Nucleolar Targeting of RelA(p65) Is Regulated by COMMD1-Dependent Ubiquitination

The NF-κB transcription factor generally exists as a heterodimer of the p50 and RelA(p65) polypeptides bound in the cytoplasm by the inhibitor protein IκB (1, 2). Following cellular stimulation, IκB is phosphorylated and then degraded by the proteasome, allowing NF-κB to translocate to the nucleus and regulate the expression of downstream genes (3). A central function of NF-κB, in particular the RelA subcomponent, is the regulation of cellular growth and apoptosis (4, 5). However, this is complex because stimulation of NF-κB can have both proapoptotic and antiapoptotic effects, dependent on the context. For instance, persistent activation of NF-κB is a recognized contributory factor in a number of common cancers, whereas stimulation of the pathway by agents such as UV radiation, serum withdrawal, and cell stress promotes cell death (6, 7).

Although the main switch in NF-κB activation is cytoplasmic release from IκB, NF-κB is also controlled within the nucleus by a number of complex mechanisms that are still not fully understood (8–10). These post-induction mechanisms of regulation determine the genes that are activated/repressed by a signal and whether the signal inhibits or mediates apoptosis. One post-induction mechanism that we have identified for regulating NF-κB transcriptional activity and apoptosis is the nuclear/nucleolar distribution of RelA. We have shown that in response to specific stress-inducing stimuli, including UV-C radiation, serum deprivation, and aspirin and related agents, RelA translocates from the cytoplasm to the nucleoplasm and then to the nucleolus (11–13). We identified a signal at the NH₂ terminus of RelA [amino acids (aa) 27–30] that is required for nucleolar translocation of the protein and showed, using a dominant-negative mutant deleted for this domain, that nucleolar translocation of RelA is causally involved in the downregulation of NF-κB–driven transcription and the induction of apoptosis. In contrast, we found that RelA accumulates in the nucleoplasm, excluded from the nucleolus, in response to the cytokine tumor necrosis factor α (TNFα). Moreover, NF-κB/RelA induced by this agent activates transcription and inhibits apoptosis (11). Understanding the pathways that regulate nuclear RelA distribution could inform on the design of molecules that modulate these pathways and, thus, NF-κB transcriptional activity and apoptosis.

Although nucleolar localization signals are thought to be involved, the signals that regulate the nucleolar localization of proteins have yet to be fully defined (14–16). One pathway that has been implicated is the ubiquitin-proteasome system. Proteins such as c-myc, p53, and ADP ribosylation factor translocate to the nucleolus in response to proteasome inhibition, leading to the suggestion that nucleolar compartmentalization regulates protein stability (17–20). This is of particular interest because RelA undergoes ubiquitination and proteasome-mediated degradation in the nucleus (21–26). Therefore, we set out to determine the role of the ubiquitin-proteasome system in the nuclear/nucleolar distribution of RelA and apoptosis.

SW480 colon cancer cell culture, HeLa cervical cancer cell culture, and treatment with aspirin have been described (11, 27). SW480 cells that stably express GFP-RelA WT or Δ27–30 were generated by selection in G418. The proteasome inhibitors MG132 and lactacystin (Calbiochem), TNF and TNF-related apoptosis-inducing ligand (TRAIL; R&D Systems), and calyculin A (Cell Signaling Technology) were prepared as per manufacturers' instructions. Cell fractions were isolated from SW480 cells as described previously (11, 13).

Proteasome activity was measured in lysates of cellular fractions (25 μg) using a 20S Proteasome Fluorometric (AMC) Assay Kit as per manufacturer's instructions (Chemicon). For each sample, standards provided were used to convert relative fluorescent units to micromolar concentrations of free AMC, and then results were presented as percentage of basal (nontreated) proteasome activity. All results shown are the mean of duplicates of at least three independent experiments with SE. NF-κB reporter assays have been described elsewhere (11).

Extracts (30 μg) were resolved on 5% to 10% SDS polyacrylamide gels, and immunoblots done following standard procedures. The following primary antibodies were used: p65(C-20), p65(F-6), nucleolin (C23), fibrillarin, glutathione S-transferase (GST), and green fluorescent protein (GFP; all from Santa Cruz); His [generated in-house (purified form)]; ubiquitin (rabbit antibody; Dako Cytomation); ubiquitin (mouse monoclonal; Stressgen); filamin A (Bethyl Laboratories); phospho-p65 (Ser⁴⁶⁸) and phospho-p65 (Ser⁵³⁶) (Cell Signaling Technology).

The wild-type (WT) and R7 ₆His-ubiquitin (Ub) plasmids, HA-Ub, and the 3enhancer CONA (3xκB ConA-Luc) reporter construct were supplied by Prof. R. Hay (28, 29). The pCMV-β-galactosidase plasmid is commercially available (Promega). The GFP expression vectors used have been described (11). The GST-COMMD1 expression vector and corresponding control were gifts from Dr. E. Burstein and have been described (30).

Plasmid transfections were carried out using Lipofectin (Life Technologies, Inc.; ref. 11). After 24 to 48 h of recovery, cells were treated with aspirin or proteasome inhibitors as specified in the text.

For transfection of small interfering RNA (siRNA), cells were transfected with control (scrambled) or COMMD1 [#1, AAGTCTATTGCGTCTGCAGAC (30); #2, GACAGGAATCTGAATTTCTGT] siRNA using Lipofectamine 2000 (Life Technologies) as per manufacturer's instructions. Cells were recovered for 48 h before treatment with aspirin.

Immunoprecipitation assays were done using 500 μg of whole-cell lysate as previously described (31). Rabbit polyclonal anti-RelA (Santa Cruz) and anti-ubiquitin (Stressgen) were used to immunoprecipitate the appropriate protein. Mouse or rabbit IgG (preimmune serum) was used as control. Complexes were resolved by SDS-PAGE and then analyzed by Western blot analysis. Precipitation of ₆His-tagged Ub using Ni-agarose (29) and GST-COMMD1 using glutathione-Sepharose (30) was carried out as previously described. Ubiquitinated and COMMD1-bound RelA were detected using anti-RelA Western blot analysis. Anti-histidine and GST were used as controls.

The localization of GFP-tagged proteins was analyzed in snap shots of live adherent cells using an Axiovert 100 inverted microscope (Zeiss) at an excitation wavelength of 488 nm. Images were captured and processed using IPLab 3.6 with scripts written in-house.

Immunocytochemistry was done using antibodies against RelA, nucleolin, and fibrillarin as described (11). Fluorescent microscopy was done with a Zeiss Axioplan microscope, 63× Plan Neofluor objective, and Chroma 83000 filter set. Images were captured using IPLab Spectrum 3.6. The percentage of cells showing nucleolar RelA was determined in at least 200 cells from at least five fields of view in three independent experiments or as specified.

Apoptosis assays were carried out on cells transfected with GFP-tagged constructs as described (11). At least 250 transfected cells (detected in FITC channel) were examined for Annexin V binding (detected in Texas red channel) for each sample. All results presented are the mean of at least three experiments ± SE.

We hypothesized that nucleolar accumulation of RelA may be caused by inhibition of proteasome activity in the nucleolus itself (17). Therefore, using aspirin as a model stress-inducing stimulus of the NF-κB pathway, we investigated the association between nucleolar accumulation of RelA and proteasome activity. We found that, as previously described (32), aspirin mediated a partial reduction in total cellular proteasome activity (Fig. 1A). This reduction was biphasic and was initially observed within 30 minutes of treatment, which preceded an accumulation of ubiquitinated proteins within the cell (Fig. 1B and C). Consistent with our finding that aspirin stimulates the NF-κB pathway, Western blot analysis indicated that partially inhibiting proteasome activity (using low-dose proteasome inhibitor MG132) was insufficient to block signal (TNF)-mediated degradation of IκBα (data not shown).

Next, we measured proteasome activity in the nuclear (which, given the insolubility of nucleoli, predominantly represents nucleoplasmic proteins) and nucleolar fractions. The purity of isolated nucleoli was confirmed by light microscopy (data not shown) and by Western blot analysis, which showed an enrichment for C23 and fibrillarin (nucleolar markers) and reduced lamin B (a nonnucleolar protein) in the nucleolar fraction (Fig. 1D). We found that basal levels of proteasome activity were low in nucleoli and increased in response to aspirin, whereas a decrease was observed in nuclei (Fig. 1D). These data indicate that RelA accumulation in the nucleolus is not a consequence of inhibition of proteasome activity in this compartment. However, decreased proteasome activity in the nucleoplasm may be a factor.

We next determined whether the effects of aspirin on proteasome activity modulate ubiquitination of RelA. We found that aspirin induces a dose-dependent increase in high molecular weight forms of RelA, suggestive of the addition of multiple ubiquitin chains (Fig. 2A). Immunoprecipitation of endogenous RelA followed by anti-ubiquitin immunoblot confirmed increased amounts of ubiquitinated RelA in response to the agent (Fig. 2A). This was further confirmed using overexpression of ₆His-tagged ubiquitin followed by precipitation of ubiquitinated proteins with Ni-agarose beads. Figure 2B shows that aspirin and MG132 (used as a positive control refs. 24, 25) mediate an increase in ubiquitinated forms of RelA in cells transfected with WT ₆His-ubiquitin. This was not observed in cells transfected with R7-₆His-ubiquitin (which is mutated at all lysine residues and thus cannot form polyubiquitin chains), indicating specificity (Fig. 2B). These data confirm that, like MG132, aspirin mediates the ubiquitination of RelA.

On examining the kinetic relationship between aspirin events, we found that ubiquitination of RelA coincided with nucleoplasmic accumulation of the protein, which precedes its appearance in nucleoli (Fig. 2C; ref. 11). Similar to aspirin, we observed an increase in high molecular weight forms of RelA in response to UV-C and found that this ubiquitination peaked when the protein was nucleoplasmic, prior to nucleolar translocation (Fig. 2D). In contrast, TNF, which induces RelA that remains in the nucleoplasm, had minimal effect on its ubiquitination status (Fig. 2D; ref. 11).

The above data suggest that an accumulation of ubiquitinated forms of RelA in the nucleoplasm may act as a signal for nucleolar targeting of the protein. To further test this hypothesis, we examined the cellular localization of RelA in response to MG132. We found that RelA is targeted to nucleoli [as indicated by the nucleolar marker nucleolin (C23) and areas devoid of 4′,6-diamidino-2-phenylindole (DAPI) staining] after treatment with this agent (Fig. 3A). Similar results were observed with the highly selective proteasome inhibitor lactacystin, confirming that proteasome inhibition induces the nucleolar translocation of RelA (Fig. 3A). Furthermore, as for aspirin/UV-C (Fig. 2C and D), time-course studies indicated that proteasome inhibitor–mediated ubiquitination of RelA occurred concomitantly with nucleoplasmic accumulation of the protein, which preceded translocation to the nucleolus (Fig. 3B–D). Interestingly, for all stimuli, there seemed to be a decrease in ubiquitinated forms of RelA as it appeared in nucleoli. This may reflect the insoluble nature of the protein at this point or suggest that RelA is deubiquitinated once in this compartment.

We have previously shown that aa27–30 are essential for aspirin/UV-C–mediated nucleolar targeting of RelA, and that the Δ27–30 deletion mutant acts in a dominant-negative manner. Live-cell imaging and high-power fluorescent microscopy of fixed cells indicated that this domain is also required for nucleolar targeting of RelA in response to proteasome inhibition (Fig. 4A). We found that GFP-RelAWT accumulates in nucleoli (identified in bright-field images and areas devoid of DAPI staining in fixed cells) in response to lactacystin and MG132 (confirming our data generated by immunocytochemical analysis of endogenous protein), whereas GFP-RelAΔ27–30 accumulates in aggregates in the nucleoplasm and is excluded from nucleoli (Fig. 4A). These data indicate the importance of this domain in stress- and proteasome inhibitor–mediated nucleolar translocation of RelA.

To further establish the relationship between ubiquitination and the nucleolar targeting of RelA, we next examined the requirement for this domain in the ubiquitination of the protein. Using anti-RelA Western blot analysis, we found that aspirin and MG132 induced an accumulation of high molecular weight forms of RelA in cells expressing GFP-RelAWT but induced no change in the ubiquitination status of RelA in cells expressing GFP-RelAΔ27–30 (Fig. 4B). Expression of the mutant protein did not inhibit aspirin/MG132–mediated increase in total protein ubiquitination, suggesting that this is a specific effect (Fig. 4B). These data were confirmed by immunoprecipitation studies that indicated an increase in ubiquitinated RelA in response to aspirin and MG132 in cells expressing GFP-RelAWT, which was abrogated in cells expressing GFP-RelAΔ27−30 (Fig. 4C). Expression of ₆His-tagged ubiquitin in cells expressing GFP-RelA WT or Δ27−30, followed by Ni-agarose pull-down, further confirmed that aspirin induced an increase in ubiquitinated forms of GFP-RelAWT but had no effect on GFP-RelAΔ27–30.

We have previously shown that mutation of the single lysine residue within aa27–30 (K28R) has no effect on aspirin-mediated nucleolar translocation of the protein (11). We found that mutation of this lysine also had no effect on aspirin- or proteasome inhibitor–mediated ubiquitination of the protein, suggesting that this is not a direct target (Fig. 4B and C). These data show that the same motif is required for ubiquitination and nucleolar translocation of RelA, providing further evidence that the two events are linked.

Previously, we reported that nucleolar targeting of RelA regulates NF-κB–driven transcription and apoptosis (11). Consistent with these studies, NF-κB reporter assays confirmed the repression of NF-κB–driven transcription in response to proteasome inhibitors (Supplementary Fig. S1A), and showed that this repression occurred after ubiquitination of RelA, paralleling the nucleolar translocation of the protein (Fig. 3B). Annexin V apoptosis assays showed that this nucleolar translocation of RelA was also associated with induction of apoptosis (Supplementary Fig. S1B). Furthermore, we found that translocation of RelA to the nucleolus was required for the apoptotic response to these agents, as this response was completely blocked in cells expressing GFP-RelAΔ27–30 (Fig. 4D). Similar results were observed in HeLa cells, indicating that this mechanism of proteasome inhibitor–mediated apoptosis is not restricted to a particular cell type (Supplementary Fig. S2). In contrast to proteasome inhibition, we found that the apoptotic response to TNFα and TRAIL was identical in colorectal cancer cells expressing GFP-RelAWT and GFP-RelAΔ27–30, indicating that this domain of RelA is dispensable for the proapoptotic or antiapoptotic effects of these agents (Supplementary Fig. S3).

Data presented thus far show a link between ubiquitination and nucleolar translocation of RelA. To determine whether this relationship was causal, we examined the effect of modulating the RelA ubiquitin ligase complex on the nuclear distribution of the protein. Our studies focused on COMMD1 because this is a rate-limiting component of the complex (24). Using a GST-COMMD1 expression vector and immunocytochemistry, we found that increased expression of COMMD1 promotes aspirin-mediated nucleolar translocation of RelA, as determined by the number of cells in the whole population showing nucleolar RelA (Fig. 5A). We also found that siRNA-mediated knockdown of COMMD1 blocks aspirin-mediated ubiquitination and nucleolar translocation of the protein (Fig. 5B and C). In fact, we found that on knockdown of COMMD1, RelA accumulates in the nucleoplasm in response to aspirin, confirming a role for the ubiquitin ligase complex in the nucleoplasmic to nucleolar translocation of the protein (Fig. 5C). Consistent with a role for the nucleolar translocation of RelA in the apoptotic response to aspirin, Annexin V apoptosis assays indicated that overexpression of COMMD1 promotes aspirin-mediated apoptosis, whereas knockdown of COMMD1 abrogates this effect (Fig. 5D).

Having established a critical role for COMMD1 in nucleolar translocation of RelA, we determined whether aspirin modulates this protein. We found that aspirin induced a time-dependent increase in COMMD1 levels. This increase was after the effect of the agents on proteasome activity and was concomitant with the ubiquitination and nucleoplasmic accumulation of RelA, prior to nucleolar targeting (Figs. 2B and 6A). Furthermore, this aspirin-mediated increase in COMMD1 levels was associated with increased binding to RelA (Fig. 6B).

Recent studies suggest that phosphorylation of RelA at Ser⁴⁶⁸ (S468) is critical for and promotes its ubiquitination and degradation (33, 34). However, we found that treatment with the phosphatase inhibitor calyculin A blocked aspirin- and MG132-mediated ubiquitination of RelA despite promoting S468/S536 phosphorylation of the protein (Fig. 6C).

Taken together, these data provide compelling evidence that the effects of aspirin on COMMD1 promote the ubiquitination of RelA and that this acts as a signal for nucleolar translocation of the protein. They also suggest that, contrary to other RelA ubiquitination pathways reported, phosphorylation at S468 is not involved in this process.

Here we report the novel observation that RelA is targeted to the nucleolus to induce apoptosis in response to proteasome inhibition and provide convincing evidence that nucleolar shuttling of RelA, and thus the apoptotic response to specific NF-κB stimuli, is regulated by ubiquitination of the protein.

Based on the data presented here and in our previous studies, we propose the model outlined in Fig. 6D. We suggest that in response to aspirin or stress, the NF-κB pathway is stimulated in parallel with partial inhibition of proteasome activity. This proteasome inhibition allows the accumulation of COMMD1, a critical component of the RelA ubiquitin ligase complex, which itself is regulated by proteasomal degradation (24, 30). Accumulated COMMD1 binds RelA, promoting RelA ubiquitination, which, under conditions of proteasome inhibition, signals for nucleolar translocation of the protein.

It is increasingly recognized that the nucleolus acts as a critical regulator of cell growth and death through sequestration of transcription factors and death-related proteins (35). However, the signals that regulate nucleolar shuttling are still poorly understood. Our conclusion that ubiquitination of RelA acts as a signal for transport to the nucleolus comes from the following findings. First, the ability of NF-κB stimuli to induce nucleolar translocation of RelA is paralleled by their ability to induce ubiquitination of the protein (Fig. 2). Second, inhibition of proteasome activity induces ubiquitination and nucleolar accumulation of RelA (Fig. 3). Third, for all stimuli, ubiquitination of RelA precedes translocation of the protein to nucleoli (Figs. 2 and 3). Fourth, nucleolar translocation and ubiquitination of RelA share a common critical domain (Fig. 4). Fifth, binding between RelA and COMMD1 increases before nucleolar translocation of the protein (Fig. 6B). Finally, and most convincingly, knocking down the activity of the RelA ubiquitin ligase complex blocks the nucleoplasmic to nucleolar translocation of the protein (Fig. 5). We did consider that RelA may be ubiquitinated and degraded in the nucleolus, which could explain some of our data. However, in the absence of COMMD1, RelA remained in the nucleoplasm, suggesting that the ubiquitination event occurs before nucleolar translocation. Although a number of proteins have been shown to translocate to the nucleolus in response to proteasome inhibition, to our knowledge, this is the first demonstration that ubiquitination acts as a nucleolar targeting signal.

Our suggestion that RelA is ubiquitinated in the nucleoplasm is consistent with the findings of Saccani and colleagues (22) who showed that RelA is ubiquitinated and degraded in the nucleus in a DNA-binding–dependent manner, and that this is an important mechanism for switching off NF-κB transcriptional activity. Based on these and the data presented here, we propose that if RelA is ubiquitinated as a consequence of, or in conjunction with, proteasome inhibition, then the protein is targeted to the nucleolus as an alternative means of switching off NF-κB activity(11).

Our model suggests that stress agents/proteasome inhibitors cause repression of NF-κB–driven transcription by inducing nucleolar translocation of RelA. However, it may be argued that it is the ubiquitination of chromatin-bound RelA that causes repression of NF-κB–driven transcription, and not nucleolar translocation per se. Contrary to this argument, we show that at the time point showing maximum MG132-induced ubiquitination of RelA, the agent has no effect on NF-κB transcriptional activity (Fig. 3C). Similarly, there is a delay of 3 to 5 hours between the peak effects of aspirin on ubiquitination of RelA and the effects of the agent on NF-κB–driven transcription (Fig. 2; ref. 11). Furthermore, for both agents, repression of NF-κB–driven transcription parallels the appearance of RelA in nucleoli (Fig. 3C; ref. 11). Therefore, we would suggest that repression of NF-κB–driven transcription in response to these agents is indeed mediated by nucleolar translocation of the protein.

We found that aa27–30 of RelA are required for ubiquitination of the protein, and our data eliminated the possibility that this was due to ubiquitination of the lysine residue within this motif. Another possibility is that deletion of aa27–30 blocks binding of RelA to its ubiquitin ligase complex, which, in addition to COMMD1, includes elongins B and C, Cul2, and the E3 ligase SOCS1 (24). This suggestion is supported by the data of Maine and colleagues (24) showing that the NH₂-terminal 180 amino acids of RelA are important for binding to this complex. We have found that deleting this domain does not abolish RelA binding to COMMD1 (data not shown), and we are currently investigating the role of the domain in binding to other critical members of the complex.

Recent studies have identified phosphorylation at S468 as critical for COMMD1-associated ubiquitination and degradation of RelA (33, 34). However, we found that, while promoting S468 phosphorylation, phosphatase inhibition blocked aspirin- and MG132-mediated ubiquitination of RelA (Fig. 6C). Interestingly, we have previously shown that the COOH terminus of RelA (from aa311) is not required for nucleolar translocation of the protein (11). Taken together, these data would suggest that RelA ubiquitination associated with nucleolar translocation of the protein is independent of S468 phosphorylation. These data may indicate that these stress agents mediate ubiquitination of RelA by an alternative COMMD1-associated pathway to that reported.

Taken together, the data presented here identify a novel mechanism for regulating the nuclear distribution of RelA/apoptosis involving COMMD1 and RelA ubiquitination. These data have considerable relevance to cancer prevention and therapy, as well as to the post-induction regulation of RelA/NF-κB.

No potential conflicts of interest were disclosed.

We thank R.T. Hay (University of Dundee, Dundee, United Kingdom) for the gifts of the HA-Ub and ₆His-ub plasmids; E. Quarnstrom (University of Sheffield, Sheffield, United Kingdom) and E. Burstein (Ann Arbor Medical Centre, Ann Arbor, MI) for kindly providing the GFP-RelAWT and COMMD1 expression vectors, respectively; and P. Perry and C. Nicol (MRC Human Genetics Unit) for their help with microscopy and figure preparation.

Grant Support: Association of International Cancer Research (02-330), Cancer Research UK (C20658/A6656), and the Melville Trust. M.G. Dunlop is also supported by Cancer Research UK Programme Grant funding (C348/A3758 and C48/A6361).

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.