|Home | About | Journals | Submit | Contact Us | Français|
Aberrant activity of the NF-κB transcription factor family, which regulates cellular responses to stress and infection, is associated with many human cancers. In this study, we define a function for NF-κB in regulation of cellular respiration that is dependent upon the tumor suppressor p53. Translocation of NF-κB family member RelA to mitochondria was inhibited by p53 by blocking an essential interaction with the heat shock protein Mortalin. However, in the absence of p53, RelA was transported into the mitochondria and recruited to the mitochondrial genome where it repressed mitochondrial gene expression, oxygen consumption and cellular ATP levels. We found mitochondrial RelA function to be dependent on its conserved C-terminal transactivation domain and independent of its sequence specific DNA binding ability, suggesting that its function in this setting was mediated by direct interaction with mitochondrial transcription factors. Taken together, our findings uncover a new mechanism through which RelA can regulate mitochondrial function, with important implications for how NF-κB activity and loss of p53 can contribute to changes in tumor cell metabolism and energy production.
Aberrant activation of the NF-κB transcription factor family, together with their activators, the IκB kinases (IKK), is associated with many human diseases, including cancer, where it has been shown to regulate many tumor cell characteristics, including survival, proliferation and metastasis (1, 2). There are five members of the NF-κB family, RelA (p65), RelB, c-Rel, p50/p105 and p52/p100, which form homo- and heterodimers (3). Although there is considerable data supporting the role of NF-κB, particularly RelA, in the regulation of cancer cell survival there is little information on the possible roles of NF-κB in the regulation of cell metabolism.
Interest in altered tumor cell metabolism was instigated by Otto Warburg’s hypothesis that a defect in cellular respiration and subsequent shift to glycolysis was the initiating step in tumorigenesis (4). One mechanism that can lead to this state is loss of the tumor suppressor p53, which results in decreased oxygen consumption and increased glycolysis (5). Interestingly, this glycolytic effect can be dependent on the NF-κB family member, RelA, which in the absence of p53 resulted in enhanced expression of the glucose transporter Glut3 (6).
RelA, together with IKK subunits, has also been identified as a mitochondrial protein (7-10). Mitochondria contain a ~17kb circular genome that codes for 13 proteins, 22 tRNAs and 2 ribosomal RNAs (Supp. Fig. 2A), all of which are involved in oxidative phosphorylation (11). There are three mitochondrial transcripts. Transcription of the light strand produces a single mRNA encoding ND6 and 8 tRNA sequences. The heavy strand produces an RNA encoding the two mitochondrial rRNA sequences, 12S and 16S rRNA. together with a transcript encoding the remaining 12 mRNAs, 14 tRNAs and the 2 rRNAs, which then undergo subsequent RNA splicing to generate separate RNA species (12). NF-κB can repress mitochondrial gene expression, including cytochrome B and cytochrome C oxidase mRNA levels, following TNF or TRAIL stimulation (8, 9), although the mechanism through which this was accomplished, or whether this was a direct effect of specific NF-κB subunits, was not established. Moreover, the mechanism and consequences of NF-κB mitochondrial localization on oxidative phosphorylation and ATP production, together with how these might contribute to the switch to glycolysis observed in cancer cells, have not been clearly defined.
Rela −/− MEF cells (provided by Professor Ron Hay (University of Dundee)) were reconstituted by lentiviral infection with vector alone (Null) or with human RelA as described previously (13). H1299wtp53 cells have been previously described (14, 15). H1299 cells were purchased directly from the American Type Culture Collection (ATCC) and older stocks of H1299 cells were verified against the new cells at the Health Protection Agency by microsatellite geneotyping. U-2 OS and PC3 cells were purchased directly from the European Collection of Cell Cultures (ECACC) and were grown in DMEM as previously described (14, 15). All cells were grown to a maximum confluency of 70% and were split 1:5 (U-2 OS, PC3 and H1299 cells) or 1:10 (Rela −/− MEF cells) every 3-4 days for a maximum of 25 passages. It should be noted that the RelA−/− MEFs used in these studies possess a mutant form of p53 (16). In the text, ‘early’ refers to passages 4-13 while ‘late’ refers to passages 16-25, with cells in the ‘transition’ period not being used. Individual experiments used a mix of cells with different passage numbers within these ‘early’ and ‘late’ periods.
Mitochondrial proteins were isolated from the cells using the ProteoExtract Cytosol/Mitochondria Fractionation Kit (Calbiochem) following the manufacturer instructions. Mitochondria were further purified by centrifugation through a 19% (v/v) Percoll gradient (Sigma).
ATP levels from 5,000 cells per sample were measured using the CellTitre-Glo® Luminescent Cell Viability Assay (Promega) following the manufacturer instructions. Luminescence was measured following 2hr incubation using a FLUOstar Omega microplate reader (BMG Labtech). Results were expressed as average relative luciferase units (RLU) per cell.
Oxygen consumption from 50,000 cells per sample was measured in a 96 well BD Oxygen Biosensor System (BD Biosystems) following the manufacturer instructions. Fluorescence was measured after a 5hr incubation using a FLUOstar Omega microplate reader at excitation 485nm and emission of 630nm. Data were expressed as normalized relative fluorescence (NRF) relative to a blank measurement for each individual well in the absence of cells and media.
DNA/siRNA transfections, immunoprecipitations, western blotting, chromatin immunoprecipitation and quantitative real-time RT-PCR were performed as described (14, 17, 18). All experiments were conducted with a minimum of 3 repeats.
In all figures oxygen consumption data is shown as mean +/− SD normalized relative fluorescence (NRF) and expressed relative to the control samples, ATP data is shown as mean +/− SD relative luciferase units (RLU)/cell and expressed relative to the control sample, mRNA data is shown as mean +/− SD normalized relative to 18s mRNA and ChIP data is shown as mean +/− SD normalized to input and expressed relative to the Gal4 control. All western blots are representative of a minimum of three individual experiments. Band intensity was quantified using NIH ImageJ software and is shown in Supp. Table 1.
Statistical comparisons between groups were analyzed by Student’s t test or where appropriate, the paired t-test. Comparisons between more than two groups were analyzed using one-way ANOVA followed by a Tukey-Kramer multiple comparisons test. Significance was determined at p<0.05. For all data * indicates p<0.05, ** indicates p<0.01 and *** indicates p<0.001.
To investigate any potential role for RelA as a regulator of oxidative phosphorylation and cellular energy production the consequences of RelA depletion by RNA interference were examined in the U-2 OS osteosarcaoma cell line. These cells, in common with many other transformed and immortalised cell lines, have a basal level of IKK and NF-κB activity (19) so these experiments were performed without an additional NF-κB activating stimulus. Interestingly, the effect of RelA knockdown varied dependent upon the time spent in culture, with increased passage of U-2 OS cells resulting in a switch from a decrease to an increase in oxygen consumption upon RelA depletion (Fig. 1A). For ease of interpretation, the control levels in earlier and later passage cells have both been normalized to 1 but absolute levels are given in the Figure legend. Addition of the antibiotic oligomycin, which inhibits the mitochondrial H+-ATP synthase, strongly inhibited cellular oxygen consumption in late passage U-2 OS cells, confirming these measurements derive from mitochondrial oxidative phosphorylation (Fig. 1B). Moreover, under these conditions, no increase in oxygen consumption following RelA knockdown was seen, further confirming NF-κB regulation of cellular respiration (Fig. 1B).
To determine whether these changes in oxidative phosphorylation resulted in changes in cellular ATP levels we examined U-2 OS cells following RelA knockdown. As with oxygen consumption, the effect seen proved dependent upon the time the cells had spent in culture. RelA knockdown in earlier passage human U-2 OS osteosarcoma cells inhibited ATP production, while RelA knockdown in later passage cells resulted in increased ATP production (Fig. 1C). Similar results were seen with knockdown of RelA in PC3 prostate cancer cells, confirming the generality of this observation (Supp. Fig. 1A). Endogenous RelA has previously been shown to be mitochondrially localized in these cells (9).
Cancer cells derive the majority of their ATP production from glycolysis. We were therefore interested in determining whether these changes in ATP levels resulted from the effects seen on oxygen consumption. Interestingly, treatment of later passage U-2 OS cells with oligomycin did not significantly change ATP levels in control cells, suggesting that, as predicted, ATP production comes primarily from glycolysis. However, oligomycin treatment did prevent the increase in ATP seen upon RelA knockdown indicating that this effect results from RelA repression of oxidative phosphorylation (Fig. 1D). Consistent with this, we saw no significant RelA dependent change in glucose consumption in late passage U-2 OS cells (Supp. Fig. 1B).
Large fluctuations in fundamental aspect of cell metabolism, such as in ATP levels and oxygen consumption, are not to be expected. Nonetheless, to confirm the functional significance of these effects we investigated down stream effects of these changes by analyzing the energy sensing protein, 5′-AMP activated protein kinase α (AMPKα) (20). Importantly, RelA knockdown differentially altered activatory Tyr172 phosphorylation of AMPKα, in a manner correlating with the effects of RelA on total ATP levels (Supp. Fig. 1C), indicating that these changes were physiologically significant.
We were interested in whether mitochondrially localized RelA was a cause of these effects on oxygen consumption and ATP levels. We confirmed the presence of RelA in mitochondrial protein preparations from U-2 OS cells with higher levels being seen in later passage cells (Fig. 2A). Since RelA is a DNA-binding protein, one mechanism through which it could exert its effects on oxygen consumption and ATP levels is through regulation of mitochondrial gene expression. We therefore examined the binding of RelA to mitochondrial DNA (mtDNA) using chromatin immunoprecipitation (ChIP). Mitochondrial gene transcription occurs bidirectionally from 3 promoters located in the D-Loop region of the genome (Supp. Fig. 2A). Consistent with the increased localization of RelA to mitochondria in later passage U-2 OS cells (Fig. 2A), significant binding of RelA to the D-loop promoter region of the mitochondrial genome was observed only in late passage cells (Fig. 2B). Primers to the cytochrome B region, which is immediately adjacent to the D-loop, confirmed this result (Supp. Fig. 2B). Due to the resolution allowed by ChIP, this result does not imply RelA is specifically binding to sites within the cytochrome B gene. Interestingly, there was a RelA dependent decrease in binding of POLRMT, the mitochondrial RNA polymerase, to mtDNA that was observed only in late passage cells (Fig. 2C). To assay what effect this change in POLRMT binding may have on mitochondrial gene expression, quantitative PCR analysis was performed using primers targeted to specific mitochondrially encoded genes. Significantly, a decrease in cytochrome B RNA and protein levels was also observed in later passage cells, which was partially reversed upon RelA depletion (Fig. 2D & Supp. Fig. 2C). No significant effect of RelA knockdown on cytochrome B RNA levels was seen in the earlier passage cells. Similar, late passage specific effects were also seen for cytochrome C oxidase I and cytochrome C oxidase III (Supp. Fig. 2D and data not shown).
Taken together, these results suggest that RelA effects on oxidative phosphorylation and ATP levels in earlier passage cells are not mediated directly through mitochondria and most probably result from regulation of nuclear gene expression. However, as the cells continue to grow in culture, a process generally associated with acquisition of a more transformed phenotype, RelA actively represses oxidative phosphorylation, at least in part through repression of mitochondrial gene expression.
To learn more about RelA function in mitochondria we created a series of fusion proteins in which a mitochondrial targeting sequence (MTS) was fused to the N-terminus of RelA, thereby allowing the effects of mitochondrial RelA to be dissociated from nuclear and other cellular effects. In addition to full length human RelA, MTS tagged versions of a C-terminal deleted RelA, encoding the amino terminal DNA binding and dimerization Rel Homology Domain (RHD), together with a specific DNA binding mutation (DBM) of full length RelA, were created. ChIP analysis of a non-MTS tagged form of the RelA DBM mutant with primers to the IκBα promoter, a known NF-κB target (21), confirmed the loss of sequence specific DNA binding ability (Supp. Fig. 3A). Western blot analysis confirmed that all MTS tagged proteins were targeted to mitochondria (Supp. Fig. 3B). Interestingly, all MTS tagged forms of RelA, including the DBM mutant, were seen to bind mtDNA by ChIP analysis, with significantly higher levels of binding seen with the C-terminally deleted RelA RHD (Fig. 2E). This result suggests that RelA does not require direct binding to κB elements within the mtDNA and that recruitment can result indirectly, through interaction with mitochondrial transcription factors.
Consistent with the data seen upon depletion of endogenous RelA (Fig. 2D), expression of exogenous, MTS tagged RelA in U-2 OS cells, resulted in repression of cytochrome B and cytochrome C oxidase III RNA levels (Fig. 2F and Supp. Fig. 3C) as well as a significant decrease in oxygen consumption and ATP levels (Supp. Fig. 3D & E). None of these effects were seen upon expression of non-tagged, wild type RelA, confirming that MTS tagging can distinguish mitochondrial from other cellular effects of RelA. Repression of oxygen consumption by exogenously expressed MTS tagged RelA but not wild type RelA was confirmed in p53 null human non-small cell lung carcinoma H1299 cells (Supp. Fig. 3F). Consistent with this data and our previous analysis (Fig. 2C), expression of MTS tagged RelA resulted in a decrease in POLRMT binding to mtDNA (Supp. Fig. 3G).
Analysis of the MTS tagged RelA mutants produced a surprising result. Although MTS-RelA RHD binds mtDNA more efficiently than any other RelA construct (Fig. 2E), it failed to repress mitochondrial gene expression, oxygen and ATP levels together with POLRMT binding (Figs. 2F, Supp. Fig. 3C-E & G). By contrast MTS tagged RelA DBM, exhibited all of these effects almost to the same levels as MTS tagged wild type RelA. Therefore, DNA binding and/or recruitment to the mitochondrial genome is not sufficient in itself to regulate oxidative phosphorylation and there is a requirement for the C-terminal RelA transactivation domain.
A number of positive and negative feedback loops between the p53 tumor suppressor and NF-κB pathways have been identified (22) and p53 suppresses NF-κB dependent tumorigenesis in a murine lung cancer model (23). Furthermore, p53 effects on glycolysis can be NF-κB dependent (6). We observed a decrease in p53 levels associated with increased growth of U-2 OS cells in culture, (Supp. Fig. 4A). Therefore, we investigated whether p53 activity might also regulate RelA mitochondrial function. Significantly in both p53 null H1299 cells and in p53 −/− MEF cells, used to avoid any issues with background levels of wild type p53, re-expression of p53 reduced RelA levels in mitochondria (Fig. 3A & B), while, induction of p53 expression by IPTG treatment of H1299wtp53 cells (15) (Supp. Fig. 4B) resulted in loss of RelA binding to mtDNA as determined by ChIP analysis (Fig. 3C).
We next determined the consequences of p53’s ability to regulate RelA mitochondrial localization. Consistent with p53-induced loss of mitochondrial RelA, siRNA depletion of endogenous RelA no longer resulted in an increase in cytochrome B or cytochrome C oxidase III RNA levels in either H1299 or p53 −/− MEF cells in which p53 was re-expressed (Figs. 4A & B, Supp. Figs 4C-H). Q-PCR and western blot analysis confirmed the effects of p53 on RelA did not result from changes in RelA protein or RNA levels (Supp. Fig. 4B, D, E, G, H).
These observations suggested that RelA dependent changes in oxidative phosphorylation and ATP levels would also be p53 dependent. Indeed, induction of p53 expression in H1299wtp53 cells abolished the increase in oxygen consumption and ATP levels seen upon siRNA depletion of endogenous RelA (Fig. 4C & D). Moreover, a complete reversal of RelA function was observed, with depletion of endogenous RelA now reducing the increase in oxygen consumption and ATP levels seen upon induction of p53. Importantly, similar contrasting effects of RelA on ATP levels were observed between wild type or p53 −/− MEF cells. Here, RelA knockdown in wild type cells resulted in a decrease in ATP levels, while in the absence of p53, RelA knockdown had the opposite effect (Supp. Fig. 4I).
Although our results demonstrated clear effects on RelA mitochondrial localization, the mechanism through which this is achieved and how this might be perturbed by p53 was not clear. RelA lacks a defined N-terminal MTS suggesting transport to mitochondria through a distinct, rate limited, mechanism to proteins exclusively or predominantly localized in this organelle. If RelA possessed a classical mitochondrial targeting sequence, a disproportionate level of protein could become localized to this organelle.
An indication of how this occurs resulted from separate analysis of proteins binding the region of Threonine 505 (T505) phosphorylation in the RelA TAD. Previously we identified phosphorylation of RelA at T505 as an important regulatory site that can help determine cell fate by inhibiting RelA’s anti apoptotic function (17, 18). To identify proteins binding to the evolutionarily conserved T505 region, we performed peptide affinity chromatography using either a T505 region peptide or a scramble control peptide (Fig. 5A). Analysis of the eluted proteins revealed a 75kDa protein that specifically bound the T505 peptide, identified by mass spectrometry as Mortalin (Fig. 5B). Western blot analysis of column fractions confirmed that Mortalin was present only in the samples from the T505 peptide columns (Fig. 5C). Importantly, immunoprecipitation of endogenous RelA demonstrated binding to endogenous Mortalin in HeLa, U-2 OS and HEK 293 cells (Fig. 5C & D). We next analyzed the effect of mutating the RelA T505 residue to alanine using RelA−/− immortalized mouse embryo fibroblasts (MEFs) reconstituted with wild type and mutant forms of RelA, to overcome background interactions from wild type RelA. Confirming the specificity of this interaction, T505A mutated RelA no longer co-immunoprecipitated with Mortalin (Fig. 5E).
Mortalin, also known as mitochondrial heat shock protein 70, is a member of the HSP-70 family that plays an important role in a number of cellular processes, including the stress response, cellular proliferation, intracellular trafficking, antigen processing, differentiation and tumorigenesis (24, 25). Mortalin is known to act as a molecular chaperone facilitating protein unfolding and transport across the mitochondrial inner membrane through interaction with translocase of inner membrane (TIM) proteins (26-28), although it can also be found in the cytoplasm of certain cell types (29). We were therefore interested in whether this interaction with Mortalin provided a basis for RelA mitochondrial localisation and function.
Consistent with this hypothesis, the RelA T505A mutant also displayed reduced mitochondrial localization (Fig. 5F) and failed to bind mtDNA, as determined by ChIP analysis (Fig. 5G and Supp Fig. 5A). Furthermore, siRNA knockdown of Mortalin in U-2 OS and H1299 cells resulted in reduced levels of mitochondrial RelA (Fig. 6A and Supp. Fig. 5B) and reduced binding of endogenous RelA to mtDNA as determined by ChIP analysis (Fig. 6B), while overexpression of Mortalin resulted in increased levels of endogenous mitochondrial RelA (Fig. 6C). Taken together, these data indicated that mitochondrial RelA localization is mediated by an interaction between the T505 region of the RelA TAD and Mortalin.
Importantly, this data provided a potential explanation for the ability of p53 to exclude RelA from mitochondria, since it has been previously reported that Mortalin also binds p53 and can sequester it in the cytoplasm (30, 31). Consistent with this hypothesis, induction of p53 expression by IPTG treatment of H1299wtp53 cells resulted in reduced RelA binding to Mortalin (Fig. 6D).
These results indicate that RelA mitochondrial localization and function is dependent upon the p53 status of the cell, with high p53 levels inhibiting RelA translocation to this organelle through inhibition of RelA association with Mortalin. As a consequence, p53 prevents RelA dependent repression of mitochondrial gene expression and oxidative phosphorylation. However RelA and p53 can also cooperatively promote oxidative phosphorylation through other mechanisms, most likely involving regulation of nuclear gene expression (Fig. 7).
This study demonstrates endogenous RelA binding to the mitochondrial genome and links this to regulation of oxidative phosphorylation. Such direct regulation of mitochondrial gene expression by normally nuclear transcription factors can potentially provide a mechanism to link the regulation of mitochondrial function with other cellular signaling pathways. We propose that this process contributes to the switch to glycolysis seen in cancer cells and acts as a complementary mechanism to the RelA dependent increase in glycolytic gene expression seen upon loss of p53 reported by Kawauchi et al (6) (Fig 7). We anticipate that pathway will act in parallel to other nuclear transcription factors, such as Stat3, which can also exhibit non-transcriptional mitochondrial functions that contribute to tumorigenesis (32).
Our data suggests that p53 prevents RelA mitochondrial localization by inhibiting its interaction with Mortalin (Fig. 6D). Since p53 has itself been shown to bind to RelA and form a transcriptionally active complex upon stimulation with TNFα or replication stress (33) as well as interact directly with Mortalin (30, 31), this could be achieved through competitive binding and sequestration. However, other mechanisms are possible and our data also suggests that RelA mitochondrial import is regulated by post-translational modification. This conclusion is implied from our identification of Mortalin as a protein binding the T505 region of RelA (Fig. 5), with the T505A mutation resulting in disruption of the RelA/Mortalin complex and inhibition of RelA mitochondrial localization (Fig. 5E-G). Although this suggests phosphorylation at this site may be required to promote this interaction, a purely structural effect cannot be ruled out.
It is probable that the physiological role of RelA in mitochondria is selective and that it will not regulate mitochondrial function in all cell types at all times. We propose that this is most likely to occur during NF-κB dependent processes in vivo, coupling the physiological response to the need for increased or decreased energy and ATP levels. For example, IKK/NF-κB signaling has been shown to affect mitochondrial function and biogenesis in skeletal muscle (34), while Guseva et al. demonstrated that TRAIL activation of NF-κB DNA binding in mitochondria in LNCAP and PC3 prostate cancer cells resulted in repression of mitochondrial gene expression (9). We suggest that tumor cells (and the cell lines derived from them) which have become dependent upon NF-κB for their ability to grow and survive, have subverted NF-κB mitochondrial function, where it contributes to the switch from oxidative phosphorylation to glycolysis, in part through fulfilling a negative regulatory role on mitochondrial gene expression.
Our results provide another mechanism linking the p53 and NF-κB pathways that will contribute to the ability of p53 to suppress the oncogenic characteristics of NF-κB (22) and inhibit NF-κB dependent tumor growth (23). We propose that p53 loss during tumorigenesis, removes an important control on RelA regulation of mitochondrial function, providing a pathway through which NF-κB can contribute towards cancer cell malignancy by fundamentally altering cellular metabolism.
We thank Ron Hay and Aichi Msaki for supplying reagents, Phil Pasdois and Andrew Halestrap for technical advice, Sonia Rocha, Derek Mann and Chris Paraskeva and all members of the NDP Laboratory as well as former colleagues at the universities of Bristol and Dundee for their advice and support. RFJ and IW were funded by CRUK grants C1443/A12750 and C1443/A6721.