Activation of NF-κB by a novel, CDK4-regulated, nucleolar stress response pathway

The nucleolus is a multifunctional organelle that plays a critical role in maintaining cellular homeostasis under stress. However, how nucleoli sense stress and coordinate specific phenotypic outcomes, remains poorly understood. Here, we identify a novel nucleolar stress response pathway that culminates in activation of NF-κB. Using multiple approaches, we show that specific disruption of the PolI complex stimulates NF-κB signalling. Unlike the paradigm of nucleolar stress, this stimulation is not caused by inhibition of rRNA transcription. We identify a novel mechanism by which specific stresses disrupt nucleoli involving CDK4 inhibition and consequently, UBF-p14ARF-dependent degradation of the PolI complex component, TIF-IA. We show this atypical nucleolar stress response is associated with a distinctive nucleolar architecture. Furthermore, we show it lies upstream of NF-κB signalling. Finally, we explore the relevance of this pathway in response to aspirin in human clinical samples and demonstrate a correlation between TIF-IA degradation and NF-κB pathway activation. Together, these data provide a conceptual advance in understanding of nucleolar stress response with therapeutic implications.


Introduction
The nucleolus is a highly dynamic sub-nuclear organelle that, in addition to its primary function as the hub of ribosome biogenesis, acts as a critical stress sensor and coordinator of stress response 1, 2 . The first step in ribosome biogenesis is transcription of preribosomal RNA (pre-rRNA) by the RNA polymerase I (PolI) complex. Pre-RNA is then cleaved and assembled with ribosomal and accessory proteins to generate the 40S and 60S ribosomal subunits. This process is the most energy consuming in a cell and as such, is closely linked to metabolic and proliferative activity. If cells are exposed to stresses/insults that threaten this activity (e.g UV-C radiation, nutrient depletion, infection, DNA damaging or toxic agents), rRNA transcription is inhibited and a cascade of nucleolar events is triggered that will either allow the cell to repair and regain homeostasis, or, if the damage is too great, undergo cell death. The nucleolar proteome contains over 4500 proteins, more than half of which are involved in processes out-with ribosome biogenesis such as transcription, cell cycle regulation, chromatin structure, ubiquitin modification, apoptosis and proliferation 3 . It is thought to be the differential release of these regulatory proteins to the nucleoplasm/cytoplasm in response to stress that ultimately determines cell fate. Outcomes that have been linked with nucleolar perturbation include differentiation, cell cycle arrest, autophagy, DNA repair, senescence and apoptosis 4 .
Although it is well established that nucleoli play an important role in the maintenance of cell homeostasis, the mechanisms that coordinate stress effects on the PolI complex, and integrate these effects into individual phenotypic outcomes, remain poorly understood. The most recognised downstream consequence of nucleolar stress is stabilisation of p53 and the pathway leading to this stabilisation has been widely reported [5][6][7] . However, it is increasingly apparent that perturbation of nucleolar function can regulate cell phenotype in a p53 independent manner 8 . Spatial proteomic studies, demonstrating hundreds of nucleolar proteins with p53 independent functions relocate from the nucleolus in response to stress, also support this notion 9,10 . Similar to p53, the NF-κB transcription factor plays a critical role in maintaining cellular homeostasis under stress 11 . The most abundant form of NF-κB is a heterodimer of the p50 and RelA polypeptides, which is generally bound in the cytoplasm by the inhibitor, IκBα 12 . Upon exposure of the cell to a myriad of stress signals, IκBα is degraded allowing NF-κB to translocate to the nucleus where it regulates expression of target genes [13][14][15] . The most recognised NF-κB stimuli are pro-inflammatory cytokines, which induce rapid/transient NFκ B activation by mechanisms that are well-defined. In contrast, stress stimuli (including serum deprivation, UV-C radiation and chemotherapeutic/preventative agents) induce slow, prolonged activation of NF-κB by mechanisms that remain poorly understood.
A common denominator of stresses that activate NF-κB is disruption of nucleoli (some of which are summarised in Supplemental Table 1) [16][17][18] . Furthermore, proteins that have a role in stress-mediated activation of NF-κB reside within this organelle. For example, CK2, which forms part of the PolI complex and facilitates rDNA gene transcription, phosphorylates IκB in response to UV-C 19,20 . EIF2α is also a nucleolar protein that plays a role in NF-κB activation in response to multiple stresses 21,22 . Previous work in this lab has shown that post induction, RelA can accumulate in nucleoli 23,24 . However, the relationship between stress-mediated nucleolar disruption and activation of the cytoplasmic NF-κB pathway has not yet been considered.
Here we define a novel, p53 independent nucleolar stress response pathway that triggers activation of NF-κB. We firstly show that artificial disruption of the PolI complex induces degradation of IκB and consequently, increased transcription of NF-κB target genes.
Unlike classical nucleolar stress, this effect is not caused by inhibition of rRNA transcription.
We show that multiple stress stimuli of NF-κB induce degradation of the critical PolI com-plex component, TIF-IA. We identify an upstream mechanism for this degradation involving CDK4-UBF-p14ARF, and show that increased nucleolar area and activation of the NF-κB pathway lie downstream of this degradation. Finally, we show an association between TIF-IA degradation and stimulation of NF-κB in response to pharmacological concentrations of aspirin in fresh, surgically resected human colorectal tumours, suggesting this pathway is triggered in a whole tissue setting and has relevance to the anti-tumour effects of the agent.

Direct disruption of nucleolar function activates the NF-κB pathway
To explore the link between stress effects on nucleoli and activation of NF-κB signalling we started by directly inhibiting upstream binding factor (UBF), an essential component of the PolI pre-initiation complex. Such inhibition was previously used by Rubbi and Milner to show nucleolar stress stabilises p53 5 . To disrupt UBF activity we initially utilised an inactive mutant (Flag-UBF-S388G) that is known to block rRNA transcription 25 .
We found that compared to expression of wild type UBF, expression of the mutant induced a significant increase in nuclear RelA, increased S536 phosphorylated RelA (an independent marker for cytoplasmic activation of the NF-κB pathway) and a greater than 5 fold increase in NF-κB-driven transcription (Figs. 1a to c). siRNA to UBF confirmed that inactivation of this protein stimulates the NF-κB pathway, as evidenced by increased nuclear RelA, S536 phosphorylated RelA and an increase in NF-κB-driven transcription comparable to that observed when cells were treated with the classic NF-κB stimulus, TNF ( Fig. 1d and Supplemental Fig. 1a). To determine whether this effect was a general consequence of PolI disruption, or restricted to inhibition of UBF, we utilised siRNA to two further complex components, PolR1A (RPA194 subunit) and TIF-IA (Rrn3p). We found depletion of these proteins also induced degradation of IκB, increased pRelA S536 and increased NF-κB-driven transcription (Figs. 1d and e and Supplemental Fig. 1a).
To further confirm that directly disrupting nucleoli activates NF-κB signalling we utilised a reporter plasmid in which transcription of the luciferase gene is driven by the full length promoter of the classic NF-κB target, IκBα. Figure 1f demonstrates that depletion of UBF induces a more than five-fold increase in transcription from the IκBα promotor.
Depletion of TIF-IA also caused a significant increase in IκBα transcription. However, the increase in trabscription was abrogated when an equivalent reporter plasmid lacking κ B sites was utilised. qRTPCR confirmed a significant increase in transcription of IκBα and another NF-κB target gene, Bcl-xl in response to PolI complex disruption (Fig. 1g).
As would be expected, siRNA to UBF, PolRIA and TIF-IA inhibited rRNA transcription (Supplemental Fig. 1b). To explore the role of this inhibition in activation of the NF-κB pathway we utilised low dose actinomycin D (specifically inhibiting PolI activity) and two highly specific small molecule inhibitors of this process, CX5461 and BHM-21 26, 27 . We found all these agents induced a significant decrease in rRNA transcription (Supplemental Fig. 1c). They are also known to stabilise p53 26, 27 . However, they had no effect on NF-κBdriven transcription (Fig. 1h).
Given that disrupting nucleolar function causes dramatic changes to the nuclear proteome 10 , we considered that depleting PolI complex components may stimulate NF-κB transcriptional activity in the absence of cytoplasmic release of NF-κB. However, blocking cytoplasmic to nuclear translocation of NF-κB, using cells we generated that constitutively express super repressor (non-degradable) IκBα, blocked the effects of TIF-IA/UBF depletion on NF-κB-driven transcription ( Fig. 1i and supplementary Fig. 1d). These data confirm that IκBα degradation is an essential step in nucleolar stress mediated induction of NF-κB transcriptional activity.
Taken together, these data indicate that direct disruption of the PolI complex activates the cytoplasmic NF-κB pathway. They also suggest that unlike the paradigm of nucleolar stress, this activation is not a consequence of inhibition of rRNA transcription per se, but due to a specific type of PolI complex disruption. To test this hypothesis, and further explore the link between nucleoli and stress-mediated activation of NF-κB, we investigated the effects of NF-κB stress stimuli on proteins of this complex.

κ B pathway
TIF-IA is known as the PolI complex component that senses stress/environmental signals and transfers these signals to the PolI transcription machinery 28,29 . It also plays an important role in nucleolar structure and in the regulation of cell growth and death. Therefore, we started by investigating this protein. The first stress stimuli we utilised was aspirin as we are interested in its pro-apoptotic activity and have shown it stimulates NF-κB in the delayed/prolonged manner characteristic of multiple inducers 30 .
Surprisingly, we found aspirin not only induced a decrease in phosphorylated TIF-IA (at the critical Ser649 residue), which is a known response to environmental changes, but also a substantial reduction in native TIF-IA (Fig.2a), which is not a reported stress response. This reduction was observed in multiple cell types and was independent of p53 status In contrast, CX-5461, BMH21, actinomycinD and TNF had a minimal effect on TIF-IA levels ( Fig. 2e).
The hallmarks of classical nucleolar stress are inhibition of rRNA transcription and segregation of nucleolar proteins 1 . As aspirin and actinomycinD had differing effects on TIF-IA, we considered they may have differing effects on these hallmarks. qRTPCR for the 47S pre rRNA transcript, 5-fluorouridine (FUrd) run on assays and immunocytochemistry revealed that while both agents caused a significant reduction in rRNA transcription and relocalisation of nucleolar marker proteins, they had distinct effects on nucleolar architecture (Figs. 2f to j and supplemental Fig. 2f). Aspirin induced a decrease in nucleolar number and a significant increase in nucleolar area. In contrast, actinomycinD induced a significant reduction in nucleolar area (Fig. 2h). Time course studies indicated that aspirin effects on rRNA transcription and nucleolar area paralleled loss of TIF-IA, suggesting this distinctive nucleolar phenotype may be associated with this loss (Fig. 2i). UV-C induced a similar nucleolar phenotype to aspirin, in keeping with this suggestion (Fig. 2j).
Aspirin and UV-C generate ceramide, a crucial lipid second messenger that induces its cytotoxic effects through activation of NF-κB 31-34 . On examination we found that the C2 and Taken together, these data indicate that specific stimuli of the NF-κB pathway induce a distinctive type of nucleolar stress that is associated with degradation of TIF-IA and enlargement/segregation of nucleoli. They also suggest a possible link between this distinctive nucleolar phenotype and stimulation of the NF-κB pathway. To help clarify this link, we investigated the mechanism underlying stress-mediated TIF-IA degradation.

Identification of p14ARF as a regulator of TIF-IA stability.
Previous studies demonstrated that basal TIF-IA is regulated by proteasomal degradation, facilitated by MDM2 or Sug-1 16,35 . However, we found that chemical inhibition of the proteasome (using MG132 or lactacystin) could not block aspirin-mediated degradation of TIF-IA ( Fig. 3a and supplemental Figs 3a to c). We also found that, although the MDM2 inhibitor Nutlin-3 increased basal TIF-IA as reported 16 , it had no effect on aspirinmediated degradation of the protein (Supplemental Fig. 3c). Similarly, the basal interaction observed between TIF-IA and Sug-1 did not change substantially after aspirin exposure (Supplemental Fig. 3d).
These data eliminated proteasomal degradation and so we considered a role for lysosomal degradation. However, we found that chemically inhibiting lysosome activity alone also had no effect on aspirin-mediated TIF-IA degradation ( Fig. 3a and Supplementary Figs e and f). Indeed, we found that the observed degradation of TIF-IA could only be blocked by inhibiting proteasomal and lysosomal activity together, indicating redundancy in the degradation pathway (Fig. 3a).
As lysosomal degradation is not reported within nucleoli, we generated a series of GFP-tagged deletion and site mutants to investigate the cellular localisation of TIF-IA degradation, and to further explore the mechanism underlying this degradation ( Supplementary Fig. 3g). These data revealed that amino acids 94-204 are sufficient for aspirin-mediated degradation of TIF-IA, but that the recognised phosphorylation sites within this region (S170/2, S199 and T200) are dispensable ( Fig. 3b and supplemental Fig. 3g).
They also revealed that GFP-TIF-IA 94-302 and 94-204, which localise in the nucleoplasm/cytoplasm, undergo aspirin-mediated degradation comparable to that of full length protein ( Fig. 3b and supplementary Fig. 3g for GFP control).
In a related study in the lab, aimed at identifying aspirin effects on the nucleolar proteome, we noted that an early response to this agent was a decrease in nucleolar p14ARF (unpublished data). This was of particular interest as the p14ARF tumour suppressor is known to play a role in cytoplasmic shuttling of PolI complex components 36 . It also regulates rDNA transcription in a p53/MDM2 independent manner and influences NF-κB signalling 37,

38
. Therefore, we investigated the role of this protein and found that overexpression of p14ARF enhanced aspirin-mediated degradation of TIF-IA while siRNA depletion abrogated this effect, especially at lower doses of the agent (Figs. 3c and d). Immunocytochemical analysis confirmed that the significant reduction in TIF-IA observed in response to aspirin in cells transfected with control siRNA, was lost in cells transfected with siRNA to p14ARF 1 (Fig. 3e). Importantly, TIF-IA accumulated within nucleoli in p14ARF depleted cells, suggesting a role for p14ARF in nucleolar shuttling of the protein (Fig. 3e). In keeping with this suggestion, immunoprecipitation assays revealed that TIF-IA and p14ARF interact and that this interaction is enhanced in response to aspirin in a dose and time dependent manner targets UBF, which complexes with both TIF-IA and p14ARF 37, 40 . Therefore, we considered that CDK4 inhibition may lie upstream of stress effects on p14ARF-TIF-IA 40, 41 and that UBF acts as a bridging protein in the process. To test this suggestion, we initially examined the link between CDK4 and TIF-IA using a highly specific CDK4 inhibitor (CDK4i). We have previously shown this inhibitor mimics stress effects on the NF-κB pathway 39 In keeping with our hypothesis, immunoblot analysis indicated that CDK4i Having established that inhibition of CDK4 mediates degradation of TIF-IA in a p14ARF dependent manner, we next investigated the role of UBF. We found that both CDK4i-and aspirin-mediated degradation of TIF-IA were paralleled by dephosphorylation of UBF at the known CDK4 phosphorylation site, S484 (Fig. 4g). We also found that siRNA silencing of UBF substantially abrogated aspirin-and CDK4i-mediated TIF-IA degradation (Fig. 4h). In contrast, silencing of PolRIA had no effect, suggesting this abrogation was specific and not a general consequence of PolI complex disruption (Fig. 4h).
To further explore the role of UBF S484 de-phosphorylation, we examined the effects of aspirin and CDK4i on TIF-IA degradation in cells expressing UBF mutated at this site.
SW480 cells were depleted for UBF then transfected with plasmids expressing either wild type UBF (Flag-UBFWT) or an S484 mutant that cannot be phosphorylated (Flag-UBF S484A). As would be expected from previous results (Fig. 4f), we found that depletion of UBF alone abrogated aspirin and CDK4i-mediated TIF-IA degradation ( Fig. 4i and Supplemental Fig. 4g). We also found that this degradation was enhanced by expression of wild type UBF, but completely blocked by expression of the S484A mutant ( Fig. 4i and Supplemental Fig. 4g). We did predict that, as it is constitutively dephosphorylated, expression of the mutant alone may mimic the effects of stress on TIF-IA. However, basal TIF-IA was similar in cells expressing wild type and mutant protein.
Taken together, these data suggest that stress mediated inhibition of CDK4 causes p14ARF dependent degradation of TIF-IA, and that aa484 of UBF is critical for this process.

NF-κB pathway
The results presented so far indicate that specific disruption of the PolI complex activates NF-κB signalling. They also reveal a novel mechanism by which the PolI complex is disrupted in response to stress and suggest this mechanism may lie upstream of NF-κB pathway activation. To further explore this relationship, we next blocked stress-mediated TIF-IA degradation using siRNA to p14ARF, then examined the downstream consequences. As p14ARF has previously been shown to regulate nuclear NF-κB activity, we did consider that siRNA to this protein could inhibit activation of the NF-κB pathway downstream of nucleolar disruption (Rocha et al., 2003). To test this possibility, we mimicked the effects of stress on the PolI complex (using siRNA to UBF) then analysed NF-κB activity in the presence and absence of p14ARF. We found that loss of PolI complex integrity activated NF-κB signalling to the same extent in the presence and absence of p14ARF (Supplemental Fig. 4i). These data indicate that in this stress response pathway, the role of p14ARF is upstream of PolI complex disruption.
To definitively establish a connection between inhibition of CDK4 activity, disruption of the PolI complex and activation of the NF-κB pathway we next utilised the UBF-S484A mutant that blocks aspirin and CDK4i-mediated TIF-IA degradation (Fig. 4i). SW480 cells were transfected with control or UBF siRNA prior to overexpression of Flag-UBF-WT or -S484A. Quantitative immunocytochemistry revealed that for both control and UBF siRNA, CDK4i induced a significant increase in nuclear levels of RelA in cells transfected with wild type UBF, but not in those transfected with the S484A phospho-mutant ( Fig. 5h and supplemental Fig. 4j). It also confirmed that depletion of UBF mimics the effects of stress and induces a significant increase in nuclear RelA alone. Taken together, these data provide extremely compelling evidence that stress disrupts the PolI complex through CDK4-UBFS484-P14ARF-TIF-IA and that this disruption activates the NF-κB pathway.
Based on the data presented here and that of others, we propose the model outlined in Figure 5i. We suggest that stress-mediated inhibition of CDK4 causes P14ARF to export TIF-IA for degradation, and that UBF acts as an intermediatory in this process. We propose that the resultant changes in nucleoli cause the release of specific proteins into the cytoplasm, which induce degradation of IκBα, p536 phosphorylation/nuclear translocation of RelA and consequently, changes in the NF-κB-driven gene transcription programme.

Relationship between TIF-IA degradation and stimulation of NF-κB signalling in human clinical samples
Overwhelming evidence indicates that aspirin has anti-tumour activity and the potential to prevent colorectal and other cancers 43-45 . To investigate the clinical significance of our results with regards to this activity, and to determine whether this novel nucleolar stress response pathway has relevance in a whole tissue setting, we treated biopsies of fresh, surgically resected human colorectal tumours with pharmacological doses (0-100 µM, 1h) of aspirin ex vivo (Fig. 6a). This aspirin concentration is comparable to salicylate levels we measured in plasma from patients given a short course of analgesic doses of aspirin 30 . It is also well within the reported therapeutic range (0.1-3mM).
Western blot analysis indicated that low dose aspirin induces TIF-IA degradation (> 2 fold reduction in protein levels) in over 50% of human colorectal cancers (4/7) (Fig. 6b). Low dose aspirin also activated the NF-κB pathway (as indicated by a significant increase in the percentage of cells with pRelA 536 staining) in 50 % of tumours (Fig. 6c). Importantly, for individual tumours, there was a very strong inverse correlation (r 2 =-0.85, n=6) between aspirin effects on TIF-IA protein levels and on phosphorylation of RelA (Fig. 6d). These data confirm that aspirin causes PolI complex disruption and activates the NF-κB pathway in primary human tumours and suggests a strong relationship between these two events in a whole tissue setting. These data have far reaching implications for understanding of the antitumour effects of this agent.

Discussion
The work presented here has great significance as we identify a novel nucleolar stress response pathway that activates NF-κB signalling. We show that unlike the paradigm of nucleolar stress, NF-κB activity is not triggered by inhibition of rRNA transcription per se, but by a distinct form of PolI complex disruption that is characterised by degradation of the inhibition and stress stimuli of NF-κB 47 . Therefore, we did consider that it is this cell cycle arrest that stimulates NF-κB signalling rather than nucleolar disruption per se. However, CX5461 and BMH21 induce cell cycle arrest 48 but we show they have no effect on NF-κBdriven transcription. Furthermore, we have previously shown that inhibiting the cell cycle does not mimic the effects of CDK4i on nucleoli or on the NF-κB pathway 49 . Therefore, we conclude that activation of the NF-kB pathway is a direct consequence of loss of nucleolar integrity.
In the classic nucleolar stress response pathway, inhibition of rRNA transcription triggers cytoplasmic release of RPL11/RPL5 which inhibits MDM2 to stabilise p53 27, 50 . We demonstrated that specifically inhibiting rRNA transcription has no effect on NF-κB-driven transcription. Therefore, we suggest that in this alternate pathway, it is loss of PolI complex integrity itself that triggers release of nucleolar proteins and ultimately, activation of NF-κB.
The nature of the nucleolar proteins that stimulate the cytoplasmic NF-κB are currently unknown but CK2 is an excellent candidate as it is found as part of the PolI complex 19 and phosphorylates IκBα in response to UV-C 20 . Another kinase of interest is NIK (NF-κB inducing kinase), which acts upstream of the IkappaB kinase (IKK) complex and is known to shuttle through nucleoli 51 . The role of these candidates will be the focus of future studies.
The mechanism reported for stress-mediated TIF-IA inactivation is modulation of its phosphorylation status, and a number of signalling cascades responsible for this modulation have been described 52 . The downstream consequences have also been well described. That In steady state conditions TIF-IA shuttles dynamically between the nucleolus, nucleoplasm and cytoplasm, but the mechanisms that govern this shuttling remain unclear 54 .
Our finding that GFP-tagged TIF-IA mutants that locate out-with nucleoli are degraded in response to aspirin/CDK4i, and that depletion of p14ARF blocks this degradation and causes nucleolar accumulation of the protein, suggests that P14ARF acts by transporting TIF-IA to the cytoplasm to be degraded. These data are

Cell lines and treatments
Human SW480, HRT18, RKO and HCT116 colon cancer cells, PNT pancreatic cells and Hela Cervical cancer cells, are available from the American Type Culture Collection (ATCC).
The p53 null derivative of HCT116 (HCT116 p53-/-) was a gift from Professor B Vogelstein (John Hopkins University School of Medicine, USA) and has previously been described 58 .
HRT18SR, a derivative of HRT18 cells that constitutively express a non-degradable IκBα were generated in this lab and have been described 30 . All cells lines were maintained at 5%

Immunocytochemistry, image quantification and FUrd assays
Immunocytochemistry was performed as previously described 23

Quantitative PCR
RNA was extracted from cells using RNeasy mini kit (Qiagen) following the manufacturer's instructions. Extracted RNA was purified using RQ1 RNase-free DNase (Promega) then cDNA generated using 1st Strand cDNA synthesis kit (Roche). Taqman assays (Thermo Fisher Scientific) and a LightCycler 480 system were used to quantify transcript levels.
Primers and probes used in taqman gene expression assay are given in full in additional information. The Comparative C T Method (or ∆ ∆ C T Method) was used for calculation of relative gene expression.

Immunoblotting, luciferase reporter and apoptosis assays
Immunoblotting, luciferase reporter and AnnexinV apoptosis assays were carried out as described previously 23, 30 . Primary antibodies used for immunoblots are described in full in supplemental data.

Immunoprecipitation
Immunoprecipitation assays were performed using 1mg whole cell lysate, prepared in NP40 lysis buffer. Mouse TIF-IA antibody (Santa Cruz Biotechnology) was used to immunoprecipitate the appropriate protein. Mouse IgG (pre-immune serum) was used as a control. Complexes were resolved by SDS polyacrylamide gel electrophoresis then analysed by western blot analysis.

Ex vivo treatment of tumour biopsies and immunohistochemistry
Biopsies of colorectal tumours were provided by a pathologist at the time of resection. All patients were consented and full ethical approval was in place (Scottish Colorectal Cancer         Resected colorectal tumour biopsies were immediately transferred to the lab, washed, immersed in culturing media in 96 well plates then exposed to 0-100uM aspirin for 1h. One piece of tissue was fixed for immunohistochemistry, while another was frozen for protein analysis. This was carried out for 7 patients.     Nuclear intensity RelA  T1  T2  T3  T4  T5  T6  T7   0