ml laminin (Sigma L2020) + 5 mg/ml heparin (Sigma H3393-50KU) + 20 ng/ml recombinant murine fibroblast growth factor 2 (FGF-2) (PeproTech 450-33) + 20 ng/ml recombinant murine epidermal growth factor (EGF) (PeproTech 315-09) in tissue culture incubators at 37 °C with 5% CO_2_. qNSC induction was performed with 50 ng/ml BMP4 (R&D Systems 5020-BP-010) and qNSCs cultures were never passaged. Mouse cells were fixed in 4% paraformaldehyde (PFA) (Thermo Fisher Scientific J61899.AK) in PBS for 10 min at RT, washed in PBS, then stored 0.02% sodium azide in PBS at 4 °C. For mouse brain fixation, mice were transcardially perfused with PBS for 3 min, followed by 4% PFA in PBS for 12 min, brains were post-fixed for 2 h in 4% PFA at 4 °C, washed in PBS, then stored 0.02% sodium azide in PBS at 4 °C. Brains were coronally sectioned at a thickness of 40 mm using a vibratome (Leica) with the entire rostral-caudal extent of the hippocampus collected. Cells to be stained were plated onto glass coverslips in proliferation medium.

Human tissue/cell collection, and cell culture

Umbilical cord blood HSPCs were extracted and processed as published (Grey et al, 2020): Umbilical cord blood was collected from full-term donors at the Royal London Hospital following ethical board approval (by the East London Research Ethical Committee (REC) reference number 06/Q0604/110) and informed consent. Mononuclear cells were isolated by density centrifugation using Ficoll-Paque (GE Healthcare). Cells were depleted for lineage markers by using an EasySep^TM^ Human Progenitor Cell Enrichment Kit (Stemcell Technologies) according to the manufacturer's instructions. Immunophenotypic HSCs were defined as Lin^-^/CD34^+^/CD38^-^/CD45RA^-^/CD90^+^/CD49f^+^, where Lin is Human Lineage Cocktail 1 (CD3, CD14, CD16, CD19, CD20, CD56), and sorted by an Aria cytometer (BD Biosciences). For flow cytometry of HSPCs, cells were initially identified based on forward and side scatters. Dead cells were excluded based on staining with 4’,6-diamidino-2-phenylindole (DAPI).

Preparation of poly-L-lysine-coated slides

In total, 5.9 mg/ml in H_2_O poly-L-lysine hydrobromide (Sigma P1524) stock solution (stored at −20 °C) was diluted 3/20 in H_2_O. To this, Kodak Photo-Flo 200 wetting agent was added 1.5:200, making up a working solution. Slides mounted in racks were dipped four times into poly-L-lysine working solution for 10 min each at RT interspersed each time by drying for 10 min at 60 °C. A staining well was created on coated slide by dispensing silicone rubber compound (RS 692-542) and slides were then stored at 4 °C until used.

EdU labelling

Tissue or cells were incubated in 10 mM EdU (Thermo Fisher Scientific, C10340) in medium: Drosophila CNSs for 2 h in Schneider’s Insect Medium; mouse NSCs in media described above for 6 h to assess relative quiescence depth following different durations of BMP-4 exposure. EdU detection reaction was performed according to the manufacturer's instructions (Thermo Fisher Scientific Click-iT EdU Imaging Kit). For combined EdU and immunofluorescence, the Click reaction was performed after secondary antibody incubation.

Immunofluorescence

Immunofluorescence was performed as published (Sousa-Nunes et al, 2009; Sousa-Nunes et al, 2011; Blomfield et al, 2019): Fixed tissues or cells were washed in PBT-0.1% (PBS + 0.1% Triton-X-100) for ~2 × 30 min, blocked, incubated with primary antibodies overnight at 4 °C, washed ~3 × 20 min in PBT, incubated with secondary antibodies for ~2 h for at RT, washed 3 × 20 min in PBT and mounted. Antibodies were diluted in PBT with 5% goat serum (for Drosophila tissues) or with 1% donkey serum (for mouse cells or tissue), except when combining immunofluorescence with fluorescent in situ hybridisation (FISH)—see below. Drosophila CNSs were blocked in 5% normal goat serum in PBT for at least 30 min and mounted in Vectashield with or without DAPI (Vector Laboratories H-2000 or H-1000). Mouse tissue was blocked in 10% normal donkey serum in PBT-1% for 2 h; mouse or human cells were blocked in 10% normal donkey serum in PBT-0.1% for 1 h; DAPI (Sigma D9542) was used at 1 μg/ml in 1:1 PBS:H_2_O for 30 min at RT; and mounting medium was Aqua-polymount (Polysciences 1860620). Primary antibodies were: mouse anti-Mira (Ohshiro et al, 2000) (supernatant from monoclonal antibody producing cells diluted to 1/50-1/5); guineapig anti-Dpn (Shaw et al, 2018) (diluted 1/5000); rabbit anti-Phospho-Histone H3 (Upstate Biotechnology 06-570, diluted 1/400); mouse, rabbit, and chicken anti-GFP (Thermo Fisher Scientific A11120, A6455, A10262 and A11120, diluted 1/1000), chicken anti-GFP (Abcam ab13970, diluted 1/1000); rabbit anti-Dcp-1 (Cell Signaling Technology 95785, diluted 1/100), mouse anti-Ki67 (BD Biosciences 550609, diluted 1/100), rat anti-Gfap (Invitrogen 13-0300; dilution 1/500); rat anti-Sox2 (eBioscience 14-9811-82 1/400); rabbit anti-Mcm2 (Cell Signaling 4007S, diluted 1/400); mouse anti-phosphoSR (Sigma MABE50, diluted 1/500); rabbit anti-Srrm2 (ThermoFisher PA5-66827, diluted 1/1000). Antibodies used for fluorescence-activated cell sorting of hematopoietic cells were: anti-Human Lineage Cocktail (BD Biosciences 340546), anti-CD34 (BD Biosciences 8G12), anti-CD38 (eBioscience HB7), anti-CD45 (Biolegend HI30), hCD45RA (eBioscience HI100), anti-CD90 (BD Biosciences 5E10), and anti-CD49f (BD Biosciences GoH3). Secondary antibodies were conjugated to either Alexa-Fluor-488, Alexa-Fluor-555, or Alexa-Fluor-647 (Molecular Probes) and used at 1/500.

FISH

Fixed tissues or cells were permeabilised in cold methanol for 10 min, rehydrated in 70% ethanol for at least 10 min followed by 1 M Tris-hydroxymethyl-aminomethane (Tris) buffer (pH 8.0) for 5 min. All aqueous solutions were prepared with diethyl pyrocarbonate (DEPC)-treated water (Thermo Fisher Scientific AM9906). For the oligo(dT) probe, hybridisation was performed with 1 ng/μl Cy3-Oligo-dT(50) (Gene Link, 26-4322-02) in 2× saline-sodium citrate buffer (SSC) (pH 7.0) containing 1 mg/ml yeast tRNA (Thermo Fisher Scientific AM7119), 0.005% (v/v) Bovine Serum Albumin (Gibco 15260-037), 10% (v/v) Dextran Sulphate (Millipore 3407181) and 25% (v/v) deionized Formamide (Severn Biotech LTD 30-63-05) for at least 2 h at 37 °C in a humidified chamber. Samples were washed once in 4× SSC and twice in 2× SSC. For oligo(dT) FISH of mouse tissue, 40 µm coronal sections first underwent heat-mediated antigen retrieval at 95 °C for 10 min in 10 mM SSC (pH 6.0). For specific mRNA probes, hybridisation chain reaction (HCR) FISH was performed as previously described (Faraway et al, 2025). For each mRNA of interest, 12 probe pairs were designed with the B4 amplifier sequences, using the HCR 3.0 Probe Maker (Kuehn et al, 2022). Cells were plated at a density of 12,500 cells per well in a 96-well plate, and HCR 3.0 was performed as described in the original protocol (Choi et al, 2018) with the following modifications. Briefly, cells were fixed in a solution of 4% PFA, 0.4% glyoxal and 0.1% methanol for 10 min, then permeabilised for 5 min in 0.5% Triton-X in PBS. Following pre-hybridisation, primary probe hybridisation was performed in a volume of 50 µl per well, with a probe concentration of 10 nM for 8 h. The 16 h (overnight) amplification step was carried out in a volume of 50 µl, with half the concentration of HCR hairpin amplifiers with 647 fluorophore (Molecular Instruments). For combined FISH and immunofluorescence, FISH was performed before primary antibody incubation in 2× SSC, PBT-0.1% (with 5% donkey serum for mouse tissue: all other samples: no serum) and subsequent steps in 2× SSC rather than PBS.

Imaging and image analyses

Fluorescence samples were scanned with Zeiss 510, 800 or 880, or Leica SP5 or SP8 scanning confocal microscopes, or with Nikon SoRa spinning disk microscope. Optical section steps ranged from 0.1 to 2 μm with picture size of 1024 × 1024 pixels. Images were processed and arranged using Fiji/ImageJ, Adobe Illustrator, Adobe Photoshop CS5, and/or PowerPoint software. Drosophila cell counts were carried out with the ImageJ Cell Counter plugin. At least ten brains per genotype were stained and, when qNSC fibres were observed ten brain lobes were quantified per genotype (a single lobe per animal), when not observed it was four lobes (a single lobe per animal). Cell culture images were acquired from three random fields from each of three coverslips; their counts, image masking, projecting and reformatting were performed with CellProfiler scripts. Oligo(dT) intensity per area was determined for nuclei (defined by DAPI staining) or cytoplasms from SUM projections when the whole cell could be imaged (cell culture), or from single optical sections of tissue (where dense cell packing and sectioning precluded imaging entire cells); staining variability was thus internally controlled when quantifying nucleocytoplasmic ratio. In tissue, only cells whose cytoplasmic edge could be confidently determined were quantified; no data points were excluded. Where possible, image analysis was performed with blinding. Graphpad Prism software was used for statistics and graphs. Statistical comparison between two groups was done using t test, paired or unpaired, depending on the experimental design. Gaussian distribution was tested using D’Agostino-Pearson omnibus normality test, and, if the condition was not met, the groups were compared using Mann–Whitney test, if unpaired, or Wilcoxon rank test, if paired. Statistical comparison between more than two groups was done using one-way ANOVA testing. Exact P values for all comparisons are reported in the figure legends. The number of samples (n) in the legends refers to biological replicates.

Protein extraction and proteomics

Cells were washed with ice-cold PBS and scraped. Cells were washed with ice-cold PBS, then scraped in Lysis Buffer (ThermoFischer Scientific 87788) + 1× Protease inhibitor cocktail (ThermoFischer Scientific 87786) + 1× EDTA (ThermoFischer Scientific 87788) + 1× Phosphatase inhibitor cocktail (ThermoFischer Scientific 78420). Cells were lysed at 4 °C for 20 min under rotation, centrifuged at 13,000 RPM at 4 °C for 20 min, and the pellet discarded. Protein extract concentrations were determined using Pierce BCA Protein Assay kit (ThermoFischer Scientific 23225) according to the manufacturer instructions. Proteomic analyses were performed on 50 µg total protein per time point. Samples were Acetone-precipitated overnight followed by Trypsin digestion using the PreOmics iST-NHS sample preparation kit (Kulak et al, 2014), labelled using 0.2 mg TMT 10-plex Isobaric Label Reagents (Thermo Scientific) and checked to ensure >99% labelling efficiency. Equal volumes of all ten labelled samples were mixed to produce a single mixed sample which was subject to high pH (HpH) reversed-phase peptide fractionation (Pierce). Nine HpH fractions were each analysed on a 145 min U3000 HPLC method. Samples were loaded in 2% Acetonitrile, 0.05% Trifluoroacetic acid onto a C18 trap column, then transferred onto an EasySpray 50 cm × 75 µm column. Peptides were separated by elution using the following conditions: 15 min 3–9% mobile phase A (0.1% Formic acid, 5% Dimethyl Sulfoxide (DMSO)), 90 min 9–30%, 15 min 30–50%, 5 min 99% and ending with 15 min at 3%. Mobile phase B was 80% Acetonitrile, 5% DMSO, 0.1% Formic acid. An SPS-MS3 method on an Orbitrap Fusion Tribrid mass spectrometer (Thermo Fisher Scientific) acquired data with settings: MS1 orbitrap, resolution 120 K, scan range 375–1500 m/z, maximum injection time 50 ms, AGC target 4E5, normalised AGC target 100%, microscans 1, RF lens 30%, profile data, MIPS mode peptide, charge states 2–6 included, dynamic exclusion 60 s +/− 10 ppm. MS2 ion trap, quadrupole isolation mode, 1.2 isolation window, CID activation, 35% collision energy, activation time 10 ms, activation Q 0.25, turbo scan rate, maximum injection time 50 ms, AGC target 1E4, normalised AGC target 100%, microscans 1, centroid data, filter precursor selection range MSn 400–1200 m/z. MS3 orbitrap scan event 1 for charge state 2, quadrupole isolation mode, 1.3 isolation window, Multi-notch Isolation True, MS2 Isolation Window (m/z) 2, number of Notches 5, activation type HCD, collision energy 65%, orbitrap resolution 50 K, scan range 100–500 m/z, maximum injection time 105 ms, AGC target 1E5, normalised AGC target 200%, microscans 1, centroid data. MS3 orbitrap scan event 2 for charge state 3 as above but with number of Notches 10. MS3 orbitrap scan event 3 for charge states 4–6 as above but with number of Notches 10. Raw data were analysed in MaxQuant (Cox and Mann, 2008) (v1.6.12.0) against a SwissProt Mus musculus protein database containing 17,482 protein entries (downloaded May 2020). TMT 10 plex quantification was selected (modification at Lysine and peptide N-terminal amino groups) along with variable modification of Methionine oxidation and N-terminal acetylation. A fixed Cysteine modification of +113.084 Da (specific to the iST-NHS kit) was added. Further data analyses were performed in Perseus (Tyanova et al, 2016) (v1.4.0.2). Common contaminants and proteins identified from decoy sequences were removed. Protein intensities were log_2_ transformed, median normalised within each sample and then normalised to Day 0. GO terms were simplified using REVIGO (Supek et al, 2011) with allowed similarity of 0.7.

Nuclei acid extraction, cell fractionation, sequencing and analyses

Genomic DNA of heterozygous FRT82B/FRT82B,2V327 Drosophila was extracted according to standard methods: Anaesthetised flies were ground in a microfuge tube in Buffer A (100 mM Tris-HCl (pH 7.5), 100 mM EDTA, 100 mM NaCl, 0.5% SDS) with a disposable tissue grinder (Kontes), incubated at 65 °C for 30 min, then twice the volume of LiCl/KAc Solution was added (1:2.5 of 5 M KAc: 1 part 5 M KAc), and incubate on ice for at least 10 min. The suspension was spun for 15 min at room temperature, at 13,000 rpm, and the supernatant was respun to clear any debris. DNA was precipitated with isopropanol and washed with 70% ethanol. The identity of the genetic lesion responsible for the phenotype was determined by sequencing of exons of CG14712, outsourced to Eurofins Genomics using primers designed with A Plasmid Editor software. Mouse cell nucleocytoplasmic fractionation and RNA isolation were performed with the PARIS kit (ThermoFisher AM1921) according to the manufacturer's instructions plus an additional ethanol precipitation step, and rehydrated in nuclease-free water. Fractionation quality was verified by quantitative reverse-transcriptase polymerase chain reaction (qRT-PCR) as published (Yap et al, 2018): Total RNAs were isolated from cells using TRIzol, as recommended, with an additional acidic phenol-chloroform (1:1) extraction step. The aqueous phase was precipitated with an equal volume of isopropanol, washed with 70% ethanol and rehydrated in 80 μl of DEPC-treated water. RNA samples were then treated with 4–6 units of Turbo DNase (Ambion AM2238) at 37 °C for 30 min to remove traces of genomic DNA, extracted with equal volume of acidic phenol-chloroform (1:1), precipitated with 3 volumes of 100% ethanol and 0.1 volume of 3 M sodium acetate (pH 5.2), washed with 70% ethanol and resuspended in DEPC-treated water. Reverse transcription was performed using SuperScript IV (ThermoFischer Scientific 18090200) and random decamer primers (Integrated DNA Technologies N10) at 50 °C for 40 min. cDNA samples were analysed by quantitative polymerase chain reaction (qPCR), which was carried out using a Light Cycler®96 Real-Time PCR System (Roche) and qPCR BIO SyGreen Master Mix (PCR Biosystems PB20.16). qPCR signals for nuclear or cytoplasmic control mRNAs (m45S and m7SL, respectively) were normalised to GAPDH mRNA expression levels in the respective fraction. The following primer pairs were used: m45S_F: CGGTGGTGTGTCGTTCCC and m45S_R: GCGTCTCGTCTCGTCTCACT; m7SL_F: GGAGTTCTGGGCTGTAGTGC and m7SL_R: ATCAGCACGGGAGTTTTGAC; mGAPDH_F: CCTGACCTGCCGTCTAGAAA and mGAPDH_R: CCCTGTTGCTGTAGCCAAAT. Next-generation sequencing was performed essentially as published (Blomfield et al, 2019): RNA concentration was quantified using the Qubit dsDNA BR/HS Assay Kit. A KAPA mRNA Hyper-Prep Kit (for Illumina) (KAPA Biosystems) was used with 200 ng of RNA + 1% spike-in mix (from the External RNA Controls Consortium) diluted to a final volume of 50 μl. Poly(A) RNA was purified with 50 μl of capture beads at 65 °C for 2 min and 20 °C for 5 min. RNA was eluted from the beads in 50 μl of RNase-free water by incubating at 70 °C for 2 min and 20 °C for 5 min. A second purification of poly(A) RNA was performed by adding 50 μl of bead binding buffer, and the sample was incubated for 5 min at 20 °C. Purified poly(A) RNA was subjected to the KAPA Hyper-Prep assay: end-repair, A-tailing, and ligation by adding 20 μl of Fragment, Prime and Elite Buffer (2×). To obtain a distribution of 200–300 bp fragment on the library, the reaction was run for 6 min at 94 °C. cDNA synthesis was run in two steps following the manufacturer's instructions. The ligation step consisted of a final volume of 110 μl of the adaptor ligation reaction mixture with 60 ml of input cDNA, 5 μl of diluted adaptor and 45 μl of ligation mix (50 μl of ligation buffer + 10 μl of DNA ligase). The Kapa Dual-Indexed Adaptors (KAPA Biosystems KK8720) stock was diluted to 1.5 μM to get the best adaptor concentration for library construction. The ligation reaction was run for 15 min at 20 °C. To remove short fragments such as adaptor dimers, two SPRISelect bead cleanups were done (0.63× SPRI and 0.7× SPRI). To amplify the library, 30 μl of Kapa HiFi HotStart PCR master mix was added to 20 μl cDNA and 12 cycles were performed. Amplified libraries were purified via a SPRISelect 1× cleanup. The quality and fragment size distributions of purified libraries were assessed by a 2200 TapeStation Instrument (Agilent Technologies). Libraries were sequenced with HiSeq 4000 (Illumina) with 100 bp single-end reads with a depth of 50 million reads per sample. Raw reads were quality and adaptor trimmed using cutadapt-1.9.1 software (Martin, 2011) then aligned and quantified using RSEM-1.3.0/STAR-2.5.2 (Li and Dewey, 2011; Dobin et al, 2013) against the mouse genome GRCm38 and annotation release 89, both from Ensembl. Differential gene expression analysis was performed in R-3.6.1 (R Core Team, 2025) using the DESeq2 package (Love et al, 2014) (version 1.24.0). Normalisation and variance-stabilising transformation were applied to raw counts before performing PCA and Euclidean distance-based clustering. Significantly differential genes were always selected using a 0.05 false-discovery rate threshold. Size factors in DESeq2 were calculated based on the summed cytoplasmic and nuclear counts for each paired sample set to reconcile technical differences between samples, whilst count differences between subcellular compartments within the same paired sample set were maintained for all genes. We performed pairwise comparisons between subcellular compartments on each day (within-day comparison) and between the subcellular distribution across days (between days) using the following formula: days_in_BMP + subcellular_compartment + days_in_BMP:subcellular_compartment + days_in_BMP:pair_within_day. Additionally, we performed a likelihood ratio test in DESeq2 to identify the genes that changed subcellular distribution across time (reduced design formula in DESeq2: days_in_BMP + subcellular_compartment + days_in_BMP:pair_within_day). Z-scores of protein-coding genes (from fracRNA-seq data) were correlated to respective protein expression (proteomic data) in R-3.6.1 (R Core Team, 2025) using the dplyr package (version 1.0.6). Scatterplots were prepared using ggplot2 (version 3.2.1) to visualise IR as one intron per gene, selecting the one with the biggest absolute change in IR value. Violin plots were prepared using ggplot2 (version 3.2.1) using the log2 fold changes for all genes after pairwise comparisons between days. Gene set enrichment analyses were performed using ClusterProfiler (Yu et al, 2012) (version 3.12.0) using the enrichGO() function for enrichment of biological process gene ontology terms. Multivalent transcript classification was as reported (Faraway et al, 2025): Multivalency scores and clusters were calculated using the GeRM package, DBSCAN and OPTICS, with all parameters exactly as described. For this manuscript, multivalency values were calculated using the longest protein-coding isoform per gene from the mouse transcriptome according to GENCODE vM22. Summed multivalency scores per cluster were also calculated as described.

Supplementary information

Peer Review File Dataset EV1 Dataset EV2 Dataset EV3 Dataset EV4 Dataset EV5 Dataset EV6 Source data Fig. 1 Source data Fig. 2 Source data Fig. 3 Source data Fig. 4 Source data Fig. 5 Figure EV2 Source Data Figure EV3 Source Data Figure EV4 Source Data Figure EV5 Source Data Figure EV6 Source Data Figure EV7 Source Data Expanded View Figures