A Role for the Autophagic Receptor, SQSTM1/p62, in Trafficking NF-κB/RelA to Nucleolar Aggresomes

NF-κB is a family of essential transcription factors that play a key role in many of the hallmarks of cancer including cell proliferation, apoptosis, inflammation, angiogenesis, and metastasis (1–3). The most abundant form is a heterodimer of the RelA(p65)/P50 subunits. In most healthy cells, this heterodimer is retained in the cytoplasm by the inhibitor protein IκBα. However, in the majority of human malignancies, including solid and hematologic tumors, NF-κB is aberrantly active in the nucleus that promotes cell growth, blocks apoptosis, perpetuates an inflammatory environment, accelerates disease progression, and conveys resistance to therapy (4–6). Indeed, the paramount importance of NF-κB in all stages of tumorigenesis has led to its aggressive pursuit as a therapeutic target. However, most therapies developed to date target the main switch in NF-κB activation, cytoplasmic release from IκB, which causes systemic inhibition and severe toxicity.

Once in the nucleus, NF-κB proteins are modulated by a plethora of coactivators, repressors, and posttranscriptional modifications (7–10). These nuclear regulatory pathways are also extremely important for cancer progression as they govern the genes that are activated or repressed in a specific context and hence, the downstream consequences on cell growth, death, differentiation, and inflammation (11, 12). However, unlike cytoplasmic activation of NF-κB, these postinduction, nuclear mechanisms of regulation remain poorly understood. Further mechanistic understanding in this area is essential so that drugs can be developed that specifically target cells with aberrant nuclear NF-κB activity, exploiting cancer cell vulnerabilities.

One nuclear pathway us and others have identified to be important for inhibiting aberrant NF-κB transcriptional activity and inducing the death of cancer cells is nucleolar sequestration of RelA (13). In response to classic stimuli of NF-κB such as TNF, RelA is retained in the nucleoplasm, excluded from nucleoli (13). In contrast, RelA translocates from the nucleoplasm to nucleoli in response to specific chemopreventive agents (e.g., aspirin, sulindac, sulindac suphone; refs. 13, 14), therapeutic agents [e.g., proteasome inhibitors, the antitumor agent 2-methoxyestradiol (2ME2), TRK inhibition; refs. 15, 16], and stress inducers (e.g., UV-C radiation and serum starvation; ref. 13). This differential distribution has important consequences as nucleoplasmic accumulation is associated with high NF-κB transcriptional activity and inhibition of apoptosis, while nucleolar sequestration is causally involved in repression of basal NF-κB transcriptional activity and the induction of apoptosis (13, 15, 16). Although some mechanisms that control this differential nuclear distribution have been reported, understanding is limited (17, 18). Indeed, nucleolar trafficking and sequestration in general are poorly understood. Directing RelA and other regulatory proteins to nucleoli to eliminate diseased cells is a very attractive therapeutic option that can only be realized by further understanding in this area.

Here, we demonstrate that RelA accumulates within intranucleolar aggresomes in response to therapeutic agents. We show that the nucleolar accumulation of RelA occurs alongside a flux of aggresome-related proteins into the organelle. We demonstrate that the autophagy receptor protein, SQSTM1/p62, also accumulates in nucleolar aggresomes in response to these agents. Furthermore, we identify a nucleolar localization signal at the N-terminus of p62 and, using a dominant-negative mutant deleted for this motif, demonstrate a role for p62 in trafficking RelA and other aggresome-related proteins from the nucleoplasm to nucleoli, to mediate apoptosis. Together, these data identify a novel role for p62 in trafficking nuclear proteins to nucleolar aggresomes under conditions of stress. They also provide invaluable information on the mechanisms that regulate the nuclear/nucleolar distribution of RelA that could be exploited for therapeutic purpose.

SW480 and SW620 cells were purchased from the ATCC in 2005 and 2019, respectively. These cell lines are derived from the same patient: SW480 from a primary adenocarcinoma and SW620s from a lymph node metastasis. Authentication of cell lines was carried out by the ATCC through their DNA-STR profile. Cell lines were Mycoplasma tested monthly in-house using MycoAlert Mycoplasma Detection Kit (Lonza). They were maintained at low passage number (<10 passages after thawing) in L-15 medium (Gibco) supplemented with penicillin (100 IU/mL), streptomycin (100 μg/mL), and 10% fetal calf serum. Aspirin was supplied by Sigma. It was solubilized in water using 10 N NaOH; then, the pH was adjusted to 7.0. All other chemicals and reagents were from standard commercial source and of the highest quality.

Immunocytochemistry was performed as described previously (13). Primary antibodies were: RelA (C20), RelA (F6) C23 (all Santa Cruz Biotechnology); Fibrillarin (cytoskeleton); EIF4A1, EIF4G2, EIF4H (all Cell Signaling Technology); HSPA2 and 8 (Enzo Biosciences); SUMO-2/3 (Invitrogen); β-Catenin (BD Transduction Laboratories). Cells were mounted in Vectashield (Vector Laboratories) containing 1 μg/mL DAPI. Images were captured using a Coolsnap HQ CCD camera (Photometrics Ltd.) Zeiss Axioplan II fluorescent microscope, 63× Plan Neofluor objective, a 100 W Hg source (Carl Zeiss) and Chroma 83 000 triple band pass filter set (Chroma Technology). Image capture was performed using scripts written for iVision 3.6 or Micromanager (https://open-imaging.com/). For each experiment, a constant exposure time was used. Image quantification was carried out using DAPI as a nuclear marker and C23 as a nucleolar marker along with ImageJ or Cell Profiler, image analysis software. At least 150 cells from at least 5 random fields of view were quantified for three independent experiments, or as specified in the text.

The GFP-RelA expression construct was a kind gift from E. Qwarnstrom (University of Sheffield, Sheffield, United Kingdom; ref. 19). DsRed-RelA was generated by subcloning RelA from GFP-RelA to pDsRed-C1 (Clontech). GFP-Fibrillarin was a kind gift from A. Lamond (University of Dundee, Dundee, Scotland). Wild-type (WT) GFP-p62 was a kind gift from Terje Johansen [The Arctic University of Norway (UiT), Tromso, Norway; ref. 20]. GFP-p62ΔPB1 was generated by cloning cDNA from SW480 cells corresponding to isoform 2 of p62 [Δ1–84] into pEGFP-C1 (Clontech). Full sequence analysis confirmed the sequence of p62ΔPB1 was identical to WT protein, other than lack of the first 84aa. GFP-p62ΔPB1ΔNoLS was generated by subcloning aa122–440 from GFP-p62ΔPB1 into pEGFP-C1, while GFP-p62ΔPB1ΔUBA was generated by subcloning aa84–395 from GFP-p62ΔPB1 into this plasmid. The D₃₃₅A/D₃₃₆A/D₃₃₇A mutations in the LC3-interacting domain were generated using a two-step approach. First, the N and C termini of GFP-p62ΔPB1 were amplified using p62sens3/D335Arev and D335Afor/p62antisens primers. The PCR products were then digested with NotI (this restriction site is introduced by the mutation) and ligated together. After a second round of PCR with primers p62sens3/p62antisens, the fragment was cloned into pEGFP-C1. All constructs were verified by sequencing. Plasmids were transfected into SW480 cells using Lipofectin (Gibco) as per the manufacturer's instructions. COMMD1 siRNA was generated and transfected into cells as described previously (17).

SW480 cells were transfected with GFP-fibrillarin, DsRed-RelA, or control plasmids (pEGFP, pDsRed), then treated with aspirin (5 mmol/L, 16 hours). Hoescht 33342 (Sigma-Aldrich) dye was added 30 minutes prior to imaging, to allow visualization of DNA and nucleoli (unstained area in nucleus). All FCS recordings were acquired using a Leica SP5 SMD confocal microscope using a ×63 1.2NA HCX PL Apo water lens and 488 or 561 nm CW lasers. Photon fluctuation data routed through a PicoQuant PRT 400 router were acquired at microsecond rates using external Single Photon Avalanche Photodiodes (MicroPhoton Devices). Autocorrelation traces were generated from the photon-counting histograms for each 5- to 30-second measurement using SymPhoTime v5.4.4 software (PicoQuant). In vitro calibration traces were fitted using the Triplet model (three-dimensional free diffusion model with triplet state) with informed diffusion values to yield V_eff and κ values. Diffusion within the cells is expected to be anomalous; therefore, the anomaly parameter was not fixed to one. The anomaly parameter (α) measures the departure from free Brownian diffusion (α = 1) to either superdiffusion (α > 1) or subdiffusion (α < 1) for a diffusing species. Autocorrelation curves with (α > 1) display the sharpest decay, whereas those with α < 1 decrease quite slowly (21).

Details of methods used to generate proteomic datasets are outlined in Supplementary Fig. S3. The web-based tool, GOrilla (target v background), was used to determine whether proteins that increase in nucleoli in response to aspirin (log₂ > 0.5) are enriched for specific Gene Ontology terms. Background was set as all proteins isolated from nucleoli.

Whole cell, cytoplasmic, nuclear, and nucleolar extracts were prepared as described previously (13). Bradford assays (Bio-Rad) were used to measure protein content. Extracts (30 μg) were resolved using 5% to 12% SDS-PAGE, and immunoblotting performed by standard procedures. Primary antibodies used were: RelA (C20), C23, GFP, B23/nucleophosmin, RPA194 (all from Santa Cruz Biotechnology), p62 (BD Biosciences), α-tubulin (SIGMA), α-actin (SIGMA), SUMO2/3 (Invitrogen), β-catenin (BD Transduction Laboratories), ubiquitin (Dako) and α-Tubulin (Sigma). Immunoprecipitations were performed on whole-cell lysates from aspirin-treated cells as described previously (17) using anti-RelA antibodies (C20, Santa Cruz Biotechnology) followed by p62 (anti-mouse, BD Biosciences) immunoblots.

Apoptosis assays were carried out using the AnnexinV iFluor 594 Apoptosis Assay Detection Kit (Abcam) and fluorescent microscopy, as per the manufacturer's instructions.

Proteins that translocate into nucleoli generally do not distribute evenly throughout the organelle, but accumulate in nucleolar bodies. These bodies have been independently termed nucleolar aggresomes, intranucleolar bodies, detention centers, or cavities (22–27). To help understand the nuclear distribution of RelA mechanistically, we determined whether the protein resides within such aggregates.

First, we used immunocytochemistry to examine the subnucleolar localization of RelA in response to stress inducers. Incontrovertible evidence indicates that aspirin has antitumor activity and the potential to prevent colorectal cancer (28). We are interested in the mechanisms underlying this effect and have previously demonstrated a role for nucleolar translocation of RelA (13). Therefore, we used aspirin as a tool stressor in these studies. We also used the proteasome inhibitor, MG132. We found, as has previously been described, that exposure to these agents induces an increase in size of nucleoli and segregation of nucleolar marker proteins to the periphery of the organelle (Fig. 1A and B; ref. 29). We also found that RelA accumulates in very distinct foci in the center of these enlarged nucleoli (Fig. 1A and B). This nucleolar distribution in response to aspirin was observed for SW480 colorectal cancer cells and their metastatic counterpart, SW620s (Fig. 1A; Supplementary Fig. S1). It was also observed for both endogenous and GFP-tagged RelA, suggesting that this is a genuine subnucleolar localization pattern (Fig. 1A,–C).

To further explore the nature of nucleolar RelA foci, we utilized fluorescent correlation spectroscopy (FCS). This approach is ideal as it can report molecular concentration and diffusion rate at multiple points within a cell at a microsecond timescale (30). Processes that slow down the diffusion rate of molecules, such as binding in large complexes or aggregates, are visible by a shift of the FCS curve to longer lag times.

SW480 cells were transiently transfected with DsRed-RelA and EGFP-fibrillarin, exposed to aspirin, then FCS employed to quantify the concentration and diffusion rates of the tagged proteins within the cytoplasm, nucleoplasm, and nucleoli (Fig. 1D). Supplementary Fig. S2 shows the average, normalized experimental, and fitted autocorrelation curves for DsRed-RelA in each compartment, with and without aspirin treatment. The mean t1 is marked. Examples of individual recordings from each compartment are also shown (Supplementary Fig. S2).

Analysis of FCS curves revealed that aspirin induces a significant increase in the concentration and decrease in the diffusion rate of DsRed-RelA in nucleoli, in keeping with accumulation of the protein within aggregates in this compartment (Fig. 1E and F). A reduction in the diffusion rate of DsRed-RelA was also observed in the nucleoplasm, although to a slightly lesser degree. In contrast, the diffusion rate of DsRed-RelA increased in the cytoplasm in response to the agent (Fig. 1E and F), consistent with release of RelA from the IκB inhibitor complex. Unlike DsRed-RelA, the diffusion rate of EGFP-fibrillarin remained constant in the nucleoplasm and nucleoli (Fig. 1F), as did DsRed and EGFP controls (Supplementary Fig. S3). The nucleolar concentration of EGFP-fibrillarin appeared to decrease in response to aspirin. However, this can probably be explained by segregation of this protein to the nucleolar periphery, out with the location of the data collection points (Fig. 1B, D, and E). Interestingly, in contrast to DsRed-RelA, aspirin induced a significant reduction in the cytoplasmic diffusion rate of EGFP-fibrillarin (Fig. 1F). We currently have no explanation for this result. However, it does confirm that the agent has compartment- and protein-specific effects on protein mobility.

Together, these data establish the power of FCS for investigating NF-κB signaling in distinct cellular compartments. They also provide convincing evidence that RelA resides within aggregates in nucleoli following exposure to specific therapeutic agents.

In a related study, stable isotope labeling of amino acids in culture (SILAC)-based quantitative proteomics was used to investigate aspirin- and MG132-induced changes to the cytoplasmic, nucleoplasmic, and nucleolar proteomes (Supplementary Fig. S4A–S4C). To explore the global proteomic landscape in which RelA aggregates are observed, we interrogated these datasets. Particularly, those derived from 10-hour aspirin (n = 3) and MG132 (n = 2), as immunocytochemistry indicated nucleolar RelA is evident under these conditions (Fig. 2A).

Our analysis revealed that both aspirin and MG132 have a particularly significant effect on the nucleolar proteome, compared with the cytoplasmic and nucleoplasmic proteomes (Fig. 2B; Supplementary Fig. S4C). They also revealed both agents induce an influx of proteins into the organelle (Fig. 2B; Supplementary Fig. S4C). Gene ontology analysis of proteins that increased in nucleoli in response to aspirin indicated enrichment (P < 0.00001) for terms associated with regulation of biological quality, protein folding, and response to topologically incorrect protein (Fig. 2C). Immunocytochemistry and Western blot analysis were used to analyze specific proteins, and protein families, found to accumulate in nucleoli in response to aspirin. These data confirmed that aspirin induces nucleolar accumulation of translational regulators (eIF4A1, eIF4H), chaperones (HSP90AA1, HSPA2, HSPA8), UPS proteins (ubiquitin, SUMO-2/3), and cell adhesion molecules (β-catenin; Fig. 2D,–G). Furthermore, such proteins colocalized with RelA in aggregates in this compartment (Fig. 2F). Interestingly, there was specificity in aspirin-mediated nucleolar accumulation of regulatory proteins as HSPA family chaperones accumulated in the compartment, but not HSPD1. Similarly, EIF4A1 and EIF4H translocated to nucleoli after aspirin treatment, but not EIF4G2. RelA was not identified in any compartment using SILAC-based proteomics, possibly due to difficulties detecting RelA using LC/MS-MS.

One mechanism by which MG132 alters cell phenotype is the induction of proteotoxic stress. Given the types of proteins that accumulate in nucleoli in response to aspirin, and the similarity between aspirin and MG132 effects on the nucleolar proteome, we considered that aspirin also induces proteotoxic stress. A marker for this is accumulation of insoluble proteins and so, we measured the percentage of detergent-insoluble protein following aspirin and MG132 exposure (Supplementary Fig. S4D). We found that aspirin did indeed cause an accumulation of insoluble protein in colon cancer cells. This occurred in a time-dependent manner and was significant 5 to 8 hours following exposure to the agent, which precedes the induction of apoptosis (31). Furthermore, the increase in insoluble protein was comparable with that observed with MG132 (Supplementary Fig. S4D).

Together, these data provide compelling evidence that RelA accumulates in nucleolar aggregates alongside specific chaperones and targets of the ubiquitin proteasome pathway. It also confirms that proteotoxic stress has a particularly significant effect on the nucleolar proteome (24).

One protein of considerable interest with regard to the shuttling of RelA to nucleolar aggresomes is P62/Sequestosome 1 (SQSTM1). P62 is a multifunctional protein that acts as a signaling scaffold and autophagy cargo receptor (32, 33). It colocalizes with cytoplasmic aggregates and plays a significant role in their clearance (20). It also shuttles to the nucleus, although its role in this compartment is still not clear (34). It interacts with the autophagosomal marker, LC-3B I/II, which has previously been shown to be present in nucleoli. Furthermore, it is known to interact with RelA and reportedly transports RelA to autophagosomes for degradation (35). Hence, we considered that p62 may play a role in the formation or clearance of RelA containing nucleolar aggresomes.

To begin to address this question, we first determined whether, as described for other stimuli (35), p62 and RelA interact in response to aspirin. Indeed, immunoprecipitation assays demonstrated that the agent induces a transient interaction between the two proteins (Fig. 3A). Next, we utilized immunocytochemistry and Western blot analysis to examine the subcellular localization of p62. These data revealed that aspirin induces the rapid and dose-dependent nuclear accumulation of p62, followed by nucleolar translocation of the protein (Fig. 3B and C; Supplementary Fig. S5A). Furthermore, nucleolar translocation of p62 occurred in parallel with nucleolar translocation of RelA, subsequent to their observed interaction (Fig. 3A and B; Supplementary Fig. S5A). On a single-cell basis, there was a direct correlation between cells that showed nucleolar RelA and those that showed nucleolar p62 (Supplementary Fig. S5B). In addition, p62 colocalized with RelA in foci in this compartment (Fig. 3D; Supplementary Fig. S5C). We did not detect an interaction between p62 and RelA at the time point they were present in foci (8 hours), which may reflect the insoluble nature of the aggregates. P62 was also found to accumulate in nucleoli alongside RelA response to MG132 (Fig. 3B and E). In contrast, LC-3B accumulated in the cytoplasm in response to aspirin (Fig. 3F).

Together, these data indicate that aspirin induces the nuclear/nucleolar accumulation of p62 and suggest the intriguing possibility that this protein transports RelA to nucleolar aggregates. To test this possibility, we set out to examine the pathways underlying nucleolar translocation of p62.

Given the importance of autophagy in the clearance of p62 and its cargo proteins (32), we hypothesized that there may be interplay between autophagy and nuclear/nucleolar translocation of the protein. Specifically, we speculated that nucleolar aggregates form as a consequence of overwhelmed or blocked autophagy. Contrary to this suggestion, we found that specifically blocking the formation of autophagosomes, using the inhibitor of autophagasomal degradation, bafilomycinA1, did not induce nuclear/nucleolar accumulation of p62. Rather, as has been well documented previously, the protein accumulated in autophagosomes (LC3B-positive cytoplasmic foci) in response to the agent (Fig. 4A and B).

Next, we determined whether inhibiting autophagic degradation modulated aspirin-mediated nucleolar translocation of p62 (Fig. 4A and B). Quantitative immunocytochemistry confirmed that bafilomycinA1 induced cytoplasmic aggregation of p62 (Fig. 4B). However, the percentage of cells showing this localization pattern was significantly reduced when cells were exposed to aspirin. Furthermore, this occurred in a manner dependent on aspirin dose (Fig. 4B). Similarly, when cells were exposed to aspirin alone, there was a significant increase in those showing nucleoplasmic/nucleolar p62, but this was significantly reduced in the presence of bafilomycinA1 (Fig. 4A and B). Lysotracker, which detects active lysosomes, confirmed that bafilomycinA1 was equally active in the presence and absence of aspirin (data not shown). These data suggest that the pathways that target p62 to autophagosomes, and those responsible for nuclear/nucleolar translocation of the protein, may be in active competition.

To translocate to autophagosomes, p62 first has to homodimerize, which is dependent on the N-terminal PB1 domain (34). This domain also allows p62 to interact with the autophagy receptor NBR1 and the protein kinases ERK, MEKK3, MEK5, PKCζ, and PKCλ/ι (36). To further investigate the mechanisms underlying nucleolar translocation of p62, we utilized a GFP-tagged fusion protein deleted for this domain (GFP-p62ΔPB1; Fig. 5A). GFP-P62 WT and GFP-p62ΔPB1 were transfected into SW480 cells and the subcellular localization of GFP-tagged proteins quantified in the presence and absence of aspirin (0, 3 mmol/L; Fig 5B and C). In the absence of aspirin, GFP-p62WT was present in cytoplasmic puncti. In contrast, GFP-P62ΔPB1 was diffuse within the cytoplasm, as would be expected given that it had lost the ability to self-aggregate. Following aspirin exposure, both proteins translocated into the nucleoplasm and accumulated in nucleoli. However, this effect is significantly more pronounced for GFP-P62ΔPB1 than for WT protein (Fig. 5B and C). These data indicate that the N-terminal PB1 domain is not required for nucleolar translocation of p62. The increased effect on GFP-P62ΔPB1 is also in keeping with our suggestion that there is competition between aggregation of p62 at autophagosomes and nucleolar targeting of the protein.

Next, we explored the role of the LC3-interacting region (LIR) and the ubiquitin-binding (UBA) domain in nucleolar transport of p62 (Fig. 5A). The UBA is required for p62 to bind to ubiquitinated proteins. LC3 binding, via the LIR, is essential for incorporation into autophagosomes. We generated an LC3 mutant (D₃₃₅A/D₃₃₆A/D₃₃₇A) that decreases P62-LC3 binding by 75% (GFP-P62ΔPB1mtLC3; ref. 34), and a C-terminal truncated version of the protein missing the UBA domain (GFP-P62ΔPB1ΔUBA). The mutants were generated in GFP-P62ΔPB1, as nucleolar translocation of this protein was not impeded by cytoplasmic aggregation. Analysis of fixed cells indicated that GFP-P62ΔPB1mtLC3 and GFP-P62ΔPB1ΔUBA accumulate in nucleoli in response to aspirin in a similar manner to GFP-P62ΔPB1 (Fig. 5D), suggesting these domains are also not essential for nucleolar targeting of the protein.

Nucleolar localization of multiple proteins is dependent on the presence of a specific nucleolar localization signal (NoLS; refs. 37–39). To determine whether p62 has such a signal, we submitted the sequence to an NoLS prediction program (40). Using this approach, we identified a domain of interest between aa 94 and 116 (IFRIYIKEKKECRRDHRPPCAQE; Fig. 6A). This domain is located just after PB1, preceding the nuclear localization and nuclear export signals (34). Alignment software revealed a stretch of basic amino acids within this region (aa 100 and 110, KEKKECRRDHR) that is common to known NoLSs (Fig. 6B). Bioinformatic analysis also revealed that this domain is conserved between species and is evident down to Danio Rerio, suggesting a strong selection pressure for the region (Fig. 6C).

To determine the role of the putative NoLS in nucleolar translocation of p62, we further deleted GFP-p62ΔPB1 to create a construct that was deleted for residues 1–120 (GFP-p62ΔPB1ΔNoLS). GFP-p62ΔPB1ΔNoLS was transfected into SW480 cells and its localization in the presence and absence of aspirin compared with GFP-p62ΔPB1. These studies revealed that both GFP-p62ΔPB1 and GFP-p62ΔPB1ΔNoLS were diffuse in the cytoplasm prior to aspirin treatment and translocated from the cytoplasm to the nucleoplasm in response to the agent (Fig. 6D). However, although GFP-p62ΔPB1 accumulated in nucleoli, GFP-p62ΔPB1ΔNoLS remained in the nucleoplasm (Fig. 6D). Image quantification confirmed that the significant increase in nucleolar GFP-p62ΔPB1 observed in response to aspirin was lost upon deletion of the putative NoLS (Fig. 6E). Western blot analysis confirmed that GFP-p62ΔPB1ΔNoLS is expressed at a similar level to GFP-p62ΔPB1 and other mutants, excluding differential expression level as an explanation for the differential localization (Fig. 6F). These data show for the first time that p62 has a bona fide nucleolar localization signal that is required for transport of the protein from the nucleoplasm to the nucleolus under conditions of stress.

Immunocytochemistry with antibodies to endogenous p62 revealed that expression of GFP-p62ΔPB1ΔNoLS blocks nucleolar translocation of endogenous protein (Fig. 6G), which is unexpected given that GFP-p62ΔPB1ΔNoLS does not have a self-binding domain. Nevertheless, it does indicate this mutant can be used as a tool to explore the role of p62 in nucleolar transport of cargo proteins.

We next utilized the p62 ΔNoLS deletion mutant to directly test whether the nucleolar localization of p62 is causally involved in nucleolar translocation of RelA. Using RelA immunocytochemistry performed on transfected SW480 cells, we found, as expected, that cells expressing GFP-P62ΔPB1 showed nucleolar localization of both p62 and RelA in response to aspirin (Fig. 7A). However, in cultures expressing GFP-P62ΔPB1ΔNoLS, there was a significant reduction in transfected cells showing nucleolar RelA (Fig. 7B). Instead, both p62 and RelA remained in the nucleoplasm (Fig. 7A). In contrast, adjacent non-transfected cells showed aspirin-mediated nucleolar accumulation of both p62 and RelA (Fig. 7A). Similar results were obtained using SW620 cells in that expression GFP-P62ΔPB1ΔNoLS abrogated aspirin-induced RelA nucleolar accumulation (Supplementary Fig. S6A–S6C). Together, these results provide compelling evidence that p62 is required to transport RelA from the nucleoplasm to nucleoli.

To determine whether the reciprocal is true, we utilized depletion of COMMD1 as we have previously shown this very robustly and specifically stops the nucleoplasmic–nucleolar translocation of RelA. Indeed, we found that the aspirin-induced nucleolar accumulation of RelA was blocked upon depletion of COMMD1 (Supplementary Fig. S6D). However, depletion of COMMD1 had no effect on nucleolar localization of p62 (Supplementary Fig. S6D), suggesting RelA does not transport p62 to nucleoli.

RelA accumulates in nucleolar aggregates alongside translation factors, chaperones and targets of the ubiquitin proteasome pathway (Fig. 2). Therefore, we considered the role of p62 in nucleolar shuttling of other aggregate proteins. We initially focused on SUMO-2/3 as this is a classic component of nucleolar aggresomes (26). Immunocytochemistry indicated clear colocalization between SUMO-2/3 and p62 in nucleolar puncti in response to aspirin (Fig. 7B). Furthermore, expression of GFP-P62ΔPB1ΔNoLS blocked nucleolar translocation of SUMO-2/3 in a similar manner to RelA (Fig. 7C). Next, we explored the role of p62 in nucleolar transport of EIF4H, which shows a more diffuse distribution following translocation into the compartment (Fig. 2D). In contrast to RelA and SUMO-2/3, we found little correlation between cells that show nucleolar p62 and those that show nucleolar EIF4H (Supplementary Fig. S6E). Furthermore, expression of GFP-P62ΔPB1ΔNoLS had little effect on nucleolar translocation of EIF4H (Fig. 7D).

We have previously demonstrated that RelA acts directly on nucleolar stress pathways to mediate apoptosis (41). Given that expression of GFP-P62ΔPB1ΔNoLS inhibits aspirin-mediated nucleolar translocation of RelA, we considered it may also inhibit the apoptotic effects of the agent. Indeed, we found that aspirin-mediated apoptosis was significantly reduced in the population of cells transfected with GFP-P62ΔPB1ΔNoLS, compared with those transfected with GFP-P62ΔPB1 (Fig. 7E).

Together, these results reveal that p62 plays a critical role in transporting aggresome-related proteins, including RelA, to nucleoli. They also suggest that this effect is specific and there are p62-dependent and independent mechanisms of nucleolar transport. Furthermore, they suggest that p62 transport of aggresome-related proteins to nucleoli in response to stress is causally involved in the regulation cellular homeostasis.

The NF-κB transcription factor is one of the most important regulators of the cellular life/death balance and its aberrant activation is associated with many of the hallmarks of cancer. Therefore, identifying means to eliminate cells with aberrant NF-κB activity could have a major therapeutic benefit. One such mechanism is nucleolar sequestration of RelA, which actively mediates apoptosis. Indeed, nucleolar sequestration has emerged as a broad mechanism for fine tuning cancer-related pathways and maintaining cellular homeostasis. However, understanding of the signals that direct RelA and other extranucleolar proteins to the organelle is limited. Here, we report the novel observation that the autophagic cargo receptor, p62, localizes to nucleoli. Furthermore, we show that p62 is responsible for transporting specific nuclear proteins, including RelA, to nucleolar aggresomes, thus inducing apoptosis. These data identify a new role for p62 in the nucleus and shed further light on mechanisms underlying nucleolar sequestration of regulatory proteins that could be used to develop a novel class of therapeutics.

Using fluorescent correlation spectroscopy and quantitative proteomics, we show that aspirin and MG132 cause RelA to accumulate in slow moving nucleolar complexes alongside aggregate-related proteins, suggestive of nucleolar aggresomes. Our conclusion that p62 transports RelA to these aggresomes is based on the following observations. First, we demonstrate that p62 colocalizes with RelA and sumoylated proteins in these nucleolar foci. Second, we demonstrate that p62 interacts with RelA prior to aggresome formation. We show a correlation between cells that have nucleolar p62 and those that have nucleolar RelA. Finally, we demonstrate that expressing a dominant negative mutant of p62, unable to translocate to nucleoli, specifically blocks nucleolar translocation of RelA. This mutant also blocks nucleolar transport of the aggresome-related protein, SUMO2/3, but has no effect on the nucleolar translocation of EIF4H, which does not aggregate in nucleoli but is diffuse within the organelle. It will now be of considerable interest to identify the mechanisms by which p62 transports cargo proteins to nucleoli and to understand the determinants of specificity. Given that ubiquitination of RelA is known to be essential for transport of the protein to nucleoli (17), we would predict that p62 transports ubiquitinated RelA to the organelle via its UBA domain. However, our studies to determine the role of this domain in nucleolar transport of RelA were inconclusive, suggesting it may be partially involved (data not shown).

The mechanisms underlying nucleolar translocation of p62 appear to be distinct from those involved in transport of the protein to autophagosomes. For example, dimerization via the PB1 domain, which is required for autophagosomal translocation, is unnecessary for nucleolar transport (34). It is not enhanced by bafilomycin treatment, which is in keeping with other studies showing nucleolar aggresome formation is unaffected by inhibition of autophagy (42). Furthermore, unlike accumulation at autophagosomes, mutating the LC3 domain has little effect. Indeed, our data would suggest that there is competition between autophagic and nucleolar accumulation of the protein. One domain we did reveal to be essential for translocation of p62 from the nucleoplasm to nucleoli was aa 94–110. This domain is conserved among species and aligns with known NoLSs. We considered that deleting this domain may disturb an essential protein–protein interaction and/or posttranslational modification. However, no proteins are documented to bind to this region and prediction programs suggest it is not modified by PTMs. One mechanism suggested for nucleolar accumulation of extranucleolar proteins is detention within the organelle via a specific nucleolar detention sequence (NoDS; ref. 43). However, alignment analysis suggests the domain of p62 essential for nucleolar translocation is most likely an NoLS, rather than an NoDS. It will now be interesting to determine what proteins bind to the NoLS of p62 in response to proteotoxic stress and how this signals for nucleolar translocation.

The idea that nucleolar aggresomes act as a means to maintain nuclear proteostasis has gained considerable traction in recent years. Recently, Frottin and colleagues (2019; ref. 44) demonstrated that upon heat stress, misfolded proteins are sequestered in the granular component of nucleoli in liquid phase. Under mild stress, this process is transient and allows misfolded proteins to be maintained in a folding competent state until proteotoxic stress is relieved. However, prolonged stress leads to a transition from liquid-like to solid phase, with loss of reversibility and dysfunction in protein quality control. Others have also suggested that the formation of nucleolar aggresomes can be both transient or irreversible, depending on the stress and context (26). We have previously demonstrated that RelA has a functional impact on proteins within nucleoli that leads to apoptosis (41). Here, we show that inhibiting the transport of RelA to this compartment, using the p62ΔNoLS mutant, abrogates the apoptotic effects of aspirin. Therefore, we suggest that that the RelA-p62 nucleolar aggresomes we observe are not transient, but allow elimination of cells that have acquired irreversible defects in nuclear protein quality control.

We have previously demonstrated that nucleolar translocation of RelA is not only involved in the direct induction of apoptosis, but is causally involved in repression of NF-κB–driven transcription. Although expression of the p62ΔNoLS mutant inhibited nucleolar translocation of RelA and aspirin-mediated apoptosis, we found it had no effect on aspirin-mediated repression of NF-κB–driven transcription (data not shown). Ubiquitination of RelA is a prerequisite for nucleolar translocation, which in itself renders RelA inactive (17). Therefore, the lack of effect of GFP-P62ΔPB1ΔNoLS on NF-κB–driven transcription may be explained by a lack of effect on RelA ubiquitination. We are now focused on unravelling the complexities and dynamics of stress-induced ubiquitination of RelA, p62 binding, and nucleolar translocation. One area of interest is the role of LRRC25 because it facilitates the interaction between RelA and p62 and induces repression of NF-κB–driven transcription (35).

Overwhelming evidence indicates aspirin has antitumor properties and the potential to prevent colorectal cancer (45). However, the agent cannot be fully recommended for prevention purpose due to its side effect profile. It is now critical to understand how aspirin acts against cancer cells so that biomarkers of response can be identified. We have previously demonstrated that an early response to aspirin exposure is inhibition of rDNA transcription and the induction of an atypical form of nucleolar stress (29). Here, we demonstrate the agent has a particularly significant effect on the nucleolar proteome, compared with the cytoplasmic and nucleoplasmic proteomes. We also show aspirin causes an accumulation of insoluble proteins in colon cancer cells and that these form aggresomes within the organelle. Interestingly, we observe this effect both in primary and metastatic colon cancer cells. Together, these data suggest that nucleoli play a complex but significant role in the antitumor effects of this agent and that nucleoli could act as a good therapeutic target and biomarker of aspirin response.

In summary, we identify a new role for p62 in maintaining nuclear proteostasis by transporting regulatory proteins to nucleolar aggresomes. Given the role of nuclear NF-κB in disease, the challenge now is to understand the specifics of proteotoxic stress leading to nucleolar aggregation of p62-RelA, so that we can specifically target RelA to this compartment, thus eliminating diseased cells.

P. Morin reports grants from Medical Research Council during the conduct of the study. R.R. Duncan reports grants from Medical Research Council during the conduct of the study. L.A. Stark reports grants from Medical Research Council during the conduct of the study. No disclosures were reported by the other authors.

I.T. Lobb: Data curation, formal analysis, investigation. P. Morin: Conceptualization, data curation, formal analysis, investigation, writing–original draft. K. Martin: Investigation. H.C. Thoms: Investigation, carried out experiments for revised manuscript. J.C. Wills: Data curation, investigation. X. Lleshi: Investigation. K.C.F. Olsen: Investigation. R.R. Duncan: Formal analysis, supervision. L.A. Stark: Conceptualization, data curation, formal analysis, supervision, funding acquisition, writing–original draft, writing–review and editing.

We would like to thank Terje Johansen (UIT, Artic University of Norway) for kindly providing the GFP-p62 plasmid. Simon Wilkinson (ECRC, University of Edinburgh) for providing tools, reagents, and advice, and Alison Dunn for expertise and help with FCS. C. Nicol provided help with figure preparation. The work was funded by WWCR (formally AICR; 10-0158 to L.A. Stark), Rosetrees Trust (A631, JS16/M225 to L.A. Stark), MRC (MR/J001481/1 MR/K01563X/1 to R.R. Duncan), University of Edinburgh scholarships (to I.T. Lobb), and BBSRC (BB/S018530/1).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.