Abstract
Nucleoporins (Nups) form the selective permeability barrier that separates the nucleoplasm from the cytoplasm. In addition to their localization at the nuclear envelope, Nups have been observed in cytoplasmic foci in many cell types. We have investigated the origin of Nup foci using C. elegans oocytes, which naturally accumulate high concentrations of Nups. We find that the foci derive from condensation of highly cohesive FG-Nups, which are maintained at concentrations right above the solubility limit in oocytes. Nup solubility is enhanced by chaperone activity and posttranslational modifications that also promote nuclear pore channel fluidity and pore disassembly during mitosis. Oocyte Nup foci dissolve during M phase and are not essential for embryonic viability. Overexpression of the highly cohesive FG-Nup, Nup98, in post-mitotic neurons leads to uncontrolled Nup aggregation and organismal paralysis, underscoring the importance of mechanisms that maintain Nup solubility in the cytoplasm.
Introduction
In all eukaryotes, the double-membraned nuclear envelope partitions the nucleoplasm from the cytoplasm and material is exchanged between the two compartments by way of nuclear pore complexes. Pore complexes are composed of at least 30 distinct nucleoporins (Nups) arranged in biochemically stable subcomplexes (Figure 1A) (Cohen-Fix and Askjaer, 2017; Hampoelz et al., 2019a). Approximately two-thirds of Nups are essential to scaffold and anchor pore complexes to the nuclear envelope. The remaining one-third contain large phenylalanine/glycine (FG) rich domains that are highly intrinsically disordered. FG-Nups are enriched in the central channel of the pore and readily from multivalent interactions both in vivo and in vitro (Frey et al., 2006; Labokha et al., 2012; Patel et al., 2007; Xu and Powers, 2013). Cohesive interactions among FG-Nups are critical for the formation of the permeability barrier and FG-Nup hydrogels recapitulate nuclear pore selectivity in vitro (Frey and Görlich, 2007; Hülsmann et al., 2012; Ng et al., 2021; Schmidt and Görlich, 2015; Strawn et al., 2004). This has led to the “selective phase” model in which the permeability barrier is established by interactions among FG-Nups that form a phase separated network (Ribbeck and Görlich, 2001; Schmidt and Görlich, 2016).
The intrinsic propensity of FG domains to form multivalent networks presents a potential danger if allowed to occur in an uncontrolled manner. Several mechanisms likely act to prevent aggregation of FG domains within the central channel of the pore. FG-Nups are extensively O-GlcNAcylated and this modification enhances Nup solubility in vitro (Labokha et al., 2012; Schmidt and Görlich, 2015) and has been proposed to relax FG domain interactions within the central channel (Ruba and Yang, 2016; Yoo and Mitchison, 2021). In the selective phase model, nuclear transport receptors (NTRs) cross the permeability barrier by binding and locally disrupting FG domain interactions (Schmidt and Görlich, 2016). NTRs have been reported to “chaperone” diverse aggregation-prone proteins including Nups (Guo et al., 2018; Harel et al., 2003; Hofweber et al., 2018; Hutten et al., 2020; Nachury et al., 2001; Padavannil et al., 2019; Walther et al., 2003), raising the possibility that NTR binding enhances FG-Nup solubility. Finally, phosphorylation plays a major role in pore complex disassembly during M phase (Kutay et al., 2021), implicating this modification in preventing interactions among Nups.
In addition to their localization at the nuclear envelope, Nups have been observed in discrete cytoplasmic foci in virtually all cell types (Cordes et al., 1996; Raghunayakula et al., 2015; Ren et al., 2019). Based on early electron microscopy studies these structures were proposed to represent annulate lamellae, a specialized subdomain of the endoplasmic reticulum rich in pore complexes (Kessel, 1989). Prior studies have proposed that annulate lamellae may act as stores of ready-made pore complexes to supplement the nuclear envelope during rapid cell divisions (Hampoelz et al., 2016; Ren et al., 2019), though the exact function of these structures remains unclear (Stafstrom and Staehelin, 1984). Cytoplasmic Nup foci have additionally been proposed to correspond to structures distinct from annulate lamellae that function in nuclear pore inheritance (Colombi et al., 2013) or pore biogenesis in the cytoplasm (Hampoelz et al., 2019b). Nups are also frequently enriched in pathological cytoplasmic inclusions that are hallmarks of neurodegenerative disease (Chandra and Lusk, 2022; Fallini et al., 2020; Hutten and Dormann, 2020), leading to the proposal that Nups become sequestered and depleted from nuclear pores under disease conditions (Gasset-Rosa et al., 2019; Zhang et al., 2018). Condensation of FG-Nup fusion oncogenes contributes to certain cancers (Chandra et al., 2022; Terlecki-Zaniewicz et al., 2021; Zhou and Yang, 2014), and recent studies have found that cytoplasmic FG-Nups drive aggregation of TDP-43 in ALS/FTLD and following traumatic brain injury (Anderson et al., 2021; Gleixner et al., 2022). Together, these observations raise the possibility that ectopic condensation of FG-Nups may drive protein aggregation and disease progression.
Here we use C. elegans oocytes, which have abundant, endogenous Nup foci (Patterson et al., 2011; Pitt et al., 2000; Sheth et al., 2010), as a model to investigate the function of Nup structures in the cytoplasm. We find that cytoplasmic Nup foci contain only a subset of Nups and do not always colocalize with membranes, indicating that these structures do not represent pore complexes. Instead, we propose that, as in the central channel of nuclear pores, cohesive interactions among FG domains drive condensation of Nups in the cytoplasm. We find that the solubility of cytoplasmic Nups is controlled by the same mechanisms that regulate network formation in the central channel, including posttranslational modifications and NTR binding. Finally, in contrast to prior models proposing that Nup foci function in different aspects of pore biogenesis, we find that these structures are transient and do not serve an essential biological role. Instead, our data suggest that uncontrolled Nup condensation presents a potential danger that must be counteracted by energy-consuming mechanisms.
Results
Cytoplasm-facing Nups form cytoplasmic foci in C. elegans oocytes
To systematically characterize Nup distribution in oocytes, we used a collection of genomically-encoded tags, transgenes, and antibodies against 16 Nups (including one or more representatives of each subcomplex) and the Nup358 binding partners RanGAP and NXF1 (Figure 1A; Table S1). In C. elegans hermaphrodites, oocytes are arranged in the oviduct in order of maturation with the oldest oocyte (−1 position) nearest the spermatheca. The −1 oocyte is activated by a secreted sperm signal to undergo nuclear envelope breakdown (NEBD) in preparation for meiotic divisions, ovulation, and fertilization (Figure 1B) (Huelgas-Morales and Greenstein, 2018). We focused our initial survey on oocytes in the −4 and −3 positions of day 2 adult hermaphrodites where Nup foci are most prominent.
As expected, all Nups tested localized to the nuclear envelope (Figures 1C and S1A). Nuclear basket and Y complex Nups additionally localized to the nucleoplasm and meiotic chromosomes, respectively, as previously described (Hamed et al., 2021; Hattersley et al., 2016; Wu et al., 2001). A subset of Nups also localized to cytoplasmic foci, including FG-Nups of the central channel and cytoplasmic filaments (Nup62, Nup98, Nup214, and Nup358) and their binding partners (Y complex Nups, Nup88, RanGAP, and NXF1) (Figures 1C, S1A, and S1B). The transmembrane Nups gp210 and NCD1 could be detected throughout the endoplasmic reticulum as previously described (Galy et al., 2008), but did not enrich in cytoplasmic foci, nor did Nup35, an inner ring complex Nup. We also analyzed the distribution of a subset of Nups in 4-cell stage embryos and obtained the same results except for Nup35, which did not form foci in oocytes but did in embryos (Figures S1C and D). We conclude that cytoplasmic Nup foci primarily enrich cytoplasm-facing FG-Nups and their binding partners (Figure 1A).
Co-staining experiments using the mAb414 antibody (Davis and Blobel, 1986), which recognizes multiple Nups, suggested that distinct Nups co-localize in the same foci (Figures S1A and D). To examine this systematically, we crossed a subset of GFP-tagged Nups pairwise with Nup62::wrmScarlet. As expected all Nups tested colocalized with Nup62::wrmScarlet at the nuclear envelope (Figure 1D). Nups that localize to cytoplasmic foci (Nup85, Nup88, Nup98, and Nup358) additionally colocalized with Nup62::wrmScarlet in all foci, confirming that the foci are assemblies of multiple Nups. Additionally, quantification of the ratio of the GFP-tagged Nup to Nup62::wrmScarlet revealed that each Nup accumulates in fixed stoichiometry relative to Nup62 at the nuclear envelope, but with variable stoichiometry in the cytoplasmic foci (Figure 1D).
A recent study reported that cytoplasmic Nup foci do not overlap with endoplasmic reticulum membranes in HeLa or Cos7 cells (Ren et al., 2019). To determine whether Nup foci in C. elegans colocalize with membranes, we visualized GFP::Nup88 in the presence of a luminal marker of the endoplasmic reticulum and nuclear envelope (Fan et al., 2020). As expected, GFP::Nup88 at the nuclear envelope colocalized with the reporter, but only a subset (20%) of GFP::Nup88 cytoplasmic foci overlapped with the endoplasmic reticulum (Figure 1E), indicating that the majority of cytoplasmic Nup foci in oocytes are not transmembrane structures.
In summary, our observations indicate that cytoplasmic Nup foci specifically accumulate cytoplasm-facing FG-Nups and their binding partners in varying stochiometric amounts, lack critical nuclear pore scaffolds including transmembrane and inner ring complex Nups, and do not always associate with membranes. These observations suggest that Nup foci are unlikely to assemble by the same mechanisms that yield mature nuclear pores and may form instead by spontaneous condensation of FG-Nups.
Nup foci assembly is driven by FG-Nups that accumulate in excess of Csat in oocytes
Condensation is sensitive to concentration: proteins de-mix into dense and dilute phases when concentration exceeds the saturation concentration (Csat), the maximum concentration allowed in the soluble, dilute phase (Alberti et al., 2019). Proteins maintained at concentrations just above Csat will form condensates, but these will account for only a small proportion of total molecules, most of which will remain in the soluble, dilute phase. To determine the amount of Nup molecules in foci, we used Imaris software to quantify the relative amount of Nup fluorescence in nuclei, the cytoplasm, and cytoplasmic foci as a percent of total fluorescence in oocytes (see materials and methods). This analysis revealed that, in day 2 adults, only a minority (<3%) of Nup molecules accumulate in cytoplasmic foci (Figure 2A). The majority distribute between a nuclear pool (~30-40%) and a diffuse cytoplasmic pool (~60-70%). The cytoplasmic pool is the least concentrated but largest by volume and is readily visualized in sum projection photomicrographs (Figure S2A). We also found that Nup foci are highly influenced by animal age and increase significantly between days 1 and 2 of adulthood (Figure S2B), therefore, animals were carefully age-matched in all experiments.
The observation that cytoplasmic Nup foci represent only a minor fraction suggests that the concentration of FG-Nups in oocytes is near Csat and that depletion of a single FG-Nup may be sufficient to lower FG-Nup concentration below Csat and eliminate Nup foci. Consistent with this prediction, we found that RNAi depletion of the cytoplasm-facing “scaffold” FG-Nups (Nup62, Nup98, Nup214, or Nup358) and the FG-Nup binding partner Nup88, strongly reduced the proportion of GFP::Nup85 in Nup foci (Figures 2B, S2C-E), without affecting GFP::Nup85 levels at the nuclear envelope (Figure 2B). In contrast, depletion of non-FG or nucleoplasmic Nups had no effect, whereas loss of Nup35 or the transmembrane Nups NDC1 or gp210 enhanced the formation of Nup foci, including inducing formation of ectopic Nup foci in the distal germline (Figures 2B, 2C, S2C, and S2F). Nup35 and NDC1 are structural Nups required for pore assembly (Mansfeld et al., 2006; Mauro et al., 2022; Ródenas et al., 2009) and, as expected, depletion of these Nups also decreased the intensity of Nups at the nuclear envelope. Together these findings suggest that high concentrations of the FG-Nups Nup62, Nup98, Nup214, and Nup358 in the cytoplasm drive Nup foci assembly. Depletion of Nup88, which is structured but interacts with multiple subcomplexes containing FG-Nups (Fornerod et al., 1997; Griffis et al., 2003; Xylourgidis et al., 2006; Yoshida et al., 2011), partially depleted Nup foci, suggesting that interactions between FG-Nups subcomplexes contribute to foci formation.
To directly test whether increased levels of FG-Nups cause foci formation, we generated a transgenic strain with an extra copy of nup214::wrmScarlet expressed under the control of the germline-specific mex-5 promoter (Fan et al., 2020). We found that overexpression of Nup214::wrmScarlet was sufficient to increase the proportion of endogenous mNeonGreen::Nup358 in Nup foci by 4-fold (Figures 3A and S3A). Prior RNAi experiments indicate that Nup214 is non-essential in C. elegans (Galy et al., 2003), therefore we used CRISPR genome engineering to generate a complete deletion of the nup214 locus and examined nup214Δ homozygous mutants for Nup foci. Using three independent markers (mNeonGreen::Nup358, RanGAP::wrmScarlet, and mAb414), we found that Nup foci were significantly reduced in nup214Δ mutant oocytes and embryos (Figures 3B, 3C, and S3B). Nup foci were also largely absent in a second independently obtained nup214 deletion allele (Figure S3C). Despite lacking robust Nup foci, nup214Δ embryos were 100% viable (Figure 3D).
Nup214 is required for Nup88 stabilization and targeting to pore complexes (Xylourgidis et al., 2006), and consistent with lack of Nup214 activity, GFP::Nup88 was largely mislocalized to the cytoplasm in nup214Δ mutants (Figure S3D). This observation further suggests that specific Nups drive foci formation, as nup214Δ mutants lack robust foci despite mislocalization of Nup88 to the cytoplasm. We conclude that Nup foci result primarily from condensation of FG-Nups, which are maintained at concentrations just above solubility in oocytes.
Nup solubility is responsive to cell cycle phase
Although Nup foci are readily apparent in growing oocytes (−3 and −4 positions in the oviduct), they are largely absent from maturing oocytes (−1 position) which undergo NEBD in preparation for the meiotic divisions (Figures 1B and 4A). To distinguish whether this difference is due to a change in Nup concentration or solubility, we compared the cytoplasmic concentration of the FG scaffold mNeonGreen::Nup358 across −3 to −1 oocytes. We found that the cytoplasmic concentration and total levels of mNeonGreen::Nup358 increase from the −3 to the −1 oocyte (Figures 4A and S4A). In the same period, the percent of mNeonGreen::Nup358 in cytoplasmic foci decreased, most dramatically between the −3 and −2 oocyte, with no detectable foci remaining in most −1 oocytes (Figure 4A). In early embryos, Nup foci cycled with each cell division, disassembling at NEBD during mitosis and reassembling in interphase (Figure 4B). The cytoplasmic Nup concentration increased at each mitosis, while total Nup levels remained stable (Figure 4B and S4B).
These observations suggest that entry into M phase increases the solubility of Nups, both in cytoplasmic foci as well as at the nuclear envelope. Consistent with this hypothesis, we found that GFP::Nup88 remained in foci in the −1 oocytes of fog-2(q71) females, which do not initiate M phase and do not disassemble nuclear pores (Figure 4C) (Schedl and Kimble, 1988). Unlike wild-type oocytes which are continuously ovulated, fog-2(q71) oocytes remain in the oviduct and accumulate ~50% more total GFP::Nup88 and ~14-fold more GFP::Nup88 in foci compared to wild-type (Figures 4C and S4C). Arrested oocytes also accumulated Nup35 and ELYS in foci, which were not observed in foci in growing oocytes (Figure S4D). fog-2(q71) arrested oocytes, however, did not condense the stress granule scaffold G3BP, despite increased levels of G3BP (Figures 4C and S4E). We conclude that the solubility of Nups in the cytoplasm is highly tuned to cell cycle stage and may be subject to the same regulatory processes that drive nuclear pore dissolution during M phase.
Phosphorylation, GlcNAcylation, and the nuclear transport receptor CRM1 promote Nup solubility
Nuclear pore disassembly during NEBD is well characterized and driven in large part by Nup phosphorylation (Kutay et al., 2021). To test whether cell cycle kinases similarly regulate the solubility of cytoplasmic Nups, we used RNAi to deplete PLK1 and CDK1, two kinases that phosphorylate Nups to drive nuclear pore disassembly and are active during oocyte maturation in C. elegans (Chase et al., 2000; De Souza et al., 2004; Huelgas-Morales and Greenstein, 2018; Laurell et al., 2011; Linder et al., 2017; Martino et al., 2017; Onischenko et al., 2005; Rahman et al., 2015). Additionally, the CDK1 binding partner cyclin B has been observed to localize to cytoplasmic Nup structures in Xenopus oocytes (Beckhelling et al., 2003). In agreement with this study, we observed that CDK1 enriches in Nup foci in C. elegans oocytes (Figure S5A). We found that RNAi depletion of both PLK1 and CDK1 increased the percent of Nup88 in foci as well as at the nuclear envelope (Figures 5A, 5B, S5B, and S5C). Inhibition of the phosphatase PP2A was reported to block nuclear pore complex and Nup foci assembly in Drosophila embryos (Onischenko et al., 2005). Similarly, we found that RNAi depletion of the scaffolding subunit of PP2A led to a striking loss of Nup foci as well as depletion of Nup88 from the nuclear envelope (Figures 5A, 5B, S5B, and S5C). We conclude that Nup phosphorylation by cell cycle kinases increases Nup solubility in oocytes and that PP2A phosphatase activity counteracts this effect.
FG-Nups are also heavily modified by O-GlcNAcylation, which has been previously proposed to limit FG domain interactions in the central channel (Ruba and Yang, 2016; Yoo and Mitchison, 2021). The anti-GlcNAc RL2 antibody stained the nuclear envelope and Nup foci in oocytes as well as embryos (Figures S5D and E). O-GlcNAcylation is catalyzed by the enzyme O-GlcNAc transferase (OGT), and endogenously-tagged OGT::GFP was enriched in Nup foci (Figure S5F). ogtΔ mutant animals lacked Nup O-GlcNAcylation as previously described (Figures S5D and E) (Hanover et al., 2005) and exhibited enhanced formation of Nup foci (Figures 5A, 5C, S5B, and S5G). We also visualized Nup foci in a loss of function allele of the C. elegans O-GlcNAcase (OGA) reported to exhibit higher levels of Nup GlcNAcylation in embryos (Forsythe et al., 2006). We did not detect a significant change in Nup foci in the oga mutant, suggesting that, in oocytes, Nups may be sufficiently O-GlcNAcylated such that loss of OGA activity does not affect Nup solubility.
Recent studies have suggested that nuclear transport receptors (NTRs) function as chaperones to prevent aggregation of intrinsically disordered proteins (Guo et al., 2018; Hofweber et al., 2018; Hutten et al., 2020). The exportin CRM1 makes high affinity interactions with the FG-Nup scaffolds Nup214 and Nup358 (Port et al., 2015; Ritterhoff et al., 2016; Tan et al., 2018). We found that endogenously tagged CRM1::mNeonGreen, as well as the NTR transportin::mNeonGreen, distribute between the cytoplasm (~75%) and nucleoplasm (~25%), with a minor fraction (<1%) present at cytoplasmic Nup foci (Figures S6A and B). RNAi depletion of CRM1 led to an increase in Nup foci formation (Figures 5A, 5D, S5B, S6C, and S6D). This effect is unlikely to be due to impaired nuclear export, as Nup foci were not altered in worms treated for 4 hours with the CRM1 inhibitor leptomycin B (LMB), which binds within the NES binding groove to prevent interaction with NES cargos (Figures 5A, 5D, S5B, S6D, and S6E) (Sun et al., 2013). RNAi depletion of transportin did not affect Nup solubility (Figures S6F and G), indicating that CRM1 may be uniquely effective at solubilizing cytoplasmic Nups in oocytes. In summary, we conclude that Nup solubility is enhanced by phosphorylation, O-GlcNAcylation, and CRM1 binding.
Nup98 overexpression in neurons leads to ectopic Nup foci, Nup depletion from the nuclear envelope, and neuronal dysfunction
Outside of oocytes, we rarely observed cytoplasmic Nup foci in wild-type hermaphrodites. The concentration of Nup358 in terminally differentiated intestinal cells was ~30% that of oocytes (Figure 6A), suggesting that somatic cells maintain FG-Nup concentration well below Csat. We therefore wondered whether overexpression of an FG-Nup might be sufficient to drive the formation of FG-Nup foci in somatic cells. Nup98 is unique among Nups as being highly cohesive and interacting with multiple structured Nups (Onischenko et al., 2017; Schmidt and Görlich, 2016). Overexpression of Nup98 from a transgene driven by the neuron-specific rab-3 promoter led to abundant cytoplasmic foci, that were not observed when Nup98 is only expressed from its endogenous locus (Figure S7A). Interestingly, the rab-3p::Nup98 animals had shorter lifespans (Figure S7B) and appeared uncoordinated (barely moving) on plates or in liquid (Figure 6B and Videos S1 and 2), consistent with neuronal dysfunction and paralysis (Dimitriadi and Hart, 2010).
To determine whether Nup98 condensates disrupts nuclear pore assembly, we examined the expression of an endogenous Nup (Nup62) in the presence of the rab-3p::Nup98 transgene. In non-neuronal cells that did not express the transgene, Nup62::wrmScarlet localized to the nuclear envelope as in wild-type (Figures 6C and D). In contrast, in neurons overexpressing Nup98, Nup62::wrmScarlet was recruited to the cytoplasmic Nup98 foci and depleted from the nuclear envelope (Figure 6D). Interestingly, neuronal overexpression of another FG-Nup, Nup358, led to cytoplasmic foci that did not recruit an endogenous Nup and did not cause paralysis, though swimming behavior was slightly altered (Figures 6E, S7C, and S7D and Videos S3 and 4). We conclude that, in somatic cells, some cytoplasmic Nup foci are tolerated, while others containing the highly cohesive Nup98 interfere with nuclear pore assembly and cause cellular disfunction.
Discussion
In this study, we have characterized the Nup condensates that form in the cytoplasm of C. elegans oocytes and early embryos. Our findings (summarized in Figure 7) indicate that, as cytoplasmic Nup levels rise in growing oocytes, cohesive interactions among Nup FG domains drive Nup condensation in the cytoplasm. Although Nup condensates appear prominent when observed by fluorescent microscopy, they account for less than 3% of total cellular Nup, as phosphorylation, GlcNAcylation, and CRM1 chaperoning maintain the majority of Nup molecules in a soluble state. Oocyte Nup condensates disassemble during M phase and are not required for embryonic viability. Ectopic assembly of Nup condensates in post-mitotic neurons, however, can deplete Nups from the nuclear envelope and cause organismal paralysis. We conclude that cytoplasmic Nup condensates do not serve an essential function and are potentially toxic compartments.
Cytoplasmic Nup foci arise by condensation of FG-Nups and their binding partners
Several lines of evidence indicate that Nup foci in C. elegans oocytes and embryos are condensates that arise when FG-Nups exceed their solubility limit. First, Nup foci only contain cytoplasm-facing FG-Nups and their binding partners and lack nucleoporins essential for pore assembly including transmembrane Nups. Second, unlike nuclear pores, Nup foci display heterogeneous Nup stoichiometry and the majority do not colocalize with membranes. Third, Nup foci account for less than 3% of total Nup molecules and depletion and overexpression of FG-Nups eliminate and enhance, respectively, foci formation, consistent with a concentration-dependent process. Depletion of Nup88, which is structured but associates with multiple FG-Nups, partially reduced foci formation, whereas depletion of the central channel FG-Nup Nup54 had no significant effect. Together, this suggests that interactions among specific FG-Nups, including Nup62, Nup98, Nup214, and Nup358, and their binding partner Nup88 scaffold cytoplasmic condensates.
Our observations do not support the view that Nup foci correspond to annulate lamellae or another type of nuclear pore precursor, which would be expected to mirror nuclear pore stoichiometry and membrane association. Electron microscopy studies reported that annulate lamellae were absent in growing oocytes and embryos and present in only ~10% of arrested oocytes in C. elegans (Langerak et al., 2019; Patterson et al., 2011; Pitt et al., 2000). In contrast, we observe Nup foci in 100% of growing or arrested oocytes as well as embryos. While the transmembrane Nups NDC1 and gp210 were not enriched in foci in oocytes, a recent study found that a subset of Nup160::GFP foci overlapped with NDC1 in C. elegans 1-cell embryos (Mauro et al., 2022). We cannot exclude the possibility that a small subset of Nup foci correspond to annulate lamellae, however, we favor the view that the majority are “accidental” condensates that form when Nup levels exceed the solubility threshold.
Cytoplasmic Nup foci have been observed by fluorescent microscopy across diverse cell types ranging from primary neurons to transformed cell lines (Cordes et al., 1996; Raghunayakula et al., 2015). Consistent with our findings, recent systematic tagging of endogenous Nups in HEK293T cells revealed that cytoplasm-facing FG-Nups and their binding partners accumulate in cytoplasmic foci, but Nup153, which faces the nucleoplasm, does not (Cho et al., 2022). Similarly, Colombi et al., 2013 reported that Nup foci exist in yeast that contain multiple FG-Nups but lack transmembrane or inner ring complex Nups, and suggested that the foci did not represent annulate lamellae. We also found that the FG-Nup Nup214 forms numerous foci in yeast cells, but nucleoplasm-facing Nup50 and Nup153 do not (Figure S7E). Together these observations suggest that most Nup foci do not represent fully formed pores, as expected for annulate lamellae. We suggest instead that Nup foci arise in cells whenever the concentration of FG-Nups in the cytoplasm exceeds the solubility threshold. Consistent with this view, depletion of scaffold nucleoporins that liberate FG-Nups enhance foci formation in C. elegans oocytes (Figures 2 and S2), yeast (Makio et al., 2009) and HeLa cells (Raghunayakula et al., 2015).
Phosphorylation, GlcNAcylation, and CRM1-mediated chaperoning limit Nup condensation
Our findings suggest that the same mechanisms that regulate Nups at nuclear pores regulate Nup solubility in the cytoplasm. Phosphorylation by the mitotic kinases PLK1 and CDK1 drive pore complex disassembly during M phase (Kutay et al., 2021) and we find that the same kinases promote Nup solubility in oocytes. Given the high number of annotated phosphorylation sites for most Nups (Hampoelz et al., 2019a), it is likely that multiple kinases act combinatorially to promote Nup solubility, including NIMA and DYRK kinases which have been implicated in nuclear pore disassembly (De Souza et al., 2004; Laurell et al., 2011; Wippich et al., 2013). Consistent with phosphorylation driving Nup solubility, cellular fractionation experiments found that Nups in the cytoplasm are phosphorylated (Onischenko et al., 2004). Our observations indicate that GlcNAcylation additionally contributes to Nup solubility in the cytoplasm, a modification that has also been proposed to prevent non-productive interactions within the nuclear pore central channel (Ruba and Yang, 2016; Yoo and Mitchison, 2021). Numerous studies have reported a protective role for O-GlcNAcylation in neurodegenerative disease (Lee et al., 2021), raising the possibility that this modification plays a general solubilizing role for aggregation-prone proteins.
We find that Nup solubility is also enhanced by the NTR CRM1. NTRs cross the permeability barrier through direct, transient association with Nup FG domains (Bayliss et al. 1999; Iovine et al., 1995). The selective phase model of transport proposes that NTR binding catalyzes the local dissociation of the FG network to enable efficient passage of cargo (Schmidt and Görlich, 2016), and it has been suggested that NTR binding also prevents non-productive Nup interactions outside of pore complexes (Frey and Görlich, 2007; Schmidt and Görlich, 2015). NTRs have also been proposed to prevent assembly of specific Nups during mitosis (Harel et al., 2003; Walther et al., 2003), and to chaperone histones (Padavannil et al., 2019), mitotic spindle components (Nachury et al., 2001), and intrinsically disordered RNA binding proteins including FUS and TDP-43 (Guo et al., 2018; Hofweber et al., 2018; Hutten et al., 2020).
NTRs are present at high concentrations in the cytoplasm and are relatively large proteins (~100-150 kD), thus their available binding surfaces make them good candidates for chaperones (Springhower et al., 2020). The crystal structure of a Nup214/CRM1 complex reveals that the Nup214 FG domain makes extensive contacts with hydrophobic patches on the surface of CRM1 (Port et al., 2015). Furthermore, CRM1 generates high affinity interactions with both Nup214 and Nup358 that are significantly stronger than the weak, transient interactions characteristic of most Nup/NTR pairs (Port et al., 2015; Ritterhoff et al., 2016; Tan et al., 2018). Both Nup214 and Nup358 are required for condensate formation in C. elegans oocytes, therefore stable interaction of these specific Nups with CRM1 likely suppresses condensation in the cytoplasm. Of note, the NTR transportin did not affect Nup solubility in C. elegans oocytes, suggesting that chaperoning behavior may be specific to CRM1.
Nup foci might not serve an essential biological role and are potentially toxic
Cytoplasmic Nup foci have been proposed to function in pore biogenesis in Drosophila oocytes (Hampoelz et al., 2019b), nuclear pore inheritance during cell division in yeast (Colombi et al., 2013), or as reservoirs of pre-formed pore complexes to support the rapid divisions of nuclei in cell culture and Drosophila embryos (Hampoelz et al., 2016; Ren et al., 2019). Our analyses do not support such roles for Nup foci in C. elegans. First, Nup foci account for only a very small proportion of FG-Nups (<3%) in oocytes, the vast majority of which exist in the cytoplasm as a soluble pool. Second, Nup foci are transient structures that dissolve fully at oocyte maturation and every M phase thereafter, making them unlikely to be a source of partially or fully assembled pore complexes. Third, we have characterized a mutant lacking Nup214, a non-essential FG-Nup, that is 100% viable despite lacking abundant Nup foci in oocytes and embryos. We suggest instead that Nup foci are “accidental” condensates that offer no benefits to cells and may in fact be precursors to toxic condensates. Aberrant condensation of Nup98 and Nup214 fusion proteins drives oncogenic transformation in certain types of leukemia (Chandra et al., 2022; Terlecki-Zaniewicz et al., 2021; Zhou and Yang, 2014) and Nups are often present in pathological inclusions in primary patient samples and models of neurodegenerative disease (Chandra and Lusk, 2022; Fallini et al., 2020; Hutten and Dormann, 2020). Our findings suggest that Nups are not merely passive clients of pathological inclusions but can initiate the formation of toxic condensates. Ectopically expressed Nup98 formed cytoplasmic condensates in neurons that recruited endogenous Nup62 away from the nuclear envelope and caused paralysis. Nup98 is highly cohesive due to its extreme hydrophobicity and interacts with Nups from multiple subcomplexes (Griffis et al., 2003; Onischenko et al., 2017; Yoshida et al., 2011); therefore Nup98 may be uniquely capable of seeding toxic condensates. Our findings are consistent with recent studies reporting that cytoplasmic FG-Nups drive aggregation of TDP-43 in both ALS/FTLD and following traumatic brain injury (Anderson et al., 2021; Gleixner et al., 2022).
The deleterious effects of Nup condensation are likely context dependent. In fully mature arrested oocytes, Nup condensation increases by ~14 fold over growing oocytes, yet is likely not damaging as the majority of arrested oocytes go on to form viable embryos when fertilized (Jud et al., 2008). Pore complexes and Nup condensates in oocytes and embryos are fully disassembled during M phase, allowing for a cycle of “renewal” with each cell division. We suggest that Nup condensation may only be dangerous in post-mitotic cells that lack M phase specific Nup solubilizers and where certain Nups are naturally long-lived (D’Angelo et al., 2009; Toyama et al., 2013).
In summary, the high propensity of FG-Nups to form multivalent networks is critical in establishing the nuclear pore permeability barrier but presents a danger if Nup condensation is allowed to occur outside of nuclear pores. Cells like oocytes that naturally accumulate and clear Nup condensates offer a powerful model to study the mechanisms that promote and reverse Nup condensation.
Materials and methods
C. elegans and yeast strains and culture
C. elegans were cultured using standard methods (Brenner, 1974). Briefly, worms were maintained at 20°C on normal nematode growth media (NNGM) plates (IPM Scientific Inc. cat # 11006-548) seeded with OP50 bacteria. We have found that Nup solubility is highly influenced by multiple factors including animal age: for all Nups tested the number and size of foci increased significantly between days 1 and 2 of adulthood (see Figure S2B). Therefore, for all experiments worms were synchronized as day 1 or 2 adults using vulval morphology to stage L4 larvae. The age of animals used for each experiment is indicated in figures and legends.
Endogenous npp-21 (TPR) was tagged with GFP using CRISPR/Cas9-mediated genome editing as previously described (Arribere et al., 2014). Endogenous npp-24 (Nup88) and npp-2 (Nup85) were tagged with G>F>P using SapTrap CRISPR/Cas9 gene modification as previously described (Schwartz and Jorgensen, 2016). G>F>P contains Frt sites in introns 1 and 2 of GFP that enable FLP-mediated, conditional knockout; in the absence of FLP, the construct behaves as normal GFP. Endogenous npp-19 (Nup35) was tagged with G>F>P based on protocols for nested CRISPR (Vicencio et al., 2019) and “hybrid” partially single-stranded DNA donors (Dokshin et al., 2018). All other endogenous edits were performed using CRISPR/Cas9-mediated genome editing as described previously (Paix et al., 2017). Transgenic Nup214 and Nup98 strains (JH4119 and JH4204) were generated using SapTrap cloned vectors as previously described (Fan et al., 2020). Standard crosses were used to generate strains with multiple genomic edits. All strains used or generated in this study are described in Table S1.
Yeast strains were generated using homologous recombination of PCR-amplified cassettes (Longtine et al., 1998). Endogenous NUP159 (Nup214), NUP60 (Nup153), and NUP2 (Nup50) were tagged by amplifying the mNeonGreen::HIS3 cassette from pFA6a-mNeonGreen::HIS3 (Thomas et al., 2019) using primers with homology to the C-termini (without the stop codon) and downstream regions of the genes. Yeast strains generated in this study are described in Table S1.
RNAi
RNAi was performed by feeding (Timmons and Fire, 1998). RNAi vectors were obtained from the Ahringer or Open Biosystems libraries and sequence verified, or alternatively cloned from C. elegans cDNA and inserted into the T777T enhanced RNAi vector (Addgene cat # 113082). RNAi feeding vectors were freshly transformed into HT115 bacteria, grown to log phase in LB + 100 μg/mL ampicillin at 37°C, induced with 5 mM IPTG for 45 min, and plated on RNAi plates (50 μg/mL Carb, 1 mM IPTG; IPM Scientific Inc. cat # 11006-529). Seeded plates were allowed to dry overnight at RT before adding L4 larvae or day 1 adults. For depletion of Nup98 (Figures 2B and S2C), RNAi feeding was performed for 6 hr at 25°C; partial depletion was used to minimize cytological defects caused by loss of Nup98. For all other experiments, RNAi feeding was performed for 18-24 hr at 25°C. For all experiments, control worms were fed HT115 bacteria transformed with the corresponding L4440 or T777T empty vector.
Immunofluorescence
For immunostaining of embryos, gravid adults were placed into 7 μL of M9 media on a poly-L-lysine coated slide and compressed with a coverslip to extrude embryos. For immunostaining of oocytes, staged adults were dissected on poly-L-lysine slides to extrude the germline, and a coverslip was placed gently on top. In both cases, slides were immediately frozen on aluminum blocks pre-chilled with dry ice. After > 5 min, coverslips were removed to permeabilize embryos (freeze-cracking), and slides were fixed > 24 hr in pre-chilled MeOH at −20°C. Slides were then incubated in pre-chilled acetone for 10 min at −20°C, and blocked in PBS-T (PBS, 0.1% Triton X-100, 0.1% BSA) for > 30 min at RT. Slides were then incubated overnight in primary antibody in a humid chamber at 4°C. Slides were washed 3 x 10 min in PBS-T at RT, incubated in secondary antibody for 2 hr in a humid chamber at RT, and washed 3 x 10 min in PBS-T at RT. Slides were then washed 1x in PBS before being mounted using Prolong Glass Antifade Mountant with NucBlue (Thermo Fisher cat # P36981). Primary antibodies were diluted as follows: mAb414 (1:1,000; Biolegend cat # 902907), αNup358 (1:250; Novus Biologicals cat # 48610002), αNup50 (1:250, Novus Biologicals cat # 48590002), αGlcNAc RL2 (1:100; Invitrogen cat # MA1-072), αNup96 (1:250, (Ródenas et al., 2012)), αNup153 (1:250, (Galy et al., 2003)), αOLLAS-L2 (1:50, Novus Biologicals cat # NBP1-06713). Secondary antibodies were diluted as follows: Cy3 Donkey αMouse IgG (1:200; Jackson cat # 715-165-151), AlexaFluor 488 Goat αRabbit IgG (1:200; Invitrogen cat # A-11034), AlexaFluor 568 Goat αRabbit IgG (1:200; Invitrogen cat # A-11011), Alexa Fluor 488 Goat αRat IgG (1:200; Invitrogen cat # A-11006), AlexaFluor 488 αGFP (1:500; Invitrogen cat # A-21311).
LMB and HXD treatment, and HaloTag and Hoechst labeling
For CRM1 inhibition, leptomycin B (LMB; Sigma cat # L2913) was diluted in OP50 bacteria to a final concentration of 500 ng/mL and seeded on NNGM plates. 10-20 day 1 adults were transferred to LMB or control vehicle plates and incubated at 20°C for 4 hr prior to imaging. For treatment of yeast with 1,6-hexanediol (HXD), log-phase yeast were pelleted, re-suspended in media containing 5% HXD (Acros Organics cat # 629-11-8), and allowed to grow for 10 min at 30°C prior to imaging.
For HaloTag labeling, Janelia Fluor 646 HaloTag Ligand (Promega cat # GA1121) was diluted in OP50 bacteria to a final concentration of 30 μM and seeded on NNGM plates. 10-20 day 1 adults were added and incubated without light at 20°C for 12-16 hrs prior to imaging. For Hoechst staining, Hoechst 33342 dye (Thermo Fisher cat # 62249) was diluted in OP50 bacteria to a final concentration of 200 μM and seeded on NNGM plates. 10-20 day 1 adults were added and incubated without light at 20°C for 12-16 hrs prior to imaging.
Embryonic viability and lifespan analysis
To measure embryonic viability, six day 1 adults were transferred to six NNGM plates (36 worms total) and allowed to lay embryos for 1 hr at 20°C. Adults were then removed and the number of embryos on each plate was counted. Embryos were then allowed to hatch, and the number of adults on each plate was counted after 3 days at 20°C. Viability counts were repeated in three independent experiments, and embryonic viability was measured as the number of surviving adults divided by the original number of embryos counted in each experiment.
To measure adult lifespan, 75 day 1 adults were transferred to five NNGM plates (15 worms per plate) and incubated at 20°C. Worms were scored daily and considered dead if they failed to move when prodded. Worms were transferred every 2 days to avoid progeny, and any worms that crawled off the plates were censored from analysis.
Swimming assay
To measure swimming behavior, 5-10 day 1 adults were transferred to a 33 mm culture dish (MatTek cat # P35G-1.5-14-C) containing 400 μL M9 media and immediately filmed using an Axiocam 208 color camera (Zeiss) mounted on a Stemi 508 Stereo Microscope (Zeiss). Swimming assays were performed at RT (~22°C). Movies were exported to ImageJ, and the number of body bends per minute was counted manually.
Imaging
For live imaging of germlines, five staged adults were transferred to the middle well of a 3-chambered slide (Thermo Fisher cat # 30-2066A) in 10 μL of L-15 medium (Thermo Fisher cat # 21-083-027) with 1 mM levamisole. 20 μm polystyrene beads (Bangs Laboratories Inc. cat # PS07003) were then added to support a coverslip (Marienfeld cat # 0107052). Germlines were imaged using an inverted Zeiss Axio Observer with CSU-W1 SoRa spinning disk scan head (Yokogawa), 1x/2.8x/4x relay lens (Yokogawa), and an iXon Life 888 EMCCD camera (Andor) controlled by Slidebook 6.0 software (Intelligent Imaging Innovations). To image entire germlines, a 20 μm Z stack (1 μm step size) was captured using a 63x-1.4NA objective (Zeiss) with the 1x relay lens. For high resolution images of oocytes, 3 μm Z stacks (0.1 μm step size) were acquired using the 63x-1.4NA objective with the 2.8x relay lens. As germline condensates are highly sensitive to imaging-induced stress (Elaswad et al., 2022), care was taken to avoid compression of germlines, and all animals were imaged only once and maintained on the slide for < 5 min.
For live imaging of embryos, five young adults were transferred to 10 μL of L-15 medium on a coverslip and dissected to release embryos. 20 μm polystyrene beads were then added to prevent compression, and the coverslip was inverted onto a microscope slide (Thermo Fisher cat # 12-550-403). Embryos were imaged as 15 μm Z stacks (1 μm step size), captured using the 63x-1.4NA objective with the 2.8x relay lens. For imaging fixed germlines and embryos, prepared slides were imaged as 15 μm Z stacks (0.5 μm step size), captured using the 63x-1.4NA objective with the 2.8x relay lens.
For live imaging of yeast, cells were grown overnight in synthetic dropout media (Thermo Fisher cat # DF0919-15-3) at 30°C and imaged in log-phase (OD600 of ~0.5) at room temperature. Yeast were imaged as 6 μm Z stacks (0.5 μm step size), captured using the 63x-1.4NA objective with the 2.8x relay lens.
Images were exported from Slidebook software and further analyzed using ImageJ or Imaris image analysis software. For presentation in figures, images were processed using ImageJ, adjusting only the minimum/maximum brightness levels for clarity with identical leveling between all images in a figure panel. Images presented in figures are maximum intensity projections (10 μm for germlines, 15 μm for embryos, 6 μm for yeast) or single focal planes as indicated in the legends.
Image quantification
The overlap of GFP or mNeonGreen-tagged Nups with Nup62::wrmScarlet (Figure 1D) was measured using single focal planes exported to ImageJ. The Nup62::wrmScarlet micrograph was used to create a mask defining the nuclear envelope as well as cytoplasmic foci as individual regions of interest (ROIs). This mask was then applied to both the GFP/mNeonGreen Nup micrograph as well as the Nup62::wrmScarlet micrograph and the integrated density was measured within each ROI. To control for cytoplasmic background, the average cytoplasmic signal for the GFP/mNeonGreen Nup was multiplied by the area of each ROI, and the resulting value subtracted from integrated density for the GFP/mNeonGreen Nup. Background normalized GFP/mNeonGreen Nup values were divided by Nup62::wrmScarlet values to obtain the ratio of GFP/mNeonGreen Nup to Nup62::wrmScalet at each ROI.
To quantify the overlap of GFP::Nup88 with membranes (Figure 1E), Z stacks of oocytes expressing GFP::Nup88 and the HaloTag::HDEL reporter were manually scored into 3 categories: 1. Complete overlap (the entire Nup88 focus overlapped with HaloTag::HDEL); 2. Partial overlap (the Nup88 focus partially overlapped or was directly adjacent to HaloTag::HDEL); 3. No overlap (the Nup88 focus did not directly contract membranes marked by HaloTag::HDEL).
To quantify the distribution of Nups in oocytes as well as total expression, Z stacks were exported to Imaris image analysis software. The “Surface” tool was first used to isolate the −3 and −4 oocyte from each germline. For each pair of −3 and −4 oocytes, the Surface tool was then used to isolate both nuclei and the “Spot” tool was used to isolate cytoplasmic foci. The percent of Nup present at the nuclear envelope/nucleoplasm was measured as the intensity sum for both nuclei divided by the total intensity sum of the oocytes. Similarly, the percent of Nup present in foci was measured as the intensity sum for all foci divided by the total intensity sum of the oocytes. Finally, the percent soluble Nup was defined as 100% minus the percentage of Nup in both nuclei and foci. Total Nup expression was measured as the intensity sum of the −3 and −4 oocytes normalized to volume. To control for autofluorescent background in all measurements, staged animals lacking fluorescent tags were imaged using identical imaging settings. The average intensity sum per volume was calculated for the −3 and −4 oocytes of germlines lacking fluorescent tags and subtracted from the intensity sum measured for oocytes with tagged Nups.
To quantify the distribution of Nups in embryos, Z stacks were exported to Imaris software. The Surface tool was used to isolate the entire embryo as well as all nuclei, and the Spot tool was used to isolate cytoplasmic foci. The percent of Nup at the nuclear envelope/nucleoplasm or foci was measured as the intensity sum of all nuclei or foci divided by the total intensity sum of the embryo, respectively. The percent soluble Nup was defined as 100% minus the percentage of Nup in nuclei and foci. For all measurements, embryos lacking fluorescent tags were used to control for autofluorescent background as described for oocytes.
The Y complex component Nup85 localizes to meiotic chromosomes, therefore a high percentage of Nup85 is present in the nucleoplasm. Therefore, line-scan analysis was used to measure the amount of GFP::Nup85 at the nuclear envelope (Figure 2B). Z stacks were exported to ImageJ and line traces were drawn to pass through the central plane of −3 and −4 oocyte nuclei as well as the image background. For each nucleus, the two peak values of the nuclear envelope rim were averaged and normalized to the image background. Line-scan analysis was also used to quantify depletion of endogenous Nups from the nuclear envelope in neurons expressing rab-3p::Nup98::mNeonGreen or UPN::Nup358::mCherry (Figures 6D and E). Line traces were drawn to pass through the central plane of nuclei identified by Hoechst staining. For each nucleus, the two peak values for the nuclear envelope rim were averaged and normalized to the average value of Nup in the nucleoplasm. To quantify TFEB::GFP accumulation in the nucleus (Figure S6E), line traces were drawn to pass through the cytoplasm as well as the nucleoplasm. The average intensity value for TFEB::GFP in the nucleus was then divided by the average value of the cytoplasm.
To quantify cytoplasmic levels of mNeonGreen::Nup358 in oocytes versus intestinal cells (Figure 6A), single focal planes capturing both the germline and intestine were exported into ImageJ. For each image, 3 ROIs in the −3 oocyte, −2 oocyte, and intestinal cell cytoplasm were measured, averaged, and normalized to the image background.
Statistical analysis
All statistical tests were performed using GraphPad Prism 9.2.0 software. For comparison of three or more groups, significance was determined using a one-way ANOVA. For comparison of two groups, significance was determined using an unpaired t-test. In all figures error bars represent 95% confidence intervals (CI). For all figures, ns indicates not significant; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.
Competing interests
G.S. serves on the Scientific Advisory Board of Dewpoint Therapeutics, Inc. The other authors declare no competing interests.
Video S1. Swimming assay with control Day 1 adults lacking rab-3p::Nup98::mNeonGreen.
Video S2. Swimming assay with Day 1 adults expressing rab-3p::Nup98::mNeonGreen.
Video S3. Swimming assay with control Day 1 adults lacking UPN::Nup358::mCherry.
Video S4. Swimming assay with Day 1 adults expressing UPN::Nup358::mCherry.
Acknowledgements
We thank all members of the Seydoux and Cochella Labs, Orna Cohen-Fix, and the Baltimore Worm Club for support and many helpful discussions. We thank Cristina Ayuso and Basma Taleb for assistance with C. elegans genome engineering, Madeline Cassani for generating strain JH3656, and the Fromme Lab for generously sharing yeast strains and plasmids. We thank the Chuang Lab for sharing the transportin::mNeonGreen strain and the Greenstein Lab for sharing the GFP::NDC1 strain. Several C. elegans strains were provided by the Caenorhabditis Genetics Center (CGC), which is supported by the National Institutes of Health Office of Research Infrastructure Programs (P40 OD010440). This work was funded by the Spanish State Research Agency (PID2019-105069GB-I00) and the National Institutes of Health (R37HD037047). L.T. is a postdoctoral fellow of the Life Sciences Research Foundation supported by the Howard Hughes Medical Institute. G.S. is an investigator of the Howard Hughes Medical Institute.