Convergent nuclear proteostasis alterations across 40S ribosomal protein deficiencies

Abstract

Defects in ribosome biogenesis cause ribosomopathies, but the pathogenic role of malformed preribosomes remains unclear. Using preribosome fractionation, live-cell and confocal microscopy, and automated imaging screens, we tracked abortive preribosomes following depletion of the 33 small-subunit ribosomal proteins (RPSs). We found that all RPS deficiencies lead to accumulation of nucleolar preribosome materials in the nucleoplasm, regardless of the affected 40S subunit maturation stage. These aberrant particles appear as dispersed complexes or persistent aggregates arising during nucleolar reassembly in late mitosis. Although nucleolar morphology remains largely intact, their accumulation impairs recycling of nucleolar factors and disrupts nucleolar–nucleoplasmic proteostasis. Distinct RPS deficiencies produce different preribosome behaviors, indicating maturation-specific properties. Notably, depletion of RPS19, the protein encoded by the most frequently mutated gene in Diamond–Blackfan anemia, causes particularly severe defects. Our findings identify nucleoplasmic preribosome aggregation and dispersion as common consequences of RPS deficits and potential contributors to 40S ribosomopathy pathogenesis.

Defects in 40S ribosomal proteins disrupt ribosome biogenesis and underlie ribosomopathies. Here, the authors show that ribosomal protein deficiencies cause aberrant preribosomes to accumulate in the nucleoplasm, impairing nuclear protein homeostasis and potentially contributing to diseases such as Diamond-Blackfan anemia.

Introduction

The nucleolus is the site of assembly of both the small (40S) and large (60S) ribosomal subunits. The former one is composed of a ribosomal rRNA (rRNA) (18S) and 33 ribosomal proteins (RPSs), whereas the latter one harbors three ribosomal rRNAs (28S, 5.8S, and 5S) and 47 ribosomal proteins (RPLs). The nucleolus emerges from a coalescence of rRNA gene clusters that are being actively transcribed by RNA polymerase I to generate the 47S pre-rRNA, a precursor that contains three rRNAs (18S, 5.8S, and 28S) separated by spacer regions¹. The synthesis of this precursor drives the formation of a bipartite particle that contains the primordial 40S and 60S subunits. After cleavage and initial processing of the 47S pre-rRNA, the bipartite particle dissociates to generate the pre-40S and pre-60S complexes that, subsequently, mature as they transit from the most internal to the most external regions of the nucleolus, to the nucleoplasm, and to the cytosol^2–5. This maturation process requires >300 ribosome biogenesis factors (RBFs) that help in the processing, folding or modification of the maturing rRNAs as well as in the assembly of the  ribosomal proteins of each subunit^3,4,6–9. Most RBFs play roles inside the nucleoli, and only a minority of them travel with either the 40S or the 60S preribosomes to the nucleoplasm or the cytosol^4,8,9. For instance, the SSU (Small-Subunit) processome, the initial independent nucleolar precursor of the 40S subunit, is a massive assembly that contains the 30S pre-rRNA, 20 RPSs, the U3 snoRNA and 49 RBFs¹⁰ (see image of representative pre-A1 state in Fig. 1A). This particle undergoes an intricate maturation process that involves pre-rRNA processing, folding, and structural compaction to generate pre-40S intermediates that resemble the mature 40S subunit. These compacted complexes are equipped with only the few RBFs required for nuclear export and final maturation, including the recruitment of late-incorporating RPSs in the cytoplasm^11–13 (see images of two pre-40S intermediate states in Fig. 1A).

Fig. 1. Schemes of 40S ribosomal subunit maturation and experimental strategy.

A Structural progression of 40S ribosomal subunit maturation from the SSU processome (pre-A1 state) to the mature 40S subunit. The SSU processome contains the 30S pre-rRNA, 20 ribosomal proteins (RPSs), U3 snoRNA and 49 ribosome biogenesis factors (RBFs, 44 different proteins). Maturation involves pre-rRNA processing, 5′-ETS degradation, recruitment of additional RPSs, and progressive release of RBFs, generating intermediate pre-40S particles in the nucleolus and cytoplasm. Images correspond to published structures^10,11,13,70 (PDB IDs: 7MQ8, pre-A1; 7WTV, state A2; 6G51, state D; 7R4X, mature 40S). B Schematic representation of the nucleolar fibrillar center (FC), dense fibrillar component (DFC) and granular component (GC), showing major 40S preribosomal intermediates. The bipartite precursor assembled on the 47S pre-rRNA contains both pre-40S and pre-60S components. The SSU processome is the earliest-characterized independent 40S precursor and gives rise to successive intermediates located in the nucleolus (21S processome, No1 and No2 particles), nucleoplasm (Nuc) and cytoplasm (Cyt). The distribution of the RBFs bystin, NOC4L, RRP12, NOB1 and LTV1 across these intermediates is indicated. The actual locations of the RBFs in the structures of successive pre-40S precursors are shown in Supplementary Fig. 1. The common nucleolar marker fibrillarin (FBL) is present in the earliest preribosomes containing the 47S or 30S pre-rRNA, but not in the other downstream precursors. DDX21, one RBF analyzed in a specific experiment of this study, is present in early pre-40S complexes. The preribosome sequential extraction (PSE) method separates complexes into SN1, SN2 and SN3 fractions according to their solubilization properties. C Experimental workflow. Cells depleted of RPS19, RPS17 or RPS26 by siRNA transfection were analyzed by northern blotting, PSE/GFP-Trap/western blotting, fluorescence microscopy, live-cell imaging and RNA FISH to identify and localize abortive pre-40S particles. All GFP-tagged RBFs were expressed from their endogenous loci under their cognate promoters and were previously shown to be functional⁴³.

The nucleolus is a unique organelle that forms around actively transcribed genomic loci and exhibits distinctive biophysical properties, giving rise to three concentrically organized subregions: the fibrillar center (FC), the dense fibrillar component (DFC), and the granular component (GC)^1,14,15. The synthesis of the 47S pre-rRNA and the initial steps of ribosome subunit assembly occur at the FC–DFC boundary, while the early maturation of pre-40S and pre-60S intermediates takes place within the GC^14,16. These three subcompartments are immiscible due to their different viscosities, which cause liquid-liquid phase separation^15,17. The movement of preribosomes in and out of the GC is proposed to be driven by interactions of the pre-rRNAs with RBFs containing intrinsically disordered regions, as well as by the radial movement of pre-rRNAs. These RNAs form a gel-like structure and become progressively more fluid as the ribosomal subunits mature and compact¹⁸. Nucleolar subcompartmentalization is primarily viewed as a mechanism to maximize the efficacy and fidelity of all ribosome synthesis steps. However, its phase-separation properties are also used by cells for ribosome-synthesis unrelated functions, such as the temporary sequestration of misfolded proteins to prevent their irreversible aggregation¹⁹, as well as the establishment of repressive chromatin regions²⁰. Like other nuclear condensates, the nucleolus disassembles and reassembles during each cell cycle. It breaks down at the onset of mitosis, dispersing preribosomes and other nucleolar components into the nucleoplasm. Reassembly occurs at the end of mitosis and involves the formation of prenucleolar bodies (PNBs), which are intermediate condensates that incorporate the dispersed preribosomes released during nucleolar breakdown^21–23. Given its central role in cellular homeostasis, eukaryotic cells have developed mechanisms to arrest proliferation when nucleolar assembly, nucleolar integrity, or ribosome production are compromised. One of the main known responses to this type of stress is the nucleolar surveillance pathway, which stabilizes and activates the tumor suppressor protein p53²⁴. p53 activation is triggered not only by major structural disruptions of the nucleolus but also by functional impairments that leave the nucleolar morphology mostly intact, for instance, when maturation of one ribosomal subunit is defective while the production of the other subunit remains unaffected^25,26.

Nucleolar dysfunctions lead to several diseases. One group, which includes both neurodegenerative and rare disorders, is driven by the aberrant expression of highly charged peptides or proteins that interfere with nucleolar structure and/or function^27–31. A second group encompasses several ribosomopathies caused by mutations in genes encoding ribosomal proteins or factors required for ribosome biogenesis^32–35. The most studied ribosomopathy is Diamond-Blackfan anemia (DBA), a disorder characterized by both erythroid and skeletal defects³⁶. DBA is caused by heterozygous loss-of-function mutations in at least 20 ribosomal protein-encoding genes, with mutations in RPS19 accounting for approximately 25% of cases^36,37. Each DBA-causing mutation leads to the haploinsufficiency of one ribosomal protein, causing a partial block in the ribosome maturation step in which the affected protein is involved^38–41. Given the diversity of DBA-associated mutations (involving different RPS and RPL genes), the disease can arise after the generation of different types of abortive preribosomes, including immature or near-mature 40S or 60S subunit precursors, which can be formed in the nucleolus, nucleoplasm, or cytoplasm^{26,38–40,42}. As a result, abortive preribosomes have been largely overlooked as a unifying pathogenic mechanism or as a possible explanation for the high prevalence of RPS19 mutations. The biochemical features, subcellular dynamics, and extent of the accumulation of aberrant preribosomes in the context of ribosomal protein deficiencies also remain poorly understood. For instance, RPSs are commonly classified in a reductionist manner into four major groups based on their roles at distinct stages of 40S subunit synthesis: intranucleolar early-assembly RPSs; intranucleolar early-maturation RPSs; nuclear export RPSs; and RPSs involved in late-maturation cytoplasmic steps. This classification is primarily based on two main features: the types of pre-rRNA intermediates that accumulate and the changes in subcellular localization of certain RBFs upon individual protein depletion conditions^26,40. However, the nature of the abortive preribosomes generated under each of those conditions remains poorly characterized. The indirect impact of those aberrant preribosomes also remains poorly characterized (e.g., effects derived from the sequestration of nucleolar RBFs within the preribosomes stalled at specific biosynthetic steps). A comprehensive understanding of those issues requires a comparative examination of the abortive preribosomes and breakdown-derived subcomplexes arising from direct and indirect 40S synthesis defects for each RPS deficiency.

In this work, we have employed a combination of differential preribosome extraction, confocal and time-lapse microscopy, and siRNA screens coupled to high-throughput automated imaging analyses to investigate the effects of individual depletion of 40S RPSs and RBFs on preribosome biogenesis and RBF recycling. This approach revealed defined patterns of abnormal preribosome dynamics that likely reflect differences in both the stability and breakdown behavior of the abortive preribosomes formed in each case. It also revealed that all RPS deficiencies induce large accumulations of aberrant preribosomal complexes in the nucleoplasm, a phenotype that remains undetectable by standard immunofluorescence with conventional nucleolar markers. Notably, our work revealed that RPS deficiencies exhibit a mixed phenotype, characterized by aggregation-prone preribosomal materials with altered biophysical properties that coexist with free subcomplexes remaining dispersed throughout the nucleoplasm. This phenotype is particularly conspicuous in the case of the depletion of RPS19, the protein encoded by the most frequently mutated gene in DBA. Together, our findings provide a comprehensive view of how 40S ribosomal subunit maturation defects alter proteostasis between the nucleolus and nucleoplasm. They also provide a more accurate functional classification of RPSs and pave the way toward a better mechanistic understanding of the early pathogenic events in DBA. This work also unveils a manifestation of nucleolar stress that has, to our knowledge, remained elusive in studies of nucleolar dysfunction conducted to date: the dispersion of large masses of nucleolar material without major disruption of nucleolar integrity.

Results

Strategy for the analysis of preribosomal complexes in RPS-deficient cells

To analyze the effects of 40S ribosomal protein deficiencies on preribosomal dynamics, we relied on new cellular and analytical tools. In the former case, we utilized HeLa cell derivatives in which the endogenous loci for NOC4L, bystin (also known as ENP1), RRP12, or LTV1 were gene edited individually using the CRISPR–Cas9 technique to incorporate sequences encoding the green fluorescence protein (GFP). To maintain functionality, the GFP-encoding sequences were incorporated at either the 5’ or 3’ end of the coding sequence based on prior functional studies conducted in our lab⁴³. The four HeLa cell derivatives carry the edited alleles in homozygosity and, therefore, all the proteins they produce are GFP-tagged. We chose bystin and RRP12 for our study because they are present in early nucleolar 40S precursors (also known as SSU processomes) and remain bound to downstream intermediate precursors along maturation steps that take place in the nucleolus, nucleoplasm, and cytoplasm^5,10–12 (Fig. 1B, see also the localization of the RBFs in the structures of 40S subunit precursors in Supplementary Fig. 1). We also selected NOC4L because it is exclusively associated with nucleolar precursors^5,10–12 (Fig. 1B and Supplementary Fig. 1). It is also worth noting that bystin, RRP12 and NOC4L become incorporated at the same step into the SSU processome, at a time when this large 40S precursor has been already structured in the almost-final configuration of the pre-A1 state. In such step, the SSU processome has already incorporated all early-assembly RPSs and underwent a large-scale conformational remodeling in which the different 18S domains become folded¹⁰ (Supplementary Fig. 1). Finally, we chose LTV1 because it is recruited before nucleolar exit of 40S preribosomes and, subsequently, is released during the final cytoplasmic maturation steps^5,10–12 (Fig. 1B and Supplementary Fig. 1). The use of these HeLa cell derivatives was compatible with: (i) tracing the formation of all known 40S subunit precursors by performing coimmunoprecipitations with the GFP-Trap reagent; (ii) the study of the localization of the interrogated proteins using epifluorescence microscopy, thus avoiding the variability associated with the use of antibodies of different target affinities; and (iii) avoiding the known misslocalization and functional problems associated with the use of ectopically expressed RBFs that, in many cases, lead to impaired ribogenesis. The full functionality (wild-type-like localization, full incorporation into preribosomes, normal pre-rRNA processing and normal production of 40S subunits) of these four GFP-tagged proteins in the HeLa cell derivatives was demonstrated in a previous study from our laboratory⁴³.

To investigate the fate of 40S preribosomal complexes upon depletion of RPSs whose haploinsufficiency causes DBA in the foregoing cell derivatives, we resorted to a small interference RNA (siRNA) approach to reduce the amount of RPS19 (also designated as eS19) (Supplementary Fig. 2A, top panel; si-RPS19-2), RPS17 (also designated as eS17) (Supplementary Fig. 2B, top panel) or RPS26 (also designated eS26) (Supplementary Fig. 2C, top panel). These three proteins are encoded by genes that are found mutated at high (RPS19, 25% of cases), moderate (RPS26, 7% of cases) and low (RPS17, 1% of cases) frequency in DBA patients^36,37. Previous studies have shown that these deficiencies block the synthesis of 40S ribosomal subunits at specific maturation steps^{26,38,40,44,45}. Based on subsequent characterization of knockdown cell phenotypes, we estimated that the siRNAs were incorporated into 85 to 95% of all transfected cells. This ratio ensures a low degree of cell-to-cell heterogeneity in our phenotypic scores. Experiments were then conducted 24 h after the transfection step, the minimal time point required to induce significant defects in 40S synthesis (see next sections), but early enough to avoid major defects in mature 40S ribosomal subunit levels according to 28S/18S rRNA ratio analyses (Supplementary Fig. 2A–D).

Control and knockdown cells were then analyzed using four methods. Firstly, we utilized a preribosome sequential extraction (PSE) protocol⁴⁶ that allows the isolation of preribosomal complexes in three fractions based on their differential solubilization properties. This method, when applied to normally growing wild-type cells, allows the purification of preribosomes in quantities compatible with both RBF coimmunoprecipitation assays and mass spectrometry analyses⁴³. We have previously validated this method, showing by coimmunoprecipitation, mass spectrometry, sucrose gradient sedimentation, and northern blot analyses that it fractionates: (i) early and intermediate nucleolar preribosomal particles (the SSU processomes containing the 30S or 21S pre-rRNAs as well as the pre-40S intermediates containing the 18S-E pre-rRNA at the No1 stage) in the so-called fraction SN3 (Fig. 1B); (ii) late-nucleolar particles (those containing the 18S-E pre-rRNA at the No2 stage) in the SN2 fraction (Fig. 1B); and (iii) the more-mature precursors (those harboring 18S-E pre-rRNA species) in the SN1 fraction (Fig. 1B)⁴³. Based on that previous evidence, we already knew that NOC4L is present in the 30S-, 21S-, and 18S-E pre-rRNA-containing complexes that fractionate in SN3 (Fig. 1B, early particles upstream of No1, and No1 complexes); that LTV1 and NOB1 are fractionated in the 18S-E pre-rRNA-containing complexes contained in SN1 and SN2 fractions (Fig. 1B, No2, Nuc and Cyt complexes); and that both bystin and RRP12 are in precursors that fractionate in the SN1, SN2, and SN3 fractions (Fig. 1B). In this work, the fractions obtained with this method were incubated with the GFP-Trap reagent to pull down the bystin-GFP bait and analyze the coimmunoprecipitation of RRP12, NOC4L, LTV1 or NOB1 using standard immunoblotting techniques (Fig. 1C, left). We complemented these studies with confocal microscopy analyses to monitor the localization of each of the endogenous GFP-tagged RBFs (Fig. 1C, middle) and, when deemed interesting, with RNA fluorescence in situ hybridization (FISH) experiments with probes for different pre-rRNA species to confirm the identity of misslocalized preribosomes (Fig. 1C, right). Collectively, this approach allowed us to check the impact of the depletion of each of the interrogated RPSs on the extent of accumulation, solubilization properties, and subcellular localization of abortive pre-40S complexes. It also enabled us to investigate the RBF recycling defects found in each case.

A deficit of RPS19 triggers a massive accumulation of nucleolar preribosomal complexes in the nucleoplasm

We first investigated the effect of the knockdown of RPS19 in the system of cells. As previously described^38,40,44,45, we observed that this knockdown promoted a strong accumulation of 21S pre-rRNA and a loss of 18S-E pre-rRNA in siRNA-transfected cells (Supplementary Fig. 3A, B). This is caused by defects in either the assembly or the maturation of post-A1 (21S pre-rRNA-containing) SSU processomes inside the nucleolus (see scheme of 40S subunit maturation in Fig. 1B or Supplementary Fig. 3C). We also found the expected increase in the ratio of 28S rRNA versus 18S rRNA under these conditions, which results from the loss of 40S ribosomal subunits (Supplementary Fig. 3B, left panel, bottom, see 28S/18S ratios). We assumed that the accumulated 21S-containing complexes generated under such conditions would be detectable as a significant increase in the amount of bystin-GFP associated with 21S-containing preribosome components, such as NOC4L and RRP12 in the SN3 fraction of siRPS19-transfected cells when compared to control cells (see scheme of 40S preribosome intermediates in PSE fractions in Supplementary Fig. 3C). However, we found using GFP-Trap coimmunoprecipitation experiments that, although the relative amount of bystin-GFP was slightly more concentrated in the SN3 fraction (Supplementary Fig. 3D, upper panel, compare lanes 4 and 6), the bystin-GFP bound to NOC4L or RRP12 was not substantially increased upon the RPS19 depletion in that fraction (Supplementary Fig. 3D, compare lanes 12 and 14). Furthermore, we did not find any accumulation of preribosomes or breakdown subcomplexes in the SN2 or SN1 fractions from the knockdown cells, a result that would be expected if the abortive 21S-containing preribosomes or their disassembly products with no intact pre-rRNA (referred hereafter as “subparticles”) had abnormal properties that made them soluble in those fractions. In contrast, we observed reduced contents of total bystin-GFP and bystin-GFP associated with LTV1 and NOB1 in the SN2 fraction obtained from the RPS19 knockdown cells when compared to controls (Supplementary Fig. 3D, compare lanes 3 and 5 in upper panels; and lanes 11 and 13 in fourth and fifth panels from top). Similarly, we found that bystin-GFP also displayed reduced binding with RRP12, LTV1, or NOB1 in the SN1 fraction from RPS19 knockdown cells (Supplementary Fig. 3E, compare lanes 6 and 7). We did not detect any major enrichment of those RBFs in the final non-solubilized pellet obtained after the three extraction steps, thus ruling out the possibility that the aberrant preribosomes were insoluble in each of the PSE buffers. All the results obtained in these experiments (summarized in Fig. 2A; RPS19 knockdown column #1) suggest that the abortive 21S-containing preribosomes, unlike the normal ones, lose their integrity when extracted with the PSE protocol. In line with the idea that these preribosomes have abnormal properties, we observed using confocal microscopy that the endogenous NOC4L-GFP, GFP-RRP12, and bystin-GFP were all lost from the internal regions of the nucleolus (Fig. 2B–D; compare insets in panels a1 and b1; panels f1 and g1; and panels j1 and k1). In contrast, they accumulated in numerous foci of variable size that were found dispersed throughout the nucleoplasm of the RPS19 knockdown cells (Fig. 2B–D, see panels b2, g2, and k2). Some of those foci were found in close contact with the nuclear envelope (Fig. 2B, C; panels b2 and g2, white arrows; see also Fig. 2F, second pair of panels from left). We also observed that a fraction of NOC4L and RRP12 located in the nucleoplasm did not accumulate in such foci, showing instead a diffuse distribution within the nucleoplasm (Fig. 2B–D, compare panels a2 and b2; and panels f2 and g2). Unlike the case of the other interrogated RBFs, we found that LTV1 did not localize in foci (Fig. 2E, panel o2). However, it also exhibited an abnormal nucleoplasmic localization completely different from the predominant cytosolic distribution found in normal cells (Fig. 2E, compare panels n2 and o2). This suggests that LTV1, which is not present in 21S-containing preribosomes in normal cells, had been entrapped in the nucleoplasm by abnormal preribosomes or subparticles. We also detected the formation of foci and the loss of bystin from the nucleolus using standard immunofluorescence assays with antibodies to bystin in parental HeLa cells transfected with the RPS19 siRNA (Supplementary Fig. 4A, B, top panels), indicating that these changes in localization are not artifacts caused by the incorporation of the GFP-tag in the reporter protein. It is worth noting, however, that the foci visualized under these conditions showed much dimmer signals and less distinct appearance than those observed by epifluorescence microscopy in bystin-GFP-expressing cells. We surmise that this is probably due either to poor penetration of antibodies into bystin or to the partial masking of the bystin epitope in those abnormal complexes. Consistent with previous reports indicating that the overall integrity and subcompartment organization of the nucleolus are not affected under these conditions²⁶, we found that fibrillarin (FBL) was present in FC-DFC modules in RPS19 knockdown cells (Supplementary Fig. 4A, compare c and d).

Fig. 2. Aberrant accumulation of ribosome biogenesis factors (RBFs) in the nucleoplasm of RPS19-, RPS17- and RPS26-deficient cells.

A Summary of changes in 18S pre-rRNA species and in bystin-RBF interactions following depletion of RPS19, RPS17 or RPS26. Data were obtained 24 h after siRNA transfection and compiled from the northern blot and protein-interaction analyses shown in Supplementary Fig. 3. Relative pre-rRNA levels were quantified from northern blots and normalized to si-ctrl. Changes in bystin interactions with NOC4L, RRP12, LTV1 and NOB1 were determined by PSE extraction, GFP-Trap pull-down and western blot analyses in cells expressing bystin-GFP. The preribosomal complexes associated with each interaction and their distribution in PSE fractions (SN1–SN3) are indicated. B–E Subcellular localization of NOC4L-GFP, GFP-RRP12, bystin-GFP and LTV1-GFP in cells depleted of the indicated RPSs. Cells were analyzed by fluorescence microscopy 24 h after siRNA transfection. Green, GFP; blue, TO-PRO-3 DNA stain. For each condition, the original GFP image (panel 1), a brightness/contrast-enhanced image (panel 2), a magnified inset, and the corresponding DNA stain are shown. White arrows indicate NOC4L- or RRP12-containing foci located near the nuclear envelope (panels b1, b2, g1, g2, i1 and i2). Images from each reporter cell line were acquired using identical microscopy settings. Brightness/contrast adjustments were applied only to panel 2 images using identical parameters across conditions to facilitate visualization of nucleoplasmic foci. Representative images are shown. Similar results were obtained in at least two independent siRNA transfection experiments (biological replicates) for each GFP reporter cell line and depletion condition. F Inverted-contrast versions of selected images from (C) (f1, g1, h1 and i1) to facilitate visualization of GFP-RRP12-containing foci. Inverted-contrast images are marked with an asterisk to match the labeling of the original panels.

The foci formed in RPS19 knockdown cells were also positive for DDX21 and nucleophosmin/B23 (Supplementary Fig. 4C, D), two proteins normally associated with nucleolar 40S preribosomes. DDX21 is an RBF that is predominantly present in bipartite and very-early 40S particles^10,47 (see scheme in Fig. 1B). Nucleophosmin/B23 is a scaffolding protein that interacts with both RBFs and pre-rRNAs to promote preribosome fluidity and phase separation⁴⁸. These results further show that the foci in RPS19 knockdown cells contain many proteins normally associated with early nucleolar 40S preribosomes. To demonstrate the presence of preribosomes, we performed RNA FISH analyses in GFP-RRP12-expressing cells using probes that detect all 18S pre-rRNA species (5’-ITS1 probe) or just the 18S pre-rRNA species present in bipartite and early 40S preribosomes (ETS1 probe) (Supplementary Fig. 5A and see also Fig. 1B for a scheme of 18S pre-rRNA maturation in the context of preribosomes). We observed that the 5’-ITS probe labeled most of the GFP-RRP12⁺ foci formed in RPS19 knockdown cells (Fig. 3A, B and Supplementary Fig. 5B–D). In contrast, the ETS1 probe only yielded positive signals with very low intensity in a small percentage of foci that, in general, were very large in size (Fig. 3A, B and Supplementary Fig. 5B–D). We found a similar detection of foci, specifically with the 5’-ITS probe, in parental HeLa cells transfected with the RPS19 siRNA (Fig. 3C, D and Supplementary Fig. 6), further indicating that these abnormal preribosome accumulations are not artefactual entities. These patterns suggest that these foci primarily contain the abortive 21S-containing preribosomes that are massively produced upon RPS19 depletion. In normal cells, 21S-containing preribosomes (post-A1 state) are SSU processomes that have undergone the cleavage of 30S pre-rRNA into 21S pre-rRNA. This state precedes subsequent maturation steps, including further pre-rRNA processing, 5’-ETS degradation and processome dismantlement, which together generate the more mature pre-40S particles destined for nuclear export (Fig. 1A)⁴. Structural studies have demonstrated that post-A1 SSU processomes contain up to 44 different RBFs, sharing high structural and compositional similarity with pre-A1 processomes (see image of pre-A1 state; Fig. 1A)¹⁰. To analyze the preribosomal complexes formed in RPS19 knockdown cells, we utilized GFP-RRP12 as bait for GFP-Trap purification, employing an extraction protocol optimized to recover SSU-processomes across varying maturation states¹⁰. Mass spectrometry analysis of RRP12-associated proteins identified 37 of the 39 core pre-A1/post-A1 RBFs, together with six additional state-specific RBFs, in the RPS19 knockdown samples (Fig. 3E). Although the extent of recovery of individual RBFs fluctuated between experiments, likely reflecting technical variations in extraction yield or siRNA knockdown efficiency, the majority of the 43 RBFs were enriched in si-RPS19 cells compared to controls in the two biological replicates analyzed (42 and 31 RBFs, respectively). The association of RRP12 with nearly the entire set of processome components, together with the non-uniform RBF enrichment patterns, indicates a population of heterogeneous complexes, likely comprising a mixture of nearly-complete (pre-A1, post-A1 or aberrant) processomes and “incomplete” ones (which might exist in vivo or arise during the extraction process). Regardless of the source of this heterogeneity, these results show that RRP12 is predominantly sequestered within large, multi-RBF assemblies in RPS19-depleted cells, with no evidence of significant accumulation as a free protein or within small subcomplexes. These findings are consistent with microscopy-based visualization of multiple processome RBFs and FISH detection of pre-rRNA within foci, further supporting the idea that RRP12-containing foci represent accumulations of abortive nucleolar preribosomes.

Fig. 3. Accumulation of preribosome-containing foci in the nucleoplasm of RPS19-deficient cells.

A, B Co-localization of RRP12 and 18S pre-rRNAs in GFP-RRP12-expressing cells 24 h after siRNA transfection, analyzed by RNA FISH using 5′-ITS1 (Cy3) and ETS1 (Cy5) probes. Complete image sets are in Supplementary Fig. 5. B Quantification of GFP-RRP12- and/or pre-rRNA-containing foci. Each point represents one nucleus. Data represent mean ± SEM from n = 3 independent transfection experiments, encompassing a total of ≥115 nuclei per condition across seven microscopy fields. Statistical significance was determined by ordinary two-sided two-way ANOVA with Bonferroni’s multiple-comparisons test. C, D Localization of 18S pre-rRNAs in parental HeLa cells 24 h after siRNA transfection by RNA FISH. Image of 5′-ITS1 detection in (C) (complete image sets in Supplementary Fig. 6). D Quantification of pre-rRNA-containing foci. Data were obtained from seven microscopy fields across n = 3 biological replicates, with ≥146 nuclei per condition. Statistics were as in (B). E Co-purification of SSU processome components with RRP12 in cells endogenously expressing GFP-RRP12, 24 h after si-RPS19 transfection. Heatmaps show fold changes in RBF abundance (normalized to GFP-RRP12) in RPS19-depleted versus control cells (two biological replicates). F–I Effects of partial RPS19 depletion. F Western blot showing RPS19 levels after efficient (si-RPS19-2) and partial (si-RPS19-1) depletion. Similar results were obtained in two independent experiments. G Images of GFP-RRP12- and LTV1-GFP-expressing cells 24 h after si-RPS19-1 transfection (complete sets in Supplementary Fig. 7). Results were reproduced in two independent experiments. H, I Pre-rRNA levels and 28S/18S rRNA ratios by northern blot and Bioanalyzer. I shows mean ± SEM (n = 3 biological replicates). J–L GFP-RRP12 dynamics during mitosis following depletion of RPS19, RPS17, or RPS26. Images in K are maximum-intensity projections (additional images in Supplementary Figs. 8–11 and Videos 1–4). L Quantification of GFP-RRP12 foci. Data represent mean ± SEM from n = 4–6 individual tracked cells per condition. Data points indicate values for single cells. Statistical significance was determined by Two-way Mixed-Effects analysis with Bonferroni’s multiple-comparisons test to compare each knockdown to the si-ctrl group at individual time points. Source data are provided as a Source Data file.

The massive formation of foci and the retention of preribosome subcomplexes in the nucleoplasm is also observed under conditions in which a 45-60% reduction is achieved by using a low-efficiency siRNA (siRPS19-1) (Fig. 3F, G and Supplementary Fig. 7). This is particularly significant, as these cells display only a moderate block in 21S pre-rRNA processing and no major reduction in mature 40S subunits (Fig. 3H, I). This is a phenotype that resembles the 40S maturation defect observed in cells from patients with RPS19 haploinsufficient DBA⁴⁵.

Aberrant RBF-containing foci of RPS19-deficient cells originate in late mitosis

We next performed time-lapse fluorescence microscopy to investigate the formation of preribosomal foci in GFP-RRP12–expressing cells transfected with RPS19 siRNAs. To this end, cultures were enriched in S phase by incubation with excess thymidine and, following thymidine removal, we monitored cells undergoing completion of cell division (for a scheme, see Fig. 3J). In control cells, we found that GFP-RRP12 displayed the expected behavior of nucleolar components: (a) dispersion into the nucleoplasm after the nucleolar breakdown that takes place in metaphase; and (b) re-accumulation during late anaphase and telophase in both the emerging nucleoli and PNBs (Fig. 3K, b and c and see also Supplementary Fig. 8 and Supplementary Video 1). PNBs are transient intermediate condensates formed by the dispersed preribosomes (containing early pre-rRNA species, early-assembly RPSs and RBFs) that are liberated during nucleolar breakdown. Such preribosomes are thought to undergo concentration-driven condensation as the nuclear envelope is re-established in late mitosis, forming discrete foci (PNBs) that normally dissolve as they are re-incorporated into the nascent nucleolus^21–23. In the case of RPS19 knockdown cells, GFP-RRP12 was normally localized in nucleoli prior to mitosis (Fig. 3K, compare panels a and f and Supplementary Video 2) and dispersed thereafter during metaphase before concentrating into foci that closely resembled PNBs (Fig. 3K, compare panels c and h). However, unlike in control cells, these foci failed to disassemble or reintegrate into the newly-formed nucleoli and persisted as such well after mitosis (Fig. 3K, compare d and e with panels i and j; Fig. 3L; Supplementary Fig. 9 and Video 2). These observations indicate that the foci observed in RPS19 knockdown cells expressing GFP-RRP12 contain defective preribosomes released from the nucleolus during metaphase, which subsequently form aberrant PNB-like accumulations highly resistant to dissolution at the end of mitosis. While these structures share the size and temporal formation of PNBs, and may exhibit liquid-like condensate properties (see Discussion), their exact physical nature remains to be characterized. Accordingly, we hereafter opt for the general term “aggregates” rather than “condensates” to refer to them.

The RPS17 deficiency elicits the accumulation of different types of aberrant preribosomal complexes in the nucleoplasm

We next monitored the formation of aberrant preribosomes in the HeLa cell derivatives knocked down for RPS17. The lack of RPS17 generates defective preribosomes that cannot be exported out of the nucleus. Due to this, cells lacking this protein exhibit reduced levels of 18S-E pre-rRNA as they lack the cytoplasmic pool of 18SE-containing pre-40S complexes⁴⁰ (Fig. 1B). Those cells also accumulate upstream pre-rRNAs (21S and 26S) (Supplementary Fig. 3A, B), probably due to the entrapment of early RBFs (e.g., PNO1, bystin or RRP12) in the abortive preribosomes or subparticles that are retained in the nucleoplasm. Those RBFs cannot be recycled back to the nucleolus, where they are needed for the maturation of the 26S- and 21S-containing preribosomes^43,49,50. Unlike the case of RPS19 knockdown cells, we found that the aberrant preribosomal complexes formed in RPS17 knockdown cells were soluble and distributed homogeneously in the nucleoplasm. Consistent with this, we observed that the contents of complexes containing bystin plus RRP12, or bystin plus NOB1, in the SN2 fraction of the PSE method were higher in RPS17 knockdown cells when compared to control cells (Supplementary Fig. 3D, compare lanes 11 and 15; Supplementary Fig. 3F, compare lanes 7 and 9). Higher contents of complexes containing either bystin plus RRP12 or bystin plus NOB1 were also found in the PSE SN1 fraction from those cells (Supplementary Fig. 3E, compare lanes 6 and 8; Supplementary Fig. 3G, compare lanes 4 and 5; for a summary, see scheme in Fig. 2A, RPS17 knockdown column #2). RPS17 knockdown cells also exhibited a mild (in the case of NOC4L and LTV1) and a strong (in the case of RRP12 and bystin) accumulation of RBFs in the nucleoplasm (Fig. 2B, compare panels a2 and c2; Fig. 2E, compare panels n2 and p2; Fig. 2C, compare panels f2 and h2; Fig. 2F, compare panels f1* and h1*; and Fig. 2D, compare panels j2 and l2). They also displayed a loss of RRP12 from the cytoplasm (Fig. 2C, compare panels f2 and h2; Fig. 2F, compare panels f1* and h1*), and a loss of both RRP12 and bystin from the interior of the nucleolus (Fig. 2C, see zoom of inset in h1; Fig. 2D, see zoom of inset in l1). Using FISH pre-rRNA analyses with 5’-ITS1 and ETS1 probes, we found increased contents of 18S pre-rRNAs in the nucleoplasm of those cells and their concomitant loss from the cytoplasm (Supplementary Fig. 6B, compare image a’ and image e’). Collectively, these data indicate that RRP12 and bystin are mostly associated with abortive preribosomes that accumulate in a diffuse manner throughout the nucleoplasm.

Time-lapse microscopy of GFP-RRP12-expresssing cells further revealed that the depletion of RPS17 elicited a diffuse accumulation of preribosomes following nucleolar disassembly in metaphase (Fig. 3K, panel l; see also Supplementary Fig. 10 and Video 3) as well as the formation of RRP12⁺ PNB-like foci at the end of mitosis. These foci are more stable than normal PNBs but less persistent and numerous than those formed in the case of RPS19 knockdown cells (Fig. 3K, note the presence of RRP12-containing foci in panels n and o but not in panels d and e; for quantifications, see Fig. 3L; see also Supplementary Fig. 10 and Supplementary Video 3). Such foci could not be detected in the foregoing RBF reporter microscopy analyses at 24 h after the RPS17 siRNA transfection (Fig. 2), probably because they are progressively “hidden” by the large diffuse accumulations of the GFP-RBF signals.

Taken together, these findings indicate that the RPS17 deficiency generates two types of preribosome accumulations in the nucleoplasm. The predominant type consists of 18SE-containing preribosomes that completed most (if not all) maturation steps inside the nucleolus and, therefore, exhibit normal solubilization properties, do not aggregate in foci in late mitosis, and accumulate in a diffuse manner in the nucleoplasm. The second type, which might correspond to the 21S or 26S-containing preribosomes generated as a secondary defect in the nucleolus, includes preribosome accumulations released in metaphase that can generate: (a) RRP12-containing aggregates that fail to dissolve properly, and (b) subparticles that do not form aggregates. The presence of these latter structures is inferred from the diffuse accumulation of NOC4L in the nucleoplasm and from the detection of bystin–NOC4L complexes in the SN2 fraction of the PSE method.

The RPS26 deficiency elicits aberrant subparticle accumulations in the nucleoplasm

The depletion of RPS26, which acts downstream of RPS17, triggers the accumulation of 18S-E pre-rRNA species in the cytoplasm due to a defect in the final 40S maturation steps^40,51 (see scheme of pre-rRNA intermediates in 40S preribosomes, Fig. 1B). It also increases the content of the 26S pre-rRNA (Supplementary Fig. 3A, B) that, similarly to the case of RPS17, is likely due to the entrapment of early-maturation RBFs in the aberrant particles that get stuck in the cytoplasm. We observed using the PSE method that RPS26 knockdown cells exhibited higher than normal levels of bystin-GFP–NOB1 complexes in the SN2 fraction (Supplementary Fig. 3F, compare lanes 7 and 11). They also showed increased contents of bystin-GFP associated with both RRP12 and NOB1 in the SN1 fraction (Supplementary Fig. 3G, compare lanes 4 and 6; for a summary of results, see scheme in Fig. 2A, RPS26 knockdown, column #3). These data could be consistent prima facie with the presence of those RBFs in the abortive preribosomes formed in the cytoplasm. In line with this, we observed using FISH that 40S preribosomes accumulated predominantly in the cytoplasm but not in the nucleoplasm (Supplementary Fig. 6, see panel B, image g’). However, we also found that complexes harboring RRP12 accumulate mostly in a diffuse manner and in a few scattered dim-fluorescence foci in the nucleoplasm rather than in the cytoplasm (Fig. 2C, compare panels f2 and i2 and Fig. 2D, compare panels j2 and m2). This suggests that RPS26 knockdown cells accumulate 40S preribosomes in the cytoplasm as well as RRP12-containing subcomplexes or subparticles, but not preribosomes with intact pre-rRNA, in the nucleoplasm. The abnormal RRP12-containing preribosomes formed in RPS26 knockdown cells do not form persistent aggregates in late mitosis (Fig. 3K, L, Supplementary Fig. 11, and Video 4), further suggesting that aggregation-prone abortive preribosomes must be highly structured and contain mostly intact pre-rRNA species.

Based on all readouts used, we can classify the phenotypic features elicited by the three interrogated RPS deficiencies in five subtypes: (1) pre-rRNA processing defects (Fig. 4A; upper panels, orange background); (2) abnormal fractionation properties when using the PSE method (Fig. 4A, set of second panels from top, green background); (3) predominant accumulation of four 40S preribosome-associated factors in the nucleoplasm (Fig. 4A, set of third panels from top, yellow background); (4) abnormal exclusion of RBFs from the nucleolus (Fig. 4A, dark yellow background panels); and (5) sustained presence of RBF aggregates in numerous foci (Fig. 4A, bottom panel, gray background). By integrating all five phenotypic features, the effects elicited by the three interrogated RPS depletions, in terms of aberrant preribosome accumulation in the nucleoplasm, can be considered as (see schemes in Fig. 4B): (1) strong effect (e.g., RPS19 depletion), which is associated with preribosomes originating inside the nucleolus. The effects found include high levels of accumulation of abortive aggregation-prone preribosomal particles, the mild accumulation of preribosomal subparticles (with no intact pre-rRNA) in the nucleoplasm, and a drastic impairment in the recycling of RBFs back to the nucleolus; (2) strong effect (e.g., RPS17 depletion), which is associated with the abundant fraction of preribosomes that have completed nucleolar maturation and with a less abundant fraction of preribosomes that originated inside the nucleolus. The effects observed in this case include a large accumulation of complexes that lack a tendency to aggregate, as well as a moderate accumulation of both aggregation-prone nucleolar complexes and dispersed subcomplexes. As above, these aberrant features also cause a strong impairment in the recycling of RBFs back to the nucleolus; and (3) moderate effect (e.g., RPS26 depletion), which is mostly associated with subcomplexes or subparticles of nucleolar origin. In this case, the defects seen include a moderate accumulation of nucleolar subcomplexes, a reduced presence of aggregation-prone preribosomes, and partial defects in RBF recycling.

Fig. 4. Distinct patterns and extents of aberrant preribosome material accumulation in the nucleoplasm induced by blockages at different steps of 40S ribosomal subunit maturation.

A Color-coded summary of changes in 18S pre-rRNA species, PSE-fractionated RBF-containing complexes, and RBF subcellular localization in cells depleted of three RPSs and five 40S RBFs. The scheme compiles the data shown in Figs. 2 and 3 and Supplementary Figs. 3, 12 and 13. Relative pre-rRNA levels (orange panel) were quantified from northern blots and normalized to si-ctrl. Changes in RBF interactions (green panel) were determined by PSE extraction, GFP-Trap pull-down and western blot analyses. The preribosomal intermediates associated with each interaction and their distribution in PSE fractions are indicated: SN1 contains nucleoplasmic (NUC) and cytoplasmic (CYT) complexes, SN2 contains NO2 nucleolar complexes, and SN3 contains 30S-, 21S- and NO1-containing nucleolar complexes. Changes in RBF localization (yellow panel) were determined by microscopy analyses. Color and pattern codes indicate both the abundance of accumulated complexes and the mode of accumulation of NOC4L, bystin, RRP12 and LTV1 in the nucleoplasm. The bottom panel summarizes the persistence of nucleoplasmic RRP12-containing foci after mitosis, determined by live-cell imaging. B Models of aberrant preribosome accumulation in RPS19-, RPS17- and RPS26-deficient cells. The scheme on the left shows the color code used for preribosomes located in the nucleolus (green), nucleoplasm (pink) and cytoplasm (blue) under normal conditions. Major features of the accumulating preribosomes or subparticles are indicated in the white boxes. Characteristics of the nucleoplasmic foci formed during mitosis are shown in the dashed box. White and pale-green nucleoli indicate defects in RBF recycling to the nucleolus.

Disruption of 40S subunit nucleolar maturation steps elicits RPS19 knockdown-like phenotypes

In order to further analyze the specificity of the phenotypes induced by RPS19, RPS17 and RPS26 knockdowns, we decided to investigate the knockdowns of transcripts encoding RBFs involved in 40S subunit nucleolar maturation (bystin, RRP12, PNO1), nuclear export (WBSCR22) and cytoplasmic maturation (TSR2). Bystin and RRP12 are required at the same 40S synthesis step as RPS19 and, therefore, the depletion of each one of those proteins blocks the maturation of preribosomes containing 21S pre-rRNA species^43,49,50. PNO1 works at an earlier synthesis step in the nucleolus, with its knockdown promoting a block in the maturation of preribosomes containing 26S pre-rRNA species^43,50. To carry out these studies, we modified the experimental approach used in previous sections in the following manner: (i) the analyses were performed 48 h after the siRNA-transfection step, given that the foregoing RBFs were depleted more slowly than the RPSs analyzed in our prior experiments. Consistent with this, we observed that the effect of the protein depletion at the 48 h post-transfection time point was comparable in terms of increments of the 28S/18S rRNA ratio to the 24 h time points used in the RPS knockdowns (Supplementary Fig. 2D–G); (ii) we utilized GFP-RRP12 rather than bystin-GFP as bait in the PSE fractionation experiments when analyzing bystin knockdown cells. In the case of the RRP12 depletion, we used the confocal microscopy and complex formation data from a previous study carried out in our laboratory⁴³. In the case of the bystin- and PNO1-depletion experiments, all data were generated in this work (Supplementary Figs.  12, 13). As summarized in Fig. 4A, the knockdown of bystin, RRP12, or PNO1 elicits phenotypic features in HeLa cells similar to those found in the RPS19 knockdown counterparts. Such features included the decrease in the recovery of preribosomes in the fractions of the PSE method (Fig. 4A, second set of panels from top, green background), the retention of LTV1 in the nucleoplasm, and the accumulation of NOC4L⁺ (in the case of bystin and RRP12 knockdown cells) and/or RRP12⁺ (in the case of bystin and PNO1 knockdown cells) foci and diffuse subcomplexes in the nucleoplasm of the appropriate siRNA-transfected cells (Fig. 4A; bottom panels, yellow background). Despite this, these alterations were milder than in the case of RPS19 knockdown cells as evidenced from the lack of some RBFs in some of the foci, the partial retention of LTV1 inside the nucleoplasm, or the partial loss of NOC4L from the nucleolus (Fig. 4A and Supplementary Fig. 13). None of those defects could be observed in the case of HeLa cells with depleted WBSCR22 or TSR2 (Fig. 4A), two proteins involved in the nuclear export (WBSCR22) and cytoplasmic maturation (TSR2) of 40S subunits^52,53. These findings indicate that additional alterations that promote a block in the maturation of 40S subunits in the nucleolus can generate, similarly to the depletion of RPS19, persistent aggregation-prone preribosomes that prevent RBF recycling. Furthermore, the differences in the extents of RBF accumulations and recycling defects induced by the specific depletions of RPS19, bystin or RRP12 indicate that blockages at the same pre-rRNA processing stage (21S-containing preribosomes) do not necessarily produce the same abortive preribosomes.

Patterns of nucleoplasmic RBF accumulation reveal functional subclasses among 40S ribosomal proteins

The above findings prompted us to address the following questions: (i) do all types of abortive preribosomes formed in the nucleolus tend to accumulate in the nucleoplasm? (ii) are the RBF accumulation patterns conserved among deficiencies of RPSs belonging to the same or different functional classes? and (iii) does the extent of accumulation vary substantially among them? To answer those questions, we implemented an automated imaging strategy to systematically analyze the impact of individually depleting all RPSs on protein homeostasis in the nucleus. To this end, we transfected HeLa cell derivatives endogenously expressing the reporters NOC4L-GFP, GFP-RRP12, bystin-GFP or LTV1-GFP with high-efficiency siRNA pools for each single RPS transcript and, 24 h later, we immunostained cells with antibodies to FBL, incubated them with 4′,6-diamidino-2-phenylindole (DAPI) to decorate the DNA, and imaged them on an automated spinning disk microscope (see experimental flowchart in Supplementary Fig. 14). As specificity controls, we also included in these screenings siRNA pools for six components of the large ribosomal subunit, two of which were known to trigger nucleolar disruption (RPL5 and RPL11)²⁶. As negative controls, we included a non-targeting (CTRL) siRNA pool and siRNA pools targeting a transcript of an RPS isoform (RPS27L) that is not expressed in HeLa cells. We chose an image analysis pipeline that aimed at identifying the following features: (i) percentage of the interrogated protein exhibiting a diffused localization pattern in either the nucleoplasm (NOC4L-GFP, bystin-GFP, GFP-RRP12, LTV1-GFP) or the cytoplasm (NOC4L-GFP, bystin-GFP, GFP-RRP12); (ii) nucleolar/nucleoplasm content ratio of NOC4L-GFP, bystin-GFP, and GFP-RRP12 (a parameter that measured losses of proteins from the nucleolus and/or diffuse gains of those proteins in the nucleoplasm); (iii) number of bright, PNB-like nucleoplasmic foci positive for NOC4L-GFP, bystin-GFP or GFP-RRP12; and (iv) integrity and overall size of nucleoli (area occupied by FBL⁺ subcellular structures). Finally, we used Z-scores for features that monitored the relative subcellular distribution of NOC4L-GFP, GFP-RRP12, bystin-GFP or LTV1-GFP (13 features in total) to cluster the effect of siRNAs according to phenotypic similarity. FBL⁺ structures were analyzed separately because, unlike the other scored parameters, they could only inform about the distribution of very early 40S preribosomes. This experimental setup exhibited high reproducibility across different features for well- and plate-replicates as well as high penetrance of the phenotypes observed under a given condition (Supplementary Fig. 15). Consistent with this, the patterns obtained with the previously characterized RPS19, RPS26 and RPS17 knockdowns were validated in these automated analyses (see below).

Based on the heatmap obtained (Fig. 5A) and the phenotypic similarity network built from the correlations among feature signatures for each pair of perturbations (Supplementary Fig. 16A), we found that the effects of RPS knockdown could be clustered in three well-defined phenotypic classes (which will be referred to hereafter as class I, II and IV), a small-size class (class III), and a more-heterogenous class (class V). In addition, two of the foregoing classes (III and IV) could be further segregated into two subtypes (III-a, III-b, IV-a, IV-b) (Supplementary Fig. 16A).

Fig. 5. Two functional classes of early-assembly RPSs revealed by distinct patterns of abnormal RBF accumulation.

A Hierarchical clustering heatmap from automated imaging analyses (Supplementary Fig. 14) of 13 features (right) monitoring the subcellular localization and distribution of NOC4L, bystin, RRP12 and LTV1 following individual knockdowns of all 33 RPS genes. Feature values were normalized to the non-targeting (CTRL) siRNA. Knockdowns of S27L (non-expressing isoform) and six RPL genes served as specificity controls. Alterations in fibrillarin-containing structures were quantified but excluded from clustering and are shown separately. Five major phenotypic classes identified by correlation analysis of RPS knockdowns are color-coded. The corresponding phenotypic similarity network is shown in Supplementary Fig. 16A. Positive and negative Z-scores apply to all features except feature #3 (LTV1 diffuse in nucleoplasm), which is shown on a 0–3 scale because of its high Z-scores. Z-score values for ten features are shown in Supplementary Fig. 17. B Heatmaps of class I (eight RPSs), class II (four RPSs), and two mixed I/II RPS knockdowns extracted from (A) for comparison of phenotypic features. C Pre-rRNA processing defects and blocked step of 40S maturation in the RPS knockdowns shown in (B), were taken from a previous systematic study of individual RPS knockdowns in HeLa cells⁴⁰ whose data have been largely validated by an independent study²⁶. The pre-rRNA processing pathway is shown on the right. D, E Positions of class I, class II and mixed I/II RPSs in the human 40S ribosomal subunit structure (PDB 6G5H). 18S rRNA is shown in orange. Remaining RPSs are dark gray. F Positions of class I (brown), class II (purple), class III (green) RPSs, and NOC4L, RRP12 and bystin in human SSU processomes at the pre-A1 (PDB 7MQ8) and post-A1 (PDB 7MQA) maturation states. NOC4L, RRP12 and bystin assemble at a late processome assembly step, after the initial structuration and folding of the three 18S rRNA domains. Transition from pre-A1 to post-A1 involves A1 cleavage to generate 21S pre-rRNA, release of pre-A1-specific RBFs, recruitment of post-A1-specific RBFs, and incorporation of RPS19. Structures were visualized using Mol* tools at the RCSB PDB.

Class I showed the most uniform and consistent phenotype that was characterized by a high degree of paired-feature correlation (Supplementary Fig. 16B). This class comprises proteins whose depletion elicits a strong accumulation of bystin-GFP (Fig. 5B, feature 2), GFP-RRP12 (Fig. 5B, feature 5), and NOC4L-GFP (Fig. 5B, feature 6) in foci; the entrapment of LTV1 in the nucleoplasm (Fig. 5B, feature 3); and the loss of the diffuse distribution of both GFP-RRP12 and bystin-GFP in the nucleoplasm and cytoplasm (Fig. 5B, features 10 to 13). These concurrent alterations suggest that this phenotype is caused by the formation of aberrant NOC4L⁺ RRP12⁺ bystin⁺ preribosomes inside the nucleolus that, as in the case of RPS19 knockdown cells, can form persistent foci. Interestingly, this phenotypic class is composed of early-assembly RPSs whose depletion leads to a block of an early 40S maturation step inside the nucleolus (maturation of 30S-containing particles) according to a prior study⁴⁰ (Fig. 5C). These proteins are also known to be recruited very early in the pathway during the assembly of the 30S-containing SSU processome (Fig. 5F)¹⁰, to conform the body of the mature 40S subunit via interactions with the 5’ and central domains of the 18S rRNA (Fig. 5D). Importantly, our results indicate that the single loss of any of the proteins belonging to this class blocks maturation of the SSU processome but does not compromise its overall stability, as inferred by the apparent presence in those particles of RRP12, bystin and NOC4L, three proteins that bind to the processome after the initial structuration and folding of the three 18S rRNA domains have already taken place¹⁰.

Similar to class I, the class II group is characterized by the loss of both GFP-RRP12 and bystin-GFP diffuse localizations in the nucleoplasm and cytoplasm, as well as the retention of LTV1-GFP in the nucleoplasm (Fig. 5B, features 10 to 13 and 3). However, unlike class I proteins, the depletion of class II proteins causes the loss of NOC4L-GFP from the nucleolus (Fig. 5A, feature 9) without eliciting the accumulation of the three nucleolar RBFs in foci (Fig. 5A, see differences in features 2, 5, and 6 between classes I and II). This phenotype suggests that the elimination of these proteins leads to the formation of aberrantly assembled NOC4L⁺ or LTV1⁺ subparticles or subcomplexes that do not form aggregates. Class II proteins are known to be recruited during the early steps of assembly of the SSU processome to bind to the central domain (RPS14 and RPS13) and 3’ major domain (RPS16 and RPS28) of the primary 18S pre-rRNA when it is being transcribed¹⁰. Those regions will form part of the body or head in the mature 40S subunit (Fig. 5E). The individual depletion of class II RPSs, like in the case of class I, arrests maturation of 30S-containing precursors⁴⁰ (Fig. 5C). Despite this similarity, our results indicate that all these early-assembly RPSs can be separated into two different classes: class I proteins, which are not essential for the overall stability of 30S-containing SSU processomes, and class II proteins, which are required for acquiring or maintaining their overall stability. In line with such differential roles in processome assembly, class I proteins cluster around the 18S 5’ domain to build a primordial body whereas two of the class II proteins (RPS16 and RPS28) bind more distantly to the 18S 3’ major domain (Fig. 5F)¹⁰, a region that undergoes a major reconfiguration during the folding of 18S domains that takes place before the incorporation of RRP12, bystin and NOC4L¹⁰. Two additional RPS deficiencies (RPS11 and RPS5) exhibited mixed class I (features 5 and 6) and class II (features 8 to 13), suggesting that they might be forming: (a) a pool of abortive 30S-containing processomes that, similarly to those in class I deficiencies, can form aggregates; and (b) a pool of subparticles that, similarly to those in most class II deficiencies, sequester and prevent the recycling of NOC4L.

Class III included proteins whose depletion leads to a strong accumulation of diffuse GFP-RRP12 and NOC4L-GFP in the nucleoplasm as well as a concomitant loss of NOC4L from the nucleolus (Fig. 6A, see heatmap on the left, features 12, 8, and 9). This class can be further subdivided into two subclasses: (i) subclass III-a, which is associated with the presence of RRP12⁺ and NOC4L⁺ foci in the nucleoplasm (Fig. 6A, features 5, 6); and (ii) subclass III-b, which is characterized by the induction of a strong accumulation of diffuse bystin in the nucleoplasm and the absence of foci (Fig. 6A, features 13, 5 and 6). The two proteins in subclass III-a are known to be recruited early in the pre-A1 (30S/26S containing) processome (RPS18) or in the post-A1 (21S-containing) processome (RPS19) (Fig. 5F)¹⁰. Both of them are in close physical proximity and are involved in the maturation of post-A1 particles inside the nucleolus (Fig. 6C, B). In contrast, the two proteins of the subclass III-b (RPS17, RPS15) are not in close physical proximity (Fig. 6C) and become recruited into the early processome (RPS17) (Fig. 5F) or during the formation of the pre-40S particle (RPS15) (Fig. 6F). These two RPSs are required for the step of nuclear export^10,13,40. These results indicate that the two preribosome-accumulation phenotypes observed under RPS19 and RPS17 depletion in the first part of our study are characteristic only of two small functional subgroups: one involved in pre-A1-to-post-A1 nucleolar maturation and the other involved in nuclear export.

Fig. 6. Functional classes of RPS18, RPS19, RPS17, RPS15 and cytoplasmic-assembly RPSs revealed by distinct patterns of abnormal RBF accumulations.

A Heatmaps of preribosome-localization phenotypic analyses of classes III–V RPS knockdowns. Heatmaps have been extracted from the global hierarchical clustering heatmap shown in Fig. 5A to facilitate the visualization, description and comparison of phenotypic features. B Pre-rRNA processing defects and step of 40S maturation blocked in the RPS knockdowns shown in (A). The information on the contents of pre-rRNA species and the 40S maturation block was taken from a previous systematic study that analyzed individual knockdowns of all RPS proteins in HeLa cells after siRNA transfection⁴⁰, and whose data have been largely validated by another independent study²⁶. The pre-rRNA processing ladder is shown on the right. C–E Positions of classes III–V RPSs in the structure of the human 40S ribosomal subunit (PDB ID 6G5H). 18S rRNA is shown in orange. RPSs that do not belong to the indicated classes are shown in dark gray. F Positions of RPSs on the previously published post-A1 and state-B cryo-EM structure models¹², and steps of incorporation of RPSs and of dissociation of RBFs during late pre-40S maturation. The color code for RPSs is the same as in the previous figures: class I RPSs in brown, class II RPSs in purple, class III RPSs in dull green, class IV RPSs in blue, and class V RPSs in bright green. Discrete steps of maturation that occur between the post A1 state and state B, and between state B and the mature 40S subunit are indicated by successive arrows. The RPSs incorporated and the association (white rectangle) and release (gray rectangles) of RBFs at each step have been described in previous studies^10–12. The steps for incorporating RPS25 and RPS30 are not entirely clear. Distinct proteins on the Post-A1 state (PDB ID: 7MQA) and pre-40S state B (PDB ID:6G4S) structures have been colored using the Mol* 3D viewer tools at the Protein Data Bank web page.

Class IV includes proteins whose depletion induces a very strong accumulation of bystin-GFP in the cytoplasm (Fig. 6A, heatmap on the right, feature 11) together with increased contents of NOC4L-GFP in the nucleoplasm (Fig. 6A, see heatmap on the right, feature 8). These phenotypic features go along with either the presence (in the case of subclass IV-a) or the absence (in the case of subclass IV-b) of GFP-RRP12⁺ foci (Fig. 6A, see heatmap on the right, feature 5). Five of the proteins included in class IV (RPS3, RPS29, RPS20, RPS27a, and RPS2) are involved in late steps of the 40S maturation in the cytoplasm⁴⁰. However, it is known that their individual depletion causes secondary blocks in the maturation of preribosomes in the nucleolus as well (Fig. 6B, see scheme on the right, note the strong accumulations of 21S and 26S pre-rRNAs). The remaining three proteins belonging to class IV (RPS27, RPSA, RPS21) had been previously associated with the early and intermediate maturation of 40S particles inside the nucleolus⁴⁰. However, our data suggest that these three proteins, which are in close topological proximity in the mature 40S subunit (Fig. 6D), must play roles in late maturation steps in the cytoplasm (see Discussion). Together, the observed phenotypic features indicate that in the cases of class IV deficiencies, there is a formation of 21S-containing and 26S-containing processomes that, although do not have the structural regions containing all three RBFs (bystin, RRP12 and NOC4L) stably assembled, are stable enough to accumulate diffusely or form persistent aggregates in the nucleoplasm.

Finally, class V is composed of RPSs whose depletion induces nucleoplasmic accumulations of NOC4L and/or RRP12 (Fig. 6A, see heatmap on the right, features 8 and 12), no significant accumulation of PNB-like foci (Fig. 6A, right panel, features 2, 5, and 6), and no retention of bystin and RRP12 in the cytoplasm (Fig. 6A, right panel, features 10 and 11). Proteins belonging to this phenotypic class are known to be required for cytoplasmic maturation of 40S subunits and, therefore, their depletion leads to the accumulation of the 18S-E pre-rRNA species (Fig. 6B, see scheme on the right). They all are assembled during early pre-A1 processome (RPS12) assembly or pre-40S late (RPS10, RPS26) maturation steps and will become positioned in the upper body, the platform or the head of the mature 40S subunit (Figs. 5F, 6F, E and note that the steps of incorporation of RPS25 and RPS30 are not totally clear). Our results show that, unlike the case of class IV proteins, the abortive cytoplasmic preribosomes generated by a deficiency in class V proteins do not sequester bystin (Fig. 6A, see heatmap on the right, feature 11) and must be blocked at a maturation step downstream of the pre-40S state C (which must sequester PNO1) (Fig. 6F, see the steps of release of RBFs boxed in gray rectangles). Due to this, the diffuse accumulation of NOC4L and RRP12 in the nucleoplasm of the class V knockdown cells must reflect subparticles generated by abortive preribosomes formed in the nucleolus that, as a consequence of the lack of availability of PNO1, do not reach a structured stable processome state. This is similar to the case previously found in class II knockdowns. The phenotypic separation of class IV and V proteins defines two functional subclasses of cytoplasmic maturation RPSs that act upstream (class IV) and downstream (class V) of bystin release¹¹ (Fig. 6F, see the step of bystin release).

Three of the RPSs interrogated in our screening (RPS23, RPS27L, and RACK1) could not be grouped in any of the phenotypic classes described above. In the case of RPS23, we observed that its depletion had negligible effects on the distributions of the RBFs interrogated in our study (Fig. 5A). In the case of RPS27L, its depletion caused some loss of bystin and RRP12 in de nucleolus and nucleoplasm respectively, but it did not elicit any accumulation of the RBF reporters (Fig. 5A). Finally, the depletion of RACK1 induced phenotypes similar to those found in cells transfected with our non-specific- and indirect-phenotype siRNA controls (Supplementary Fig. 16A). Incidentally, we found that the phenotype of cells depleted in RPL5 or RPL11, two proteins implicated in the maturation of 60S subunits, resembled that found in class V knockdown cells (Fig. 5A and Supplementary Fig. 16A). This suggests that the role of these 60S pathway proteins in the maintenance of nucleolar structure is probably required for the correct functioning of the 40S synthesis machinery (Fig. 5A, note the strong increase in the areas of FBL-containing structures, which is indicative of the unfolding of the nucleolus²⁶).

In summary, the above findings identify three distinct modes of accumulation of preribosomal particles and subparticles in the nucleoplasm caused by primary blocks in the nucleolus, defined by the class I, class II and class III-a deficiencies. They also reveal more heterogeneous patterns, which resemble those of class II or class III-a, caused by primary blocks in the cytoplasm (classes IV and V) that secondarily impair processome assembly or maturation. Collectively, the results also indicate that the formation of abortive preribosomes in the nucleolus always leads to large accumulations of particles and/or subparticles in the nucleoplasm, regardless of whether 40S maturation is interrupted by primary or secondary nucleolar blocks.

Extent of nuclear proteostasis alterations in RPS deficiencies

We finally used the data described in the previous section to quantify the relative impact of each RPS knockdown on nuclear proteostasis. To this end, we quantified: (i) the presence of RBFs in persistent foci, a parameter that identifies a pool of aberrant preribosomal complexes that are stable enough to reach the mitotic phase of nucleolar reformation and form aggregates; (ii) the presence of dispersed RBFs, a parameter indicative of the formation of aberrant subparticles or small subcomplexes of nucleolar origin that are not rapidly dissolved; and (iii) the presence of LTV1, a parameter that gives an indication of the number of free or complexed LTV1 that are aberrantly retained inside the nucleus under each of the experimental conditions used (Supplementary Fig. 17). Using this approach, we could identify 12 RPSs whose depletion elicits a very similar pattern (Fig. 7A, grouped in a dashed green rectangle): strong accumulation of NOC4L⁺ bystin⁺ RRP12⁺ foci (Fig. 7A, see foci quantification in columns on the left) that is accompanied by a strong accumulation of diffuse LTV1 in the nucleoplasm. This group includes the two class II/I (mixed phenotype) proteins (RPS5, RPS11), all class I proteins (RPS3A, RPS4X, RPS6, RPS7, RPS8, RPS9, RPS15 A, RPS24), and the two class III-a proteins (RPS18, RPS19). Depletion of these proteins consistently reduces the average area of FBL⁺ structures (Fig. 5A, bottom panel, blue boxes), a result consistent with the localization of FBL in preribosome-containing foci. Notably, the depletions of three proteins of this latter class (RPS11, RPS18, RPS19) also display a pronounced accumulation of diffuse NOC4L or NOC4L plus RRP12 in the nucleoplasm (Fig. 7A). These concurrent alterations indicate a stronger impairment in RBF recycling, likely due to the formation of subparticles that are more persistent and stable than those generated in cells depleted in class I proteins. We found that the most pronounced phenotypes (in terms of persistent foci, presence of dispersed subcomplexes, and retention of LTV1) were those found in the case of cells depleted of class III-a proteins (Fig. 7A). This subclass includes RPS19, the most frequently mutated RPS in DBA patients^36,37.

Fig. 7. Nuclear preribosome proteostasis defects across RPS deficiencies: extents and patterns of accumulation, and proposed nature of the accumulating materials.

A Bubble plot showing changes in the nucleoplasmic contents of NOC4L, bystin and RRP12 (separated into diffusely-distributed and in foci) and LTV1 in RPS-knockdown cells. Z-score values used to generate the plot are shown in Supplementary Fig. 17. Raw data are included in the source data. Source data are provided as a Source Data file. Accumulations of diffuse and foci-associated nucleolar-derived materials are shaded in red and orange, respectively. Asterisks indicate transient (RPS17 knockdown) or scarce dim-fluorescent (RPS26 knockdown) foci detected by confocal microscopy. Twelve RPS knockdowns displaying accumulation of multiple RBFs in foci together with nucleoplasmic retention of LTV1 are highlighted by the green rectangle. Five Class IV RPS knockdowns with mixed (foci and diffuse) phenotypes are highlighted by the purple rectangle. RPS deficiencies associated with DBA and their incidence are shown as blue bubbles. The primary 40S maturation block is indicated on the left and the preribosome-accumulation class on the right. int., intermediate. B Schemes summarizing nucleoplasmic preribosome accumulation patterns and their proposed contents. Colors indicate the normal localization of 40S preribosomes: nucleolus (green), nucleoplasm (pink) and cytoplasm (blue). Co-localization of multiple RBFs and pre-rRNA, their appearance during mitosis, and the predominant accumulation of processome-like assemblies in RPS19-deficient cells support the interpretation that foci are aggregates of large abortive SSU processomes. In contrast, diffuse materials likely represent products of processome disassembly or misassembly lacking intact pre-rRNAs. Primary nucleolar defects (left) generate class I, class II and class III-a phenotypes, whereas primary cytoplasmic defects (right) trigger secondary nucleolar maturation defects resembling class II or class III-a phenotypes. Proteins belonging to each class are indicated on the right. White and dim-green nucleoli denote strong and partial defects in RBF recycling to the nucleolus, respectively. Schemes illustrating the primary and secondary phenotypes of Class III-b (RPS17 and RPS15 deficiencies) are shown in Fig. 4B.

The individual depletion of five class IV proteins (RPS29, RPS3, RPS20, RPS27a, RPSA) (Fig. 7A, grouped in a dashed purple rectangle) also induces a mixed phenotype that partially resembles, in terms of the extents of the accumulations, the one induced by the depletion of class III-a proteins. This phenotype is characterized by abundant foci and dispersed subparticles. However, the lack of NOC4L and RRP12 in those foci indicates that the abortive preribosomes found in this case are post-A1 processomes that are less structured or stable than those formed in cells depleted of class III-a proteins.

Finally, we found that the rest of the depletions (the four RPSs of class II, the two members of class III-b, three RPSs of class IV, and all RPSs of class V) do not generate persistent NOC4L⁺ bystin⁺ RRP12⁺ foci. Instead, they induced variegated patterns of distribution of nucleolar subparticles in the nucleoplasm (Fig. 7A). One possible explanation for the lack of foci in these cases is that the abortive particles are not structured enough to form large aggregates. We do not exclude, however, that the presence of diffuse or aggregated complexes could be underestimated in these cases for several reasons: (i) very dim intrinsic fluorescence that could preclude microscopy detection under our screening conditions; (ii) very transient lifespan of such structures; and/or (iii) lack of proper incorporation of the RBF-GFP reporters used in our study in such structures. Despite this, it is worth noting that even in the case of the knockdown cells lacking aggregates, these cells exhibit abnormally high (>4 fold) levels of mislocalized RBFs (LTV1, NOC4L and/or RRP12) in the nucleoplasm (Fig. 7A). For instance, the DBA-associated RPS10 or RPS26 knockdowns, which both block 40S subunit maturation in the cytoplasm⁴⁰, lead to a 4- and an 8-fold increase in RRP12 and NOC4L levels in the nucleoplasm, respectively (Fig. 7A and Supplementary Fig. 17A, C).

Collectively, these results demonstrate that virtually all RPS deficiencies can disrupt nuclear proteostasis. They also reveal that the accumulation of aggregation-prone preribosomal materials is a predominant feature shared by many deficiencies that, directly or indirectly, alter the normal maturation of 40S particles in the nucleolus. Moreover, we have observed that the depletion of RPS19 or RPS18 causes in each case a particularly severe disruption of the nucleolar–nucleoplasmic proteostasis equilibrium, as inferred from the detection of highly abundant and persistent aggregation-prone particles as well as by the extensive dispersion of subparticle components throughout the nucleoplasm. This pattern is consistent with a more severe 40S synthesis defect induced by the RPS18 or RPS19 knockdowns than the rest of RPS depletions interrogated in our study.

Discussion

In this work, we have used a comprehensive approach to analyze the alterations in preribosome homeostasis caused by ribosomal protein deficiencies. This strategy allowed us to identify specific patterns of RBF-containing foci, as well as dispersed materials that spuriously accumulate in the nucleoplasm rather than in the nucleoli of cells upon depletion of individual RPSs. Preribosome-containing foci had already been noticed before in certain ribosomal protein deficiencies^38,39,54, although their significance remained uncertain because those observations lacked quantitative or systematic characterizations. The nucleoplasmic accumulations of dispersed nucleolar materials had not been previously examined either. Here, we show that the RBF-containing foci resemble the PNBs that form during nucleolar reassembly in late mitosis, both in their accumulation of preribosomes and in their timing of nucleation during the cell cycle. Notably, the foci generated by most RPS deficiencies contain RRP12, bystin, and NOC4L, three proteins that are incorporated into the SSU processome at a stage when all the early RPS proteins had already been assembled and the 18S domains have been folded onto the 5’-ETS particle¹⁰. This suggests that these foci contain large and structurally complex particles. This idea is further supported by our proteomic analysis in RPS19 knockdown cells, showing an enrichment of RRP12 complexed to most SSU processome components. Given that phase separation of preribosomes depends on intrinsically disordered regions (IDRs) within RBFs⁵⁵, the foci might arise because the abortive preribosomes achieve condensation-prone states analogous to those occurring during normal 40S subunit synthesis. However, because these foci lack the dynamic behavior characteristic of PNBs, it is also possible that they are more solid- or semi-solid-like structures rather than liquid-like or gel-like condensates, formed by the self-association of abnormal preribosomes reaching high local concentrations when the nuclear envelope reforms at the end of mitosis. Accordingly, throughout this study, we have employed the general term “aggregate” to allude to these structures, as it does not explicitly define their physical state. As for the case of the dispersed RBF-containing materials observed in the nucleoplasm, they likely represent subparticles generated by the disassembly of abortive preribosomes. This interpretation is supported by the apparent absence of intact pre-rRNA species co-localizing with the dispersed materials in our pre-rRNA FISH assays. Moreover, the specific accumulation patterns differ according to the protein depleted (e.g., NOC4L^– RRP12^– bystin^– LTV1⁺ in class I deficiencies or NOC4L⁺ RRP12⁺ bystin^– LTV1⁺ in class III-a deficiencies), indicating that distinct abortive preribosomes generate specific subparticles.

Our data indicate that the abortive preribosomes generated inside the nucleolus directly by a lack of an individual RPS fall into three major groups that likely reflect differences in their structural stability (see summary cartoon in Fig. 7B). (i) Aggregation-prone 30S-containing particles: these possess high structural stability but do not render stable subparticles (generated upon depletion of class I RPSs). (ii) Structurally-unstable 30S-containing particles: these tend to accumulate as subparticles when released into the nucleoplasm (generated upon depletion of class II RPSs). (iii) Aggregation-prone, highly stable, 21S-containing particles: these generate persistent subparticles (generated upon depletion of class III-a RPSs). The nearly identical RBF accumulation patterns across class I deficiencies likely reflect defects in the stability of the 18S 5’ and central domains, which similarly compromise the formation of a pre-A1 state proficient for 30S pre-rRNA processing. Likewise, the highly similar phenotypes induced by the two Class III-a RPSs (RPS18 and RPS19) suggest the existence of a discrete maturation event during post-A1 state formation that requires both proteins.

Regarding the RPS deficiencies that block cytoplasmic maturation of 40S subunits, we found that these cases appear to generate abortive preribosomes inside the nucleolus with features that are analogous to, although more heterogeneous than, those induced by the depletion of class II or class III-a RPSs (Fig. 7B). This heterogeneity likely reflects variations in the number of early RBFs (e.g., PNO1, RRP12, bystin) that become sequestered within the abortive cytoplasmic particles. Such sequestration can impede the maturation of 26S pre-RNA-containing (as in the case of PNO1 depletion⁵⁰) or 21S pre-RNA-containing (as in the case of bystin or RRP12 depletion^43,49) particles. It is also likely that the combined sequestration of the three early RBFs causes more severe defects in the assembly of processomes than the single loss of each of them, given their essential roles for both the pre-A1 and post-A1 states¹⁰. In the cases of the class IV deficiencies, we propose that the reason for the accumulation of nucleolar materials in the nucleoplasm is an indirect effect of bystin sequestration in the cytoplasm (Fig. 6F, see the steps of release of RBFs boxed in gray rectangles). Notably, five of the eight class IV deficiencies generate foci that are single positive for either RRP12 or NOC4L but never RRP12⁺ NOC4L⁺ at the same time. This suggests that the abortive 21S-containing processomes, unlike those of class III-a, have not fully assembled the structural regions containing bystin, RRP12 or NOC4L. However, those complexes must possess sufficient structural stability to reach high local concentrations and form persistent aggregates and subparticles. As an additional finding of this study, the coherent pattern of phenotypes induced by the depletion of cytoplasmic-maturation RPSs allowed us to reassign specific roles to three proteins (RPS27, RPSA, RPS21) that did not correlate with previously assigned functions in 40S maturation. Evidence supporting the role of these proteins in late cytoplasmic maturation steps includes: (i) strong cytoplasmic retention of bystin-GFP upon their depletion; (ii) high phenotypic similarity between these depletion phenotypes and those involving RPS27a or RPS2; and (iii) the topological proximity of RPS21, RPS2, and RPSA within the mature 40S subunit.

Across all classes, and irrespective of whether the defect is primary or secondary, abortive preribosomes consistently accumulate in the nucleoplasm and cause severe defects in RBF recycling. Notably, the most pronounced nuclear accumulation phenotypes occur when maturation is blocked in particles that have already generated the 21S pre-rRNA. This suggests that the processomes that reached this stage have acquired a highly structured state, rendering subparticles more resistant to cellular dissolution compared to those blocked at earlier stages. This is particularly the case with the RPS18 or RPS19 deficiencies. In the case of RPS19, the high structural stability of the abortive particles is also consistent with the fact that this protein is incorporated in the pathway after the pre-A1* stage, a point at which the processome has already completed major conformational changes required to reach a structurally stable post-A1 state¹⁰.

Although it is expected that siRNA-mediated knockdowns will generate more pronounced phenotypes than RPS haploinsufficiencies, it is worth noting that this experimental approach has previously been quite useful in unveiling mechanisms that were subsequently validated in DBA patient-derived cells and animal models. Those include, for example, the identification of defects in 40S synthesis, reduced protein translation, and p53 activation upon nucleolar stress^{38–40,44,45,56–60}. In this context, our data support the idea that nucleoplasmic aggregates and subparticles, as well as defects in RBF recycling, will likely arise in DBA cells as well. Consistent with this idea, we demonstrated that a partial depletion of RPS19 elicits a significant accumulation of aggregates and diffuse subparticles in the nucleoplasm under conditions in which the defects in 21S pre-rRNA processing are highly similar to those observed in erythroid precursors from some RPS19-deficient DBA patients⁴⁵. We also found a strong accumulation of dispersed particle subcomplexes mixed with either persistent (RPS28, RPS7, RPS15a, RPS24, RPS29, RPS20) or transient (RPS17) aggregates in seven other RPS deficiencies known to cause DBA. In addition, we detected large accumulations of dispersed nucleolar subcomplexes in the two remaining DBA-linked RPS deficiencies analyzed in our study (RPS10 and RPS26), although we did not observe persistent aggregates in these cases. However, we cannot exclude the possibility that, as in the case of RPS17, more transient aggregates that could not be detected by our automated imaging analyses could also be produced under these conditions. This is a plausible option, given that we detected nucleoplasmic foci emitting dim epifluorescence signals in RPS26-knockdown cells using standard confocal microscopy.

Based on the present results, we propose that aberrant preribosome accumulations and subparticles likely contribute to DBA pathogenesis. Indeed, the sequestration of early RBFs within these structures is expected to exacerbate 40S insufficiencies beyond the effect of the RPS haploinsufficiency itself, given that all stalled pre-A1 and post-A1 complexes contain dozens of essential early RBFs¹⁰. Additionally, nucleoplasmic protein aggregates are known to induce genomic instability, proteostasis stress, and nuclear dysfunction^19,61,62. In the context of our study, it is particularly intriguing a recent report showing that mutant versions of nucleophosmin form oncogenic subnucleolar condensates that cause cellular transformation and leukemic expansion through sequestration of export factors and gene-expression regulators⁶³. The perinuclear localization of the preribosome-containing foci described here, together with the sequestration of nucleophosmin, suggests that these abnormal preribosomes might disrupt nucleolar or nucleoplasmic processes other than ribosome synthesis and interfere with nucleocytoplasmic transport. Finally, given that the mechanism of activation of p53 in RPS haploinsufficiencies still remains ill-defined^64,65, it will also be interesting to examine in the future whether the abortive preribosome materials and aggregate structures reported here contribute to the activation of this tumor suppressor protein.

Collectively, our findings demonstrate that disturbances in 40S ribosomal maturation can profoundly disrupt the nucleolus-nucleoplasm proteostasis equilibrium and drive the formation of abnormal nucleoplasmic aggregates of preribosome materials without compromising nucleolar integrity. These observations broaden the concept of nucleolar stress and identify new cellular phenotypes (RBF delocalization, preribosome extranucleolar aggregation and diffuse accumulation of subparticles) that should be incorporated into future analyses of disease-associated mutations or nucleolus-targeting drugs. These phenotypic signatures can also be valuable to illuminate pathogenic pathways that have so far remained elusive in ribosomopathies.

Methods

Cell culture

The HeLa parental cell line was obtained from the American Type Culture Collection (ATCC). This cell line and all the HeLa-derived cell lines employed in the study were cultured in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum, 100 U/ml penicillin/streptomycin, 2 mM L-Glutamine and maintained under standard tissue culture conditions. Cell lines have been routinely tested to ensure they are free of mycoplasma. All culture reagents were obtained from Gibco-Thermo Fisher Scientific.

Generation of CRISPR-Cas9 gene-edited cell lines

In this study, six HeLa-derived cells that endogenously express specific RBFs tagged with GFP were employed. All of them were generated in our laboratory. The construction of the NOC4L-GFP-, GFP-RRP12-, bystin-GFP- and LTV1-GFP-expressing cell lines was already described in a previously published study⁴³. Briefly, for all six cell lines, the editing of each individual locus [NOC4L (gene ID: 79050), RRP12 (gene ID: 23223), BYSTL (gene ID: 705), LTV1 (gene ID: 84946), DDX21 (gene ID: 9188) and NPM (gene ID: 4869)] involved the generation of a single plasmid encoding both Cas9 nuclease and the specific scaffold/guide RNA (sgRNA), and a second plasmid supplying the donor DNA for GFP sequence insertion through homology-directed repair (HDR). To construct the first plasmid, guide RNA sequences were selected using open-access online tools (crispr.mit.edu and benchling.com) that identified protospacer-adjacent motif (PAM) sequences and provided on-target and off-target scores. The sgRNA sequences were cloned into the pX330 plasmid (Cat. No. #42230, Addgene) using synthetic oligonucleotides (listed in table in Supplementary Table 1 included in Supplementary Information file) that were annealed and ligated into the BbsI-digested vector.

The second plasmid was generated by inserting the HDR donor sequences into a cloning vector. The DNA donor comprised the GFP sequence flanked by two DNA segments of approximately 1000 base pairs each, which were homologous to the genomic region to be edited. For construction of the HDR plasmids, two DNA fragments carrying sequences homologous to both sides of the genome editing site were ligated, upstream and downstream, to the cDNA sequence encoding GFP. The fragments were generated by PCR from HeLa genomic DNA. The HDR plasmids used to generate the NOC4L-GFP-, GFP-RRP12-, bystin-GFP- and LTV1-GFP- expressing cells have been previously described⁴³. Oligonucleotides used to generate the NPM-GFP and DDX21-GFP HDR plasmids are listed in Supplementary Table 1 (included in Supplementary Information file). For the NPM-GFP HDR plasmid (pBN50), the left-arm and right-arm fragments were cloned into pEGFP-N3 (Clontech) at the PstI-BamHI and BsrGI-MfeI sites, respectively, thereby creating a fusion of the last codon of NPM and the first codon of GFP. Subsequently, the left-arm, GFP, and right-arm fragments were excised and cloned into the pBlueScript plasmid (Stratagene) at the XhoI-EcoRI sites. Finally, a silent mutation in the PAM sequence was introduced to avoid cleavage of the repair template by Cas9. For the DDX21-GFP (pJE4) HDR plasmid, the left and right-arm fragments were introduced into a pBlueScript plasmid that already contained the GFP sequence, which was flanked by two multicloning sites (pJG1). The right-arm fragment was cloned at the XhoI-NheI sites, and the left-arm fragment at the EcoRI-BamHI. A list of all plasmids generated and used in this study can be found in Supplementary Table 2 (included in Supplementary Information file).

To perform the GFP knock-in by CRISPR-Cas9, HeLa cells were transfected with 1-3 mg of a mixture of Cas9/sgRNA and HDR donor plasmid (1:2 molar ratio) using the JetPEI transfection reagent (Cat. No. 101-10 N, Polyplus). The cells were re-transfected with the HDR donor plasmid 24 h later and sorted based on GFP fluorescence intensity 4–5 days after the second transfection. Individual GFP-positive cell clones were isolated, expanded, and analyzed by western blot and PCR to select those clones carrying the GFP knock-in modification in all the alleles of the targeted locus.

siRNA-mediated transcript knockdown

For most of the siRNA-mediated transcript knockdown experiments, 10 nM of siRNA duplexes (listed in Supplementary Table 3 included in Supplementary Information file) targeting individual transcripts were reverse transfected in GFP-tagged RBF-expressing HeLa cells using 8 µg/ml of Lipofectamine RNAiMAX (Cat. No. 13778150, Invitrogen-Thermo Fisher Scientific) in 500 µl Opti-MEM (Cat. No. 31985070, Gibco-Thermo Fisher Scientific) for 15 minutes in 6-well plates. Cells were harvested or fixed 24, 48 or 72 h after transfection, as indicated in the figures. The siRNA duplexes were purchased from Ambion-Thermo Fisher Scientific (Silencer Select siRNA, Cat. No. 4390827, Invitrogen-Thermo Fisher Scientific). Negative controls were either untreated (Untr.) cells or cells transfected with control scrambled (si-ctrl) siRNA (Cat. No. 4390844, Ambion-Thermo Fisher Scientific).

For the systematic image-based analyses, 20 nM of Dharmacon ON-TARGETplus™ siRNA pools (Horizon Discovery) targeting each RPS were complexed with 8 µl/ml Lipofectamine RNAiMAX in 10 µl Opti-MEM for 10 min and pipetted in quadruplicate arrays on 384-well CellCarrier optical bottom plates (Cat. No. 6057302, Revvity). Each single RPS transcript was targeted by four independent siRNA duplexes (list of all siRNAs is in the table in Supplementary Data 1, provided as a separate Excel file). Non-targeting or ECT2-targeting siRNA duplexes were equally arrayed as mock and penetrance controls, respectively. HeLa cells endogenously expressing GFP-tagged NOC4L, RRP12, bystin or LTV1 were reverse transfected by plating them on top of transfection mixtures at 1800 cells/well in 40 µl of total volume. At 24 h after transfection, cells were fixed and processed for immunofluorescence.

Northern blot analyses

Total cellular RNA was prepared using the Trizol method with TRI-Reagent (Cat. No. AM9738, Ambion-Thermo Fisher Scientific), in accordance with the manufacturer’s guidelines. A total of 4 µg of RNA was incubated with three volumes of HCHO-RNA buffer (1X MOPS, 6.5% formaldehyde [Cat. No. F8775-500ML, Merck], 50% formamide [Cat. No. F9037, Merck]) for 12 minutes at 65 °C for RNA denaturation and then mixed with one volume of loading buffer (50% glycerol, 1 mM EDTA [pH 8.0], 0.01% xylene cyanol, 0.01% bromophenol blue and 125 µg/ml ethidium bromide [Cat. No. E1510, Merck]). The RNA was subjected to electrophoresis on 1.2% agarose/formaldehyde gels in 1x MOPS buffer. Subsequently, the gels were immersed in an alkaline solution (10 mM NaOH, 3 mM NaCl) for 20 min at room temperature, rinsed with H₂O three times, and incubated with 20X saline-sodium citrate (SSC) buffer (3 M NaCl, 300 mM sodium citrate) for 45 min at room temperature. Then, RNAs were transferred to Amersham Hybond N+ nylon membranes (Cat. No. 10024484, GE Healthcare) by capillarity with 20X SCC transfer. After 48 hours of transfer, ultraviolet (UV) cross-linking was conducted to covalently bind the RNAs to the membranes. Finally, the membranes were pre-hybridized with Church and Gilbert’s buffer (1%BSA, 1 mM EDTA, 0.5 M sodium phosphate buffer [pH 7.1], 7% SDS) and hybridized with ³²P-labeled oligonucleotide probes (see Supplementary Table 1 in Supplementary Information file) in the same buffer. Filters were then washed for 20 min with two different solutions (Solution A [2X SCC, 0.5% SDS] and Solution B [0.5X SCC, 0.05% SDS]) consecutively, after which they were exposed to X-ray films at −80 °C. To assess the 28S/18S rRNA ratio in cells after individual depletions, aliquots of total RNA were analyzed using an Agilent 2100 Bioanalyzer™. Quantification of 21S pre-rRNA species was performed using Fiji, with results expressed as ratios relative to 30S pre-rRNA.

Preparation of cell lysates

For total cellular lysate preparation, cells were transfected and plated in 6-well plates until the indicated time point. They were then washed three times with phosphate-buffer saline (PBS) solution and lysed chemo-mechanically in 200 μl of RIPA-C buffer (25 mM Tris-HCl [pH 7,6], 150 mM NaCl, 1% NP-40, 1% Triton X-100, 1% sodium deoxycholate, 0.1% Sodium Dodecyl Sulfate (SDS), 1 mM NaF, 1 mM Na₃VO₄, 1 mM s-glycerophosphate and Cømplete protease inhibitor cocktail tablet [Cat. No. 05056489001, Roche]). The extracts were precleared by centrifugation at 20,000 x g for 10 min at 4 °C, and the protein concentration was determined with the Pierce BCA Protein Assay Kit (Cat. No. 23225, Thermo Fisher Scientific) following the manufacturer’s guidelines. The SN1, SN2 and SN3 fractions were prepared using the Preribosome Sequential Extraction (PSE) method, as described in a previously published step-by-step protocol⁴⁶. Briefly, cells from two 10-cm dishes (80% confluency) were harvested in ice-cold PBS, pooled, pelleted, and frozen in liquid nitrogen. The cell pellets were resuspended thoroughly in 1 ml of SN1 buffer (20 mM HEPES-NaOH [pH 7.5], 130 mM KCl, 10 mM MgCl₂, 0.05% Igepal CA-630), supplemented with Cømplete protease inhibitor cocktail, followed by centrifugation in a microfuge (1300 × g for 3 min at 4 °C). The resulting supernatant was collected and stored as the SN1 fraction. The pellet was then washed with 1 ml SN1 buffer, after which the resulting pellet was resuspended in 600 μl of SN2 buffer (10 mM HEPES-NaOH [pH 7.5], 10 mM NaCl, 5 mM MgCl₂, 0.1% Igepal CA-630, 0.5 mg/ml heparin, 600 U/ml RNasin (Promega)] supplemented with 100 U RNase-free DNase I (Qiagen), and incubated for 10 min at room temperature with gentle mixing. The lysate was centrifuged (12,300 × g for 10 min at 4 °C) and the supernatant was collected as the SN2 fraction. The remaining pellet was resuspended in 800 μl of SN3 buffer (20 mM HEPES-NaOH [pH 7.5], 200 mM NaCl, 4 mM EDTA, 0.1% Igepal CA-630, 0.04% sodium deoxycholate, 4 mM imidazole, 0.1 mg/ml heparin, 1 mM dithiothreitol (DTT), Cømplete, 600 U/ml RNasin) and incubated for 20 min at room temperature with moderate agitation. The extract was centrifuged (12,300 × g, 10 min, 4 °C) and the supernatant was collected as the SN3 fraction. Equivalent volumes of each fraction (15 µl of SN1, 9 µl of SN2 and 12 µl of SN3) were analyzed by western blot to visualize the total protein content (TCL) of each fraction.

GFP-Trap pull-down assays

Two 15-cm or four 10-cm plates of BYSL-GFP- or GFP-RRP12- expressing HeLa cells were used for each condition to prepare SN1, SN2 and SN3 extract fractions with the PSE method. The obtained fractions were pre-incubated with 25 μl of binding control agarose beads (Cat. No. bab-20, Chromotek) for 1 h at 4 °C in a rotating wheel to remove material that non-specifically binds to GFP-Trap beads. After centrifugation, each supernatant was transferred to a clean tube and incubated with 15 µl of GFP-Trap beads (Cat. No. gta-20, Chromotek) for 2 h at 4 °C. Beads were then washed five times with the corresponding SN1, SN2 or SN3 buffer. The whole sample of purified material was resuspended in SDS-Page loading buffer (SPLB) and analyzed, in parallel with the total-cell protein (TCL) samples, by western blot.

Western blot analyses

Protein extracts were separated by electrophoresis in acrylamide gels and transferred onto nitrocellulose filters using the iBlot Dry Blotting System (Thermo Fisher Scientific). Membranes were blocked in 5% dry milk in TBS (25 mM Tris-HCl [pH 8.0], 150 mM NaCl) for at least 30 minutes and then incubated overnight with the appropriate primary antibodies (Supplementary Table 4 in Supplementary Information file) at 4 °C. After three washes with TBS-T (25 mM Tris-HCl [pH 8.0], 150 mM NaCl, 0.1%Tween-20), membranes were incubated with secondary antibodies (Supplementary Table 4) for 45 min at room temperature. Then, membranes were washed three times with TBS-T and immunoreacting bands were developed by the chemiluminescent method using Pierce ECL Western Blotting Substrate (Cat. No. #RPN2106, Cytiva – Amersham).

Preribosome purification and mass spectrometry analyses

Nucleolar preribosomes for mass spectrometry analyses were prepared following the method developed by the laboratory of Sebastian Klinge¹⁰. A total of four independent samples were analyzed, comprising two distinct biological replicates for the control condition and two distinct biological replicates for the si-RPS19-transfected condition. For each individual sample, twelve 10-cm plates of HeLa cells endogenously expressing GFP-RRP12 were cultured to approximately 80% confluency per condition (control and si-RPS19-transfected). To obtain two purifications per sample, cells were processed in parallel sets of six 10-cm plates. Cells were washed twice with ice-cold PBS, harvested by scraping, and pooled into 15 ml tubes. The cell suspension was pelleted by centrifugation in a benchtop centrifuge (1000 rpm, 5 min, 4 °C), washed with ice-cold PBS, transferred to a microcentrifuge tube, and pelleted again (2500 rpm, 5 min, 4 °C). The resulting pellets were snap-frozen in liquid nitrogen and stored at −80 °C. For preribosome extraction, frozen pellets were resuspended in 500 µl Basic Buffer (25 mM HEPES pH 7.6, 65 mM NaCl, 65 mM KCl, 10 mM MgCl₂, 1 mM EDTA, 10% glycerol, 0.05% Triton X-100, 1 mM DTT, 0.5 mM PMSF, 1x Complete protease inhibitor) to solubilize the cytoplasmic fraction, followed by centrifugation at maximum speed for 3 min at 4 °C. After discarding the supernatant, the pellet was washed once with Basic Buffer and resuspended in 500 µl Active Buffer (25 mM HEPES pH 7.6, 10 mM KCl, 10 mM MgCl₂, 1 mM CaCl₂, 1 mM EDTA, 1 mM spermidine, 5% glycerol, 0.1% Triton X-100, 100 mM arginine, 25 mM ATP, 1 mM DTT, 0.5 mM PMSF, 1X Complete protease inhibitor, 8000 U DNase I). Nucleolar preribosomes were extracted by incubating the suspension for 20 min at room temperature on a rocking shaker, followed by centrifugation at maximum speed for 20 min at 4 °C. The resulting supernatant was collected, diluted with 500 µl Active Buffer, and subjected to GFP-Trap purification. To minimize non-specific binding, lysates were pre-cleared by incubation for 1 h at 4 °C with equilibrated binding control agarose beads (Chromotek, Cat. No. bab-20). Lysates were then incubated for 2 h at 4 °C on a rotating wheel with 15 µl equilibrated GFP-Trap beads (Chromotek, Cat. No. gta-20). Beads were washed twice with Wash Buffer 1 (25 mM HEPES pH 7.6, 75 mM NaCl, 75 mM KCl, 5 mM MgCl₂, 0.5 mM EDTA, 2.5% glycerol, 0.1% Triton X-100, 100 mM arginine, 1 mM DTT, 0.5 mM PMSF, 1x Complete protease inhibitor) and twice with Wash Buffer 2 (25 mM HEPES pH 7.6, 75 mM NaCl, 75 mM KCl, 5 mM MgCl₂, 0.5 mM EDTA, 2.5% glycerol, 100 mM arginine).

For mass spectrometry analyses, the beads from two parallel purifications per sample were combined and washed twice with arginine-free Wash Buffer 2 before on-bead proteolytic digestion. Protein identification was performed at the Proteomics Unit of the Centro de Investigación del Cáncer following standard procedures. For on-bead proteolytic digestion, protein-bound beads were washed four times with a buffer containing 25 mM HEPES (pH 7.6), 75 mM NaCl, 75 mM KCl, and 5 mM MgCl₂. Simultaneous disulfide bond reduction and alkylation were performed by adding tris(2-carboxyethyl) phosphine (TCEP) and 2-chloroacetamide (CAA) to final concentrations of 5 mM and 20 mM, respectively, followed by incubation at 95 °C for 10 min. After cooling to room temperature, sequential proteolysis was carried out: first, LysC was added at an enzyme-to-protein ratio of 1:100 (w/w) and incubated for 3 h at 37 °C with shaking at 700 rpm. Subsequently, sequencing-grade modified trypsin (Promega) was added at a 1:50 (w/w) ratio and incubated overnight at 37 °C under continuous shaking at 700 rpm. Samples were briefly centrifuged at 600 x g; the peptide-containing supernatant was recovered in a new Protein LoBind tube (Eppendorf) and acidified with 1% formic acid (FA). Finally, the samples were desalted using in-house-made StageTips packed with C18 Empore disks (3 M) and vacuum-dried prior to LC-MS/MS analysis.

A nano-UHPLC system (nanoElute, Bruker Daltonics) coupled to a timsTOF Pro hybrid trapped ion mobility-quadrupole time-of-flight mass spectrometer (Bruker Daltonics) via a CaptiveSpray nano-electrospray ion source (Bruker Daltonics) was used for reversed-phase LC-MS/MS analysis. Mobile phase A consisted of 0.1% FA in water, and mobile phase B consisted of 0.1% FA in acetonitrile (ACN). Peptides were dissolved in 0.1% FA and 2% ACN, and a total of 2 µl/150 ng of sample was loaded onto a trapping cartridge (PepMap, Thermo Fisher Scientific), and separated on a 75 µm ID x 25 cm, 1.9 µm C18 Aurora Series column (IonOpticks) at 50 °C. The separation was performed at a flow rate of 300 nl/minute using a 125-minute total run profile encompassing a 100-minute gradient: from 2 to 25% B over 90 min, followed by 25 to 37% B over 10 min, and finally reaching 95% B over 10 min. The column was washed at 95% B for 10 min and re-equilibrated at 2% B for 5 min.

MS data were acquired in Parallel Accumulation-Serial Fragmentation (PASEF) data-dependent acquisition (DDA) mode⁶⁶. The source parameters were set as follows: capillary voltage, 1500 V; dry gas temperature, 180 °C; and dry gas flow, 3 l/min. Each acquisition cycle consisted of one full MS scan and 10 PASEF MS/MS frames, resulting in a total cycle time of 1.1 s. The MS and MS/MS spectra were recorded in the m/z range of 100–1700. The ion mobility (1/K₀) range was set to 0.6–1.6 V·s/cm² with an accumulation time of 100 milliseconds per frame. A polygon filter was applied in the m/z and ion mobility plane to exclude singly charged ions from PASEF precursor selection. For MS/MS, an intensity threshold of 2500 and a target intensity of 20,000 were used, with a charge state range of 2–5. Collision energy was stepped linearly as a function of mobility from 20 eV at 1/K₀ = 0.6 V·s/cm² to 59 eV at 1/K₀ = 1.6 V·s/cm².

Raw data files (.d) were processed using FragPipe (v23.1) integrated with the MSFragger search engine (v4.1)⁶⁷. MS/MS spectra were searched against the Homo sapiens UniProtKB/Swiss-Prot protein sequence database (downloaded on 2026-03-26, UP000005640), which included reversed decoy sequences and a common contaminants list. Search parameters were set as follows: trypsin as the proteolytic enzyme with up to two missed cleavages allowed; carbamidomethylation of cysteine as a fixed modification; and oxidation of methionine and N-terminal acetylation as variable modifications. Both precursor and fragment mass tolerances were set to 20 ppm and the minimum peptide length required for identification was set to seven amino acids. For data validation, protein and peptide identifications were filtered at a strict 1% false discovery rate (FDR) at both the peptide-spectrum match (PSM) and protein level using the Philosopher module⁶⁸. Reverse (decoy) hits and common contaminants were strictly excluded from the dataset prior to quantification. Functional data annotation was comprehensively retrieved directly from the UniProt protein knowledgebase.

To estimate the relative abundance of proteins, iBAQ (intensity-based absolute quantification) values were calculated within FragPipe using the integrated IonQuant module (v1.10.12)69⁶⁹. Each protein’s total intensity was divided by the number of its theoretically observable tryptic peptides (between 7 and 30 amino acids in length), requiring a minimum of 1 unique peptide for identification. This normalization allowed ranking of proteins by molar abundance, generated independently for each LC-MS/MS biological replicate to enable downstream comparative analysis. The abundance of co-purifying RBFs was determined using these calculated iBAQ values normalized directly to the amount of bait protein (GFP-RRP12) detected in each corresponding sample. The mass spectrometry data are accessible at the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD077805.

Confocal microscopy analyses

For immunofluorescence analyses, untreated or siRNA-transfected cells were fixed in PBS containing 4% paraformaldehyde (Cat. No. 1040021000, Merck) for 15 minutes and permeabilized for 10 min in 0.5% Triton TX-100 (Cat. No. X-100, Merck) in TBS-T. Cells were blocked with 2% bovine serum albumin in TBS-T for at least 30 min and then incubated with primary antibody (Supplementary Table 4 in Supplementary Information file) for 2 h at room temperature. Finally, after four washes with TBS-T, cells were incubated with secondary antibody (Supplementary Table 4) for 45 min, and nuclei were stained with DAPI (4’6-diamidino-2-phenylindole) (Cat. No. D1306, Thermo Fisher Scientific) before being mounted with Mowiol. Cells endogenously expressing GFP-tagged RBFs, transfected with the appropriate siRNA, were fixed with 4% paraformaldehyde in PBS for 15 min, permeabilized for 10 minutes with 0.5% Triton TX-100 in TBS-T, and nuclei were stained TO-PRO-3 (Cat. No. T3605, Thermo Fisher Scientific), as indicated in the figures. Preparations were mounted onto Mowiol prior to microscopy observation. Imaging was performed on a Leica TCS SP8 X (Leica Microsystems) confocal microscope, driven by the LAS-XTM version 3.1.5 16308 software, using a x63/1.4 oil immersion optical lens (HC PL APO SC2) (optical section: 0.896 μm). GFP-Alexa Fluor 488, Alexa Fluor 594, DAPI, and TO-PRO-3 channels were excited at 488, 594, 405 and 641 nm, respectively. For each reporter cell line, images were acquired using the same laser power and detector-gain conditions. No manipulations were made other than the brightness/contrast adjustments performed in some cases to improve the visualization of specific features (indicated in figure legends).

Fluorescence in situ hybridization (FISH)

For fluorescence in situ hybridization (FISH) experiments, parental or GFP-RRP12-expressing HeLa cells were seeded directly onto poly-L-lysine-coated coverslips and incubated overnight to allow cell attachment. Cells were fixed in 4% paraformaldehyde (PFA) (Cat. No. 1040021000, Merck) in PBS for 30 min at room temperature, followed by two washes in PBS. Cells were then permeabilized in 70% ethanol at 4 °C for 18 h, and subsequently washed twice with 2X SSC containing 10% formamide (Cat No. 4650-OP, Merck). For hybridization, 5’-fluorescent-labeled probes (purchased from Isogen; listed in Supplementary Table 1 in Supplementary Information file) were diluted in freshly prepared hybridization buffer (10% dextran sulfate (Cat No. 9011-18-1, Merck), 2X SSC, 10% formamide, 0.5 µg/ml tRNA (Cat No. 9011-18-1, Thermo Fisher Scientific), 250 µg/ml BSA, 10 mM ribonucleoside vanadyl complexes (Cat No. R3380, Merck), and 1 ng/µl probe), and applied as drops (40–50 µl) to the cells, and coverslips were incubated overnight at 37 °C in a humidified chamber in the dark. Following hybridization, coverslips were washed once for 30 min at 37 °C with 2X SSC containing 10% formamide, and once with PBS. Nuclear DNA was stained with DAPI (1:10,000 in PBS; Cat. No. D1306, Thermo Fisher Scientific) for 3–5 min at room temperature in the dark. Coverslips were then mounted on microscope slides using ProLong™ Glass Antifade Mountant (Cat. No. P36984, Thermo Fisher Scientific) and stored in the dark at 4 °C until imaging. Imaging was performed using a Leica TCS SP8 X confocal microscope operated with LAS-X v3.1.5 software, equipped with a 63x/1.4 oil immersion objective (HC PL APO 63×/1.4 Oil CS2). Fluorophores were excited using a white light laser at appropriate wavelengths.

Time-lapse live-cell imaging

Cultures were enriched in G1/S-blocked cells by incubation with 2.5 mM thymidine (Cat No. T9250, Merck) in DMEM complete culture medium for 24 h. Following the thymidine treatment, cells were washed thoroughly with PBS and transfected with the indicated siRNAs on chambered 4-well slides (Cat No. 80426, Ibidi). 7 h after transfection, SiR-DNA (Cat No. SC007, Spirochrome) was added to the culture medium at a final concentration of 0.25 µM for DNA staining. Cells were then transferred to a Leica TCS SP8 confocal microscope equipped with an environmental chamber maintaining 37 °C and 5% CO₂. Live-cell imaging was performed using a 63× oil immersion objective. Time-lapse acquisition was initiated immediately, and images were collected every 10 min over an 18 h period, using excitation at 488 nm (GFP channel) and 640 nm (far-red channel for DNA). Time-lapse movies were generated in Fiji by importing image stacks directly from Leica files. Images were displayed in grayscale using a binary mask and exported as MP4 files at five frames per second, using a 16:9 canvas without rescaling the original image content

Image analysis and quantification

Quantification of GFP-RRP12- and pre-RNA-containing nucleoplasmic foci was performed using a custom macro developed in ImageJ/Fiji (NIH). Multichannel 16-bit TIFF images from FISH experiments were analyzed in batch. For each image, fluorescence channels were split into DAPI (blue), GFP or Alexa Fluor 488 (green), and Cy3 or Cy5 (red). The green channel was used to detect RRP12-GFP or AF488-labeled 5’ITS1 probes; the red channel for Cy3- or Cy5-labeled probes (5’ITS1 or ETS1). Colocalization between green and red signals was interpreted as yellow. Binary masks were generated separately for green and red channels using manual thresholding. Pre-nucleolar bodies (PNBs) were detected within each cell nucleoplasm, which was segmented based on DAPI staining. Only fluorescent signals within the DAPI-positive areas were included for quantification. To reduce background noise and exclude nucleoli, a size filter was applied such that only particles with areas between 0.0007 and 0.015 in² were considered PNBs. For each nucleus, the number of foci was counted independently in the green, red, and colocalized (yellow) channels.

For time-lapse datasets, the acquired images were subjected to maximum intensity projection along the Z-axis for each time point in the green fluorescent channel. Only cells that underwent a cell division after release from arrest and protein depletion were included. Time zero (t = 0) was defined as the frame in which the selected cell was in late anaphase or telophase. Foci were segmented using manual thresholding and manually refined to exclude background noise and large nucleolar structures. For each cell, both the number and areas of PNB-like foci were quantified. Data were subsequently imported into GraphPad Prism (version 8; GraphPad Software) for graphical representation and statistical analysis. Multiple comparisons were performed using Bonferroni’s test. Adjusted p-values are indicated in the figures. The supplementary time-lapse movies were generated in Fiji by importing image stacks directly from Leica files. Images were displayed in grayscale using a binary mask and exported as MP4 files at 5 frames per second, using a 16:9 canvas without rescaling the original image content. All raw data of image quantifications are provided in the Source Data file.

Quantitative reverse transcription polymerase chain reactions

Total RNA was extracted and purified from siRNA-transfected cells using RNeasy Mini kit (Cat. No. 74104, Qiagen) according to the manufacturer’s protocol. mRNA abundance was analyzed by RT-qPCR employing the iScript One-Step RT-PCR kit with SYBR Green (Cat. No. 4389986, Applied BioSystems) and the StepOnePlus Real-Time PCR System (Cat. No. 4376600, Applied BioSystems). The abundance of the GAPDH mRNA was used as an internal normalization control. Primers used for transcript quantification are listed in Supplementary Table 1 (included in Supplementary Information file).

Automated image analyses

A total of 32 fields were captured per condition on an HCS Opera cell::explorer station (PerkinElmer) equipped with 405, 488, 561 and 633 laser lines, with a 20X water immersion objective at maximum 200 ms exposure. The whole screen set for all four reporters was run independently twice. Each siRNA was examined through four internal replicates, with inspection of 32 independent fields, and an average of 1200 cells analyzed. The experimental setup exhibited high reproducibility across different features for well- and plate-replicates as well as high penetrance of the phenotypes under a given condition (examples shown in Supplementary Fig. 15). All raw microscopy images acquired automatically with the Opera platform are available at the Zenodo repository (10.5281/zenodo.17952029).

The analysis pipeline, running on the Acapella platform (v. 2.4.1., PerkinElmer), is available at the Zenodo repository (10.5281/zenodo.21100904). Throughout the analysis, images were used exactly as collected and were not subjected to any post-acquisition processing or editing. Briefly, nuclei were first segmented and filtered for size, roundness and brightness to remove potential artifacts and mitotic cells, together with border cells with incomplete nuclei images (see examples in the Supplementary Information file). Subcellular regions of interest (nucleus, perinuclear cytoplasmic region) were delineated for averaged intensity measurements in either EGFP or fibrillarin channels. Bright EGFP and fibrillarin structures were segmented separately from the nucleoplasmic background (see examples in the Supplementary Information file). Described features (see Fig. 5) were extracted both at single-cell resolution (for penetrance and distribution assessment) and at the well average. Average Z-scores of each feature of interest across its 4 well replicates were computed for each siRNA as normalized to the average control mean and standard deviation.

Hierarchical clustering shown in Fig. 5A was computed with average Z-scores for all features, except for feature 3 (presence of diffuse LTV1 in nucleoplasm), which was scaled to a 0-3 range, using open-source MORPHEUS™ software (https://software.broadinstitute.org/morpheus). From this information, a phenotype similarity matrix was computed among all siRNA treatments; correlations above 0.6 were used as edges to depict this information as a network with Cytoscape 3.0 (Supplementary Fig. 16A). An analogous procedure was used to infer a similarity matrix among all analyzed features (Supplementary Fig. 16B).

Reporting summary

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

Author contributions

S.G.G. performed all systematic siRNA knockdown experiments for the automated-imaging microscopy experiments, confocal microscopy analyses to replicate and validate data, immunofluorescence and FISH assays, time-lapse microscopy, generated the GFP-DDX21-expressing cell line and contributed to the design of the project, analyses of results and writing of the manuscript. M.S.-A. set up the automated-image acquisition procedures, monitored the replicability and confidence parameters, generated the script for subcellular segmentations, performed all hierarchical, network and quantitative image analyses and contributed to the analysis of the results. R.M.R.-C. carried out most of the final microscopy analysis shown in the first part of the study and performed siRNA validation, western-blot and northern-blot experiments. B.N. performed the PSE fractionation/GFP-Trap pull-down experiments, carried out some initial microscopy analysis not shown in the final manuscript and generated the NPM-GFP-expressing cell line. M.F. helped with immunofluorescence assays and preribosome purifications for MS analyses. M.A.d.P. contributed to the design of the automated imaging microscopy analyses. X.R.B. contributed to the analyses of data, the discussion of results and the writing of the paper. M.D. designed the project, analyzed the data and wrote the paper.

Peer review

Peer review information

Nature Communications thanks John Woolford and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

M.D. research has been funded by grants (PID2020-118378GB-I00 and PID2023-152015NB-I00) co-funded by MCIN/AEI/10.13039/ 501100011033/plus the European Research Development Fund “A way of making Europe” of the European Union. The X.R.B.’s project leading to these results has received funding from the Spanish Association against Cancer (GC16173472GARC, RETOS245779LLOV), the Castilla-León government (CSI018P23), the MCIN/AEI/10.13039/501100011033/ plus the European Research Development Fund «A way of making Europe» of the European Union (PID2021-122666OB-I00, PLEC2022-009217, PID2024-156980OB-I00), and the «Programa Excelencia» of the Fundación Científica AECC 2022 (EPAEC222641CICS). The activities of the Centro de Investigación del Cáncer of Salamanca are supported by the «Escalera de Excelencia» of the Education Ministry of the Castilla y León autonomous government plus the European Research Development Fund (CLU-2023-2-01), the CSIC-MaX program, and the AECC-Centers Program. MS-A is recipient of support from the MICIN (RYC2020-029690-I, PID2021-128106NA-I00) and grant CNS2023-14483 issued by MICIU/AEI/10.13039/501100011033 and European Union NextGeneration EU/PRTR. M.A.d.P. research was supported by grants from the MICIN/AEI/European Research Development Fund (ARDF/FEDER) “A way of making Europe” (PID2020-118658RB-I00 and PDC2021-121572-I00), and Fundación Obra Social La Caixa (AtheroConvergence, HR20-00075. The CNIC is supported by the Instituto de Salud Carlos III (ISCIII), the Ministerio de Ciencia e Innovación (MCIN) and the Pro CNIC Foundation. It is also a Severo Ochoa Center of Excellence (grant CEX2020-001041-S funded by MICIN/AEI/10.13039/501100011033).

Data availability

All raw images associated with the automated microscopy analysis of the NOC4L-GFP-, bystin-GFP-, GFP-RRP12- and LTV1-GFP-expressing cell lines transfected with 41 ribosomal protein siRNAs have been deposited in the Zenodo repository under 10.5281/zenodo.17952029. The raw data associated to mass spectrometry analyses of RRP12-containing complexes in control and RPS19 knockdown cells have been deposited in the PRIDE database under the dataset identifier PXD077805. The rest of raw data are provided in the Source Data file. Source data are provided with this paper.

Code availability

The custom analysis pipeline and script developed for the high-throughput automated microscopy image processing run on the Acapella platform (v. 2.4.1) have been deposited in the Zenodo repository under 10.5281/zenodo.21100904.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-75682-6.