IκBα resides in the cytosol where it retains the inducible transcription factor NF‐κB. We show that IκBα also localises to the outer mitochondrial membrane (OMM) to inhibit apoptosis. This effect is especially pronounced in tumour cells with constitutively active NF‐κB that accumulate high amounts of mitochondrial IκBα as a NF‐κB target gene. 3T3 IκBα−/− cells also become protected from apoptosis when IκBα is specifically reconstituted at the OMM. Using various IκBα mutants, we demonstrate that apoptosis inhibition and NF‐κB inhibition can be functionally and structurally separated. At mitochondria, IκBα stabilises the complex of VDAC1 and hexokinase II (HKII), thereby preventing Bax recruitment to VDAC1 and the release of cytochrome c for apoptosis induction. When IκBα is reduced in tumour cells with constitutively active NF‐κB, they show an enhanced response to anticancer treatment in an in vivo xenograft tumour model. Our results reveal the unexpected activity of IκBα in guarding the integrity of the OMM against apoptosis induction and open possibilities for more specific interference in tumours with deregulated NF‐κB.
IkBα interacts with VDAC and hexokinase II at mitochondria to prevent cytochrome c release and apoptosis of tumor cells, independently of its function as an NF‐κB inhibitor.
IκBα resides at the outer mitochondrial membrane.
IκBα associates with VDAC1 and hexokinase II to inhibit the dissociation of HKII.
IκBα prevents Bax‐mediated cytochrome c release for apoptosis induction.
This novel anti‐apoptotic activity of IκBα is independent of NF‐κB retention.
IκΒα functions as inhibitor of the inducible transcription factor NF‐κB by sequestering its subunits in the cytosol (Baeuerle & Baltimore, 1996). For the activation of NF‐κB, IκBα is phosphorylated, ubiquitinylated and degraded by the proteasome (Hayden & Ghosh, 2008). This allows the remaining dimeric subunits, composed of relA (p65), c‐rel, relB, p50 or p52, to enter the nucleus, bind promoters of a wide range of target genes and initiate their transcription. In most cells, the activation of the NF‐κB target genes implements a state of cell death resistance (Luo et al, 2005). As a target gene, IκΒα is upregulated by NF‐κB and inactivates the transcription factor in a feedback loop (Le Bail et al, 1993). In this capacity, IκΒα exerts a pro‐apoptotic activity as it inhibits the anti‐apoptotic NF‐κB. Whether IκBα has, besides NF‐κB inhibition, additional activities and whether it directly contributes to apoptosis inhibition as a NF‐κB target gene are unknown.
The mitochondrial outer membrane permeabilisation (MOMP) is acknowledged to be a pivotal event in the signalling for apoptosis (Spierings et al, 2005). It leads to the release of pro‐apoptotic factors such as cytochrome c that can then assemble the apoptosome, a caspase‐activation platform (Jiang & Wang, 2004). MOMP is accomplished by Bcl‐2 family members, in particular Bax and Bak, that form pores in the OMM. This activity must be tightly controlled; however, the underlying mechanism of how this is accomplished is only incompletely understood. VDAC1 is a protein at the OMM that can recruit cytosolic Bax for MOMP. This, in turn, is abrogated by hexokinase II (HKII), which can bind to VDAC1 and prevent the interaction with Bax (Pastorino & Hoek, 2008).
In this study, we uncovered a novel and unexpected apoptosis function of IκΒα. We found that it localises to the OMM where it interacts with VDAC and HKII to stabilise this complex and prevent Bax‐mediated cytochrome c release for apoptosis. A range of experiments including the use of specific IκBα mutants with different sub‐cellular localisations or distinct domains, IκBα variants with opposing effects on NF‐κB and conditions without or with constitutive NF‐κB activity indicate that the novel anti‐apoptotic activity of IκBα at the OMM can be separated from its known ability to inhibit NF‐κB. When IκBα is reduced by RNAi in tumour cells with constitutively active NF‐κB, it leads to a re‐sensitisation for apoptosis induction in vitro and in vivo.
IκBα localises to the outer mitochondrial membrane
In order to explore the sub‐cellular residence of IκBα, we used confocal immunofluorescence microscopy and determined that in HeLa, MCF7 and 3T3 cells IκBα partly co‐localised with mitochondria, while in the cell lines MDA‐MB‐231, PC3 and HCT‐116 that display constitutive NF‐κB activity (Nakshatri et al, 1997; Dejardin et al, 1999; Suh et al, 2002; Sakamoto et al, 2009), we observed an almost complete spatial overlap (Fig 1A, Supplementary Fig S1A). Treatment of intact mitochondria isolated from HeLa cells with proteinase K revealed that the enzyme gained access to IκBα as it could substantially reduce the IκBα signal in a protein blot and likewise that of the OMM protein Tom20 as a control, while the inner mitochondrial protein cytochrome c was protected (Fig 1B). With isolated mitochondria from the other cell lines, we likewise found that IκBα could be cleaved by proteinase K, suggesting that it was associated with the OMM (Fig 1C). Using the same fractionating protocol from Fig 1B, we observed a pronounced accumulation of IκBα in the mitochondrial faction of PC3 cells and a more equal distribution between the cytosolic and mitochondrial fraction in MCF7 cells (Supplementary Fig S1B). The co‐localisation of IκBα in PC3 cells was predominantly to mitochondrial VDAC1 when compared with an ER marker (Supplementary Fig S1C) and could be observed in HCT‐116 and MDA‐MD 231 cells as well and to a lesser degree also in primary HUVEC cells (Supplementary Fig S1D). Mitochondrial IκBα was still susceptible to degradation upon TNF stimulation of the cells (Supplementary Fig S1E).
Reduction of IκBα changes the cells’ apoptosis sensitivity dependent on its presence at mitochondria
NF‐κB is known to be constitutively active in many malignant cells, which would lead to an upregulation of its target gene IκBα. Using Oncomine, we indeed found that IκBα is increased in many microarray studies on tumours when compared to normal cells (Supplementary Fig S2A). Our experiments revealed that in MDA‐MB‐231 and PC3 cells, which harbour prominent total IκBα levels and constitutive NF‐κB, its protein accumulated at mitochondria (Fig 2A, Supplementary Fig S2B). In HCT‐116 cells that are also NF‐κB‐positive, IκBα likewise showed a noticeable concentration at mitochondria, even though its total IκBα was less pronounced. In order to further investigate the role of IκBα in those cells, we stably knocked down IκBα in MDA‐MB‐231 cells using lentiviruses. Constitutively active NF‐κB remained unaffected, indicating that it is activated by a process other than IκBα degradation in these cancer cells (Prasad et al, 2009) (Fig 2B and C). We tested a number of target genes and in line with the unaltered NF‐κB binding activity their expression levels did not change (Supplementary Fig S2C).
The treatment with various apoptosis inducers revealed that IκBα reduction rendered these cells significantly more sensitive to apoptosis (Fig 2D, Supplementary Fig S2D and E). When IκBα was targeted in PC3 cells, which likewise display constitutively active NF‐κB and a high amount of IκBα at mitochondria (Fig 2A), the cells similarly became more sensitive to apoptosis induction than when transduced with a scrambled shRNA sequence (Fig 2E–G, Supplementary Fig S2D and F). Annexin V staining revealed a good correlation with DiOC6/PI staining in the MDA‐MD 231 and PC3 cells (Supplementary Fig S2G, upper panels). The same effect could be observed at additional time points (Supplementary Fig S2H). We also used an additional RNAi construct that was less efficient in targeting IκBα and still sensitised the cells for apoptosis (Supplementary Fig S2I).
To also explore the role of IκBα at the OMM for apoptosis regulation in cells with inducible NF‐κB and low mitochondrial IκBα, we resorted to 3T3 cells. We engineered a construct with IκΒα fused to β‐actin (IκBα–β‐actin) to retain it in the cytosol and another chimera, using a N‐terminal IκBα fusion to the mitochondrial localisation sequences of mitoNEET (Wiley et al, 2007), to specifically target it to the OMM (Supplementary Fig S3A). We then reconstituted IκBα−/− 3T3 cells with the IκBα variants and with IκBα‐wt (Supplementary Fig S4A). Diverse signals for apoptosis such as ionomycin and etoposide as well as ANT1 (adenosine nucleotide translocator‐1) transfection caused enhanced apoptosis in the IκBα–β‐actin‐reconstituted cells in comparison with the IκBα−/− cells, most likely through inhibition of NF‐κB and the consequent reduced transcription of anti‐apoptotic NF‐κB target genes (Fig 2H, Supplementary Fig S4B). Reconstitution with MEET‐IκBα, on the other hand, not only overcame this apoptosis sensitisation, but led to a decrease in apoptosis (Fig 2H). NF‐κB binding was inhibited to the same extent in the MEET‐ΙκΒα‐ and IκBα–β‐actin‐reconstituted cells allowing comparison of their sensitivities to cell death independently of differences in NF‐κB target gene activation (Supplementary Fig S3B). Moreover, upon transient transfection, only the mitochondrial MEET‐IκBα but not IκBα–β‐actin could reduce apoptosis (Supplementary Fig S4C). The reconstitution of the IκBα−/− cells with WT IκBα led to an intermediate, but still significant apoptosis reduction for most signals (Fig 2H). We confirmed this with additional cell death stimuli (Supplementary Fig S4D). To directly compare cellular systems with constitutively active versus inducible NF‐κB leading to normal or increased mitochondrial IκBα, respectively, we quantified apoptosis by serum starvation when IκBα was reduced in 3T3, PC3 and MDA‐MB‐231 cells. In accordance with our expectation, we observed a stronger apoptosis sensitisation in the cells with constitutive NF‐κB and mitochondrial IκBα accumulation (Fig 2I, Supplementary Fig S4E). These results were supported by additional experiments: In cells with prominent mitochondrial IκBα localisation such as HCT‐116, a partial IκBα reduction by RNAi was sufficient to cause a significant sensitisation for apoptosis by ANT1 and t‐Bid (Fig 2A, Supplementary Fig S4F). On the other hand, targeting IκBβ did not lead to a change in apoptosis sensitivity (Supplementary Fig S4G).
IκBα at the outer mitochondrial membrane inhibits apoptosis
In order to investigate the effects of mitochondrial IκBα accumulation, we targeted IκBα to the OMM and to the mitochondrial intermembrane space, respectively, using a N‐terminal IκBα fusion to the mitochondrial localisation sequence of Smac/DIABLO (S/D) (Ozawa et al, 2007) and the above‐mentioned mitoNEET chimera (Fig 3A). Immunofluorescence confirmed the localisation of the fusion constructs to mitochondria and proteinase K digest on isolated mitochondria in combination with Western blotting supported their residence in the expected compartments (Supplementary Fig S5A). With the OMM playing a pivotal role in apoptosis induction, we tested the activity of the IκBα fusion constructs in an apoptosis assay. MitoNEET‐IκBα (ΜΕΕΤ‐IκBα) completely abolished apoptosis induced by tBid and ANT1 transfection. Bax and VDAC1 were likewise efficiently inhibited by MEET‐IκBα, while apoptosis induced by caspase‐8 was only weakly affected (Fig 3B). WT IκBα was also an efficient apoptosis inhibitor (Fig 3B), while the S/D‐IκBα fusion and IκBβ, which differs mainly in its N‐terminus from IκBα, had no effect on apoptosis by pro‐apoptotic genes (Fig 3C and D). ANT1 expression caused apoptosis in those cells as indicated by PARP cleavage and its inhibition by Bcl‐XL (Supplementary Fig S2G, lower panel). A fusion of IκBα to the mitochondrial localisation sequence of yeast mcr1, which is associated with the OMM (Hahne et al, 1994), also led to inhibition of ANT1‐induced apoptosis in 293T, MCF7 and HeLa cells (Supplementary Fig S5B–D). Also, serum starvation‐induced activation of caspases was efficiently inhibited by WT IκBα and ΜΕΕΤ‐IκBα (Supplementary Fig S5E). To determine the relation of apoptosis and NF‐κB inhibition, a non‐degradable form of IκBα (ΙκBα‐SR) and an IκBα mutant devoid of NF‐κB inhibition (IκΒα‐110A3) (Sachdev et al, 1998) were co‐transfected with ANT1 and Bax and found to be as effective as WT IκBα for apoptosis inhibition (Fig 4A). Both IκBα constructs could be detected associated with mitochondria (Supplementary Fig S6A). A reporter assay confirmed that under these conditions, the endogenous NF‐κB system was not activated (Fig 4B). The two IκBα constructs as well as the WT IκBα were also active against apoptosis induced by genotoxic drugs (Supplementary Fig S6B). In order to explore whether the effect on apoptosis could structurally be separated from the known activity of IκBα to inhibit NF‐κB, we generated MEET‐IκBα deletion mutants including one of the IκBα N‐terminus and one of its C‐terminus that are dispensable and required for NF‐κB inhibition, respectively (Hatada et al, 1993). An inverse structural requirement was found for apoptosis inhibition: the N‐terminus of IκBα was necessary for apoptosis inhibition, while the C‐terminus could be deleted without compromising apoptosis repression by IκBα (Fig 4C, top and lower left panel). The fusion to mitoNEET sequences ensured that all deletion mutants were localised to mitochondria (Supplementary Fig S6C). To determine the minimal IκBα domains sufficient for apoptosis inhibition, we generated additional mutants (Supplementary Fig S6D) and found that a construct encompassing the N‐terminal residues 1–109 was able to diminish apoptosis (Fig 4C, top and lower right panel). Apoptosis repression was also observed with the IκBα 1–109 construct without the mitoNEET moiety (Supplementary Fig S6E and F). Altogether these experiments strongly suggested that the effects on apoptosis exerted by IκBα were independent of NF‐κB inhibition.
IκBα stabilises the VDAC–hexokinase II complex via its interaction with VDAC1
In an effort to explore the function of IκBα at mitochondria, we tried to identify its interaction partners at this organelle. Using immunoprecipitations, we found that IκBα interacted, directly or indirectly, with several mitochondrial proteins such as ANT1 and CK1 located in the IMM and the intermembrane space, respectively, and also with VDAC1 at the OMM (Fig 5A). No interaction was detected with TOM20 and cytochrome c. Since VDAC1 localises to the OMM and as IκBα exerts its anti‐apoptotic effect at the same locale, it constituted a relevant target. Accordingly, IκBβ, which was found to be impaired in apoptosis inhibition (Fig 3E), was not able to interact with VDAC1 (Supplementary Fig S7A). Moreover, all IκBα mutants that were active for apoptosis repression (Fig 4C), also associated with VDAC, while the mutant lacking the N‐terminus only, showed reduced binding (Fig 5B). We also observed an interaction with VDAC1 for WT IκBα and its mutant 1–109 [which is still able to inhibit apoptosis (Supplementary Fig S6F)] without the mitoNEET moiety (Supplementary Fig S7B). ShRNA‐mediated knock‐down of VDAC1 by two independent RNAi constructs, but not VDAC3, reduced the potential of MEET‐IκBα to inhibit apoptosis (Fig 5C and D, Supplementary Fig S7C) and completely abrogated its activity in reconstituted 3T3 cells (Fig 5E). The interaction with VDAC1 appears to be lost upon apoptosis induction before any biochemical features of apoptosis were observed as confirmed by DiOC6/PI staining (Fig 5F, Supplementary Fig S7D). These experiments suggested that IκBα targets VDAC1 to inhibit apoptosis. As VDAC1 is known to recruit the pro‐apoptotic Bcl‐2 family member Bax, which causes MOMP and apoptosis induction (Shimizu et al, 1999), we investigated a possible functional interaction. We co‐incubated isolated mitochondria with or without in vitro‐translated IκBα and added increasing amounts of recombinant Bax. The release of cytochrome c served as a read‐out for the permeabilisation of the OMM as a pivotal step in apoptosis induction. Upon addition of IκBα, the release of cytochrome c was completely abrogated at all concentrations of Bax used (Fig 6A). In PC3 cells, the presence of IκBα caused a considerably delayed cytochrome c release and less efficient Bax integration than in the absence of IκBα when apoptosis was induced by staurosporine (Fig 6B). HKII is known to interact with VDAC1 at the OMM, which inhibits the recruitment of Bax to this protein complex (Pastorino et al, 2002). Upon apoptosis induction, HKII is released from VDAC1 and Bax associates with VDAC1 to form a pore in the OMM (Pastorino et al, 2005). Immunofluorescence revealed a spatial overlap of HKII, VDAC1 and IκBα (Supplementary Fig S7E), and we also observed the co‐localisation of VDAC1 with IκBα in various cell lines including primary HUVEC cells (Supplementary Fig S1D). The interaction of endogenous HKII and IκBα was detected via co‐immunoprecipitation (Fig 6C). Mapping of the interaction revealed that residues 1–70 of IκBα were dispensable and that the N‐terminal 1–109 residues interacted with HKII (Fig 6D). Reducing HKII in the cells rendered WT IκBα and MEET‐IκBα unable to inhibit ANT1‐ and ionomycin‐induced apoptosis (Supplementary Fig S7F). The reduction of either VDAC1 or HKII had no effect on the association of IκBα with the remaining complex partner (Supplementary Fig S7G). With VDAC1 being associated with the outer mitochondrial membrane, this indicates that HKII is dispensable for its mitochondrial targeting. Mitochondria from IκBα‐ or control vector‐expressing HeLa cells were isolated and assayed for the interaction between VDAC1 and HKII upon addition of increasing concentrations of the HKII‐release peptide HXK2VBD, which comprises the first N‐terminal residues of HKII. The association between VDAC1 and HKII was undisturbed in the presence of IκBα, while it was almost completely abolished in its absence (Fig 6E). Conversely, when we increased the IκBα level in non‐apoptotic cells, we observed a stronger interaction between HKII and VDAC1 (Supplementary Fig S7H). When we used clotrimazole to disrupt the VDAC1–HKII association in IκBα−/−‐deficient and MEET‐IκBα‐reconstituted cells (Fig 6F, top panels) as well as in HeLa cells transfected with WT IκBα (Fig 6G), we likewise observed an increased stabilisation of the VDAC1–HKII interaction in the presence of IκBα. A co‐IP revealed that the interaction between VDAC1 and HKII was reduced by clotrimazole in the absence of IκBα, while this effect was prevented in the presence of MEET‐IκBα (lower panels Fig 6F).
IκBα knock‐down in cancer cells with constitutive active NF‐κB sensitises to anticancer compounds in vivo
MDA‐MB‐231 cells that display constitutively active NF‐κB and prominent IκBα accumulation at mitochondria (Fig 2A) were engineered to stably express shIκBα or scrambled constructs (Fig 2B). They were then bilaterally engrafted in nude mice, which were treated with PBS or doxorubicin for 5 weeks, and tumour size measurements were taken twice weekly. Tumours with reduced IκBα decreased in volume upon drug treatment, while the control cells were much less affected (Fig 7, Supplementary Fig S8A).
We have shown here that, unexpectedly, IκBα, besides its inhibitory effect on NF‐κB, also guards the integrity of the OMM and thereby controls the intrinsic, mitochondrial pathway for apoptosis. This novel activity of IκΒα is especially pronounced in cells that accumulated IκBα at mitochondria such as tumour cells with constitutive NF‐κB activity (Fig 2B–G). In cells with baseline mitochondrial IκBα, the protective effect was revealed only upon complete knock‐down of IκBα (Fig 2H and Supplementary Fig S4D), probably as it was otherwise obscured by its known activity to inhibit the anti‐apoptotic transcription factor complex NF‐κB.
IκΒα's ability to regulate MOMP and apoptosis is accomplished by stabilising the complex of HKII and VDAC1, thereby inhibiting Bax‐mediated cytochrome c release (Fig 6). HKII is known to bind to VDAC1, the most abundant VDAC isoform, via its N‐terminus and its putative BH4 domain (Pastorino & Hoek, 2008). This seems to favour the oligomeric association of VDAC proteins and keeps these proteins unable to interact with Bax for apoptosis induction. It is well established that the pro‐apoptotic multi‐domain members of the Bcl‐2 family form proteinaceous or, indirectly, lipidic channels in the OMM as a critical step in apoptosis induction (Shamas‐Din et al, 2013). The observed importance for VDAC for Bax‐induced apoptosis could be explained by VDAC acting as a recruitment or interaction partner for the latter, similar to the described Bak–VDAC2 complex (Roy et al, 2009). With the novel activity at the OMM described here, IκBα exerts the same effect as the enterobacterial fimA protein, which likewise stabilises the VDAC–HKII interaction and inhibits apoptosis (Sukumaran et al, 2010). Our results indicate that IκBα can make contact with VDAC1 and HKII independently of each other and hence might act like as a molecular clamp for this complex (Supplementary Fig S8B).
The role of VDAC in apoptosis has been questioned in one study that used VDAC1‐ and VDAC3‐deficient and VDAC2 knock‐down cells (Baines et al, 2007). However, VDAC proteins are proven modulators in apoptosis signalling and individual members control several apoptosis cascades through mitochondria (Shoshan‐Barmatz et al, 2006, 2009; Simamura et al, 2008; Abu‐Hamad et al, 2009; Tomasello et al, 2009; McCommis & Baines, 2012). Our discovery could explain previous reports according to which IκBα expression is anti‐apoptotic, which, however, was not specifically addressed by the authors (Vannucchi et al, 2005).
The interaction of IκBα with the VDAC1–HKII complex is specific: IκBβ is not able to associate with VDAC1 and inhibit apoptosis (Fig 3D, Supplementary Fig S7A). We have also found cell lines with constitutively active NF‐κB that harboured similar IκBα levels in whole‐cell lysates but dramatically differential IκBα association with the OMM (Figs 1A and 2A). The localisation of NF‐κB subunits to mitochondria was revealed previously (Bottero et al, 2001; Cogswell et al, 2003), and in particular, Bottero et al also found the association with the OMM using immunoelectron microscopy and proteinase K digest but its role for apoptosis regulation was not investigated.
IκBα's role for apoptosis inhibition at the OMM is spatially and functionally separable from its well‐known function to inhibit the transcription factor NF‐κB (Fig 4). Our mapping experiments indicated that, in contrast to the inhibition of NF‐κB, the N‐terminus of IκBα is responsible for its repression of apoptosis (Fig 4C). Moreover, specific IκBα mutants with either reduced or constitutive NF‐κB retention (Fig 4A), and IκBα variants with differential localisation (Supplementary Fig S3) allowed separate the pro‐ and the anti‐apoptotic activity of IκBα. Finally, experiments under conditions with or without constitutive NF‐κB activity (Figs 2 and 4B) indicated that the novel activity of IκBα to repress apoptosis at the OMM is distinct from its known ability to inhibit NF‐κB.
In most cell types, the activation of the transcription factor NF‐κB leads to an anti‐apoptotic response through the upregulation of cell death repressors. Hence, as the inhibitor of NF‐κB, IκΒα exerts a pro‐apoptotic function, which is counteracted by its novel and unexpected activity to guard the integrity of the OMM against apoptosis induction as presented here. Therefore, during its degradation upon activation of NF‐κB, while it is re‐synthesised as a consequence of the transcriptional activity of NF‐κB, IκΒα contributes to the balance of survival and cell death signals at all time points in the NF‐κB activation cycle. This is especially important, as many of the signals that activate NF‐κB constitute cell stress. We have observed an effect on apoptosis in cells with inducible NF‐κB only when IκBα was knocked out. Partial reduction by knock‐down did not show this effect (Supplementary Fig S4D and F).
IκBα's activity to inhibit apoptosis at the OMM is spatially and functionally separable from its function to inhibit the transcription factor NF‐κB, even though in normal 3T3 cells mitochondria‐targeted IκBα exerts both effects (Fig 2H). In this scenario, mitochondrial IκBα affects mitochondria and NF‐κB at the same time as it can establish the relevant interactions simultaneously.
In numerous tumours, on the other hand, IκBα is rendered unable to inhibit NF‐κB based on a number of heterogenous mutations (Sethi et al, 2008). As a consequence, NF‐κB is constitutively active and this is required for the survival of the malignant cells (Prasad et al, 2009). Many studies such as Shukla et al detected that advanced, more aggressive tumour stages show an increase of active NF‐κB together with an increased accumulation of IκBα as its target gene (Shukla et al, 2004). We have focused here on such tumour cells that are NF‐κB‐positive and nevertheless express (NF‐κB inhibition‐inactive) IκBα, features that have been found in many malignant cancers (Sethi et al, 2008). Under such conditions, NF‐κB is not available for IκBα as a binding partner. Because of this and since IκBα is a target gene of NF‐κB, its protein will accumulate and interact with VDAC at mitochondria to inhibit apoptosis (Supplementary Fig S8B). In cells with inducible NF‐κB, IκBα has the dual activity of stabilising VDAC1–HKII and masking the localisation sequence of RelA. IκBα can perform these two activities simultaneously due to different domains that facilitate this. Nevertheless, in such cells, the mitochondrial accumulation of IκBα is reduced as its interaction partner RelA is known to associate with the cytoskeleton forming a cytosolic anchor for IκBα (Are et al, 2000). Of note, the C‐terminus of IκΒα is mutated in Hodgkin's lymphoma, while its N‐terminus remains intact (Cabannes et al 1999). This disturbs only its effect on retaining NF‐κB and preserves its anti‐apoptotic effect at the OMM. In those tumour cells, the balance of the dual activities of IκBα is disturbed so that it exerts only its anti‐apoptotic effect aided by its accumulation at the OMM as a NF‐κB target gene (Fig 2, Supplementary Fig S8A). IκBα knock‐down in such cells sensitised them for apoptosis and reduced tumour burden in a mouse model (Fig 7).
Constitutive active NF‐κB is targeted by pharmacological interference in inflammatory and autoimmune diseases (Karin et al, 2004). Given that the majority of solid tumours activate NF‐κB, this protein complex is also a target in cancer therapy (Baud & Karin, 2009). These efforts, however, are complicated by the immunomodulatory effects of the transcription factor. Our findings on IκBα as a novel and major regulator of the VDAC/HKII complex that exerts an anti‐apoptotic activity, which can be functionally and structurally separated from NF‐κB inhibition, reveal more specific interference options and provide a rational for the development of targeted therapies to disrupt this complex.
Materials and Methods
Cell culture and plasmid transfection
293T, 293FT, MCF7 (Metkar et al, 2003), HeLa, 3T3, 3T3 IκBα−/− (Beg et al, 1995), MEF WT, HCT‐116 and MDA‐231 cells were maintained in DMEM (Sigma, UK) and PC‐3 in RPMI medium containing 10% FCS (Sigma), 2 mM glutamine (Invitrogen, UK), 100 μg/ml streptamycine, 100 U/ml penicillin (Invitrogen) and sodium pyruvate (Invitrogen). 293T, 3T3 and MEF cells were transfected using X‐fect (Clontech, UK); MCF7, HeLa and HCT‐116 and cells were transfected with Effectene (Qiagen, UK) following the manufacturer's instructions. 3T3 WT and 3T3 IκBα−/− cells were published (Beg et al, 1995; Basso et al, 2005). MDA‐MB‐231, PC3 and HCT116 cells have constitutive NF‐κB activity (Nakshatri et al, 1997; Dejardin et al, 1999; Suh et al, 2002; Sakamoto et al, 2009). 3T3 EV, IκBα–β‐actin‐ and MEET‐IκBα‐reconstituted cells were treated with H2O2 (250 μM), etoposide (150 μM), menadione (15 μM) and ANT transfection for 24 h and with ionomycin (10 μM) and clotrimazole (40 μM) for 4 h.
Reagents and antibodies
Hydrogen peroxide (H2O2), arsenic trioxide (As2O3), etoposide (E1383), clotrimazole (CTZ), menadione, ionomycin, doxorubicin, docetaxel, staurosporin and proteinase K were purchased from Sigma. Monoclonal anti‐β‐actin, monoclonal anti‐hexokinase II, horseradish peroxidase‐conjugated (HRP) goat anti‐rabbit and mouse monoclonal anti‐hexokinase II were from Sigma‐Aldrich. Antibodies for VDAC1 (N18, D16 and monoclonal), IκΒα (C‐21), TOM20, GRP75, Hsp60 and GAPDH were from Santa Cruz Biotechnology (Tebu‐Bio, UK). Cytochrome c and cyclophilin D (Ppif) and CK1 antibodies were purchased from BD Biosciences (UK). Polyclonal anti‐PARP and rabbit monoclonal anti‐hexokinase II were supplied from Cell Signaling (UK). HRP goat anti‐mouse antibody was obtained from Jackson Laboratories (USA) and goat anti‐rabbit Alexa Fluor 555 and 594 or goat anti‐mouse Alexa Fluor 488 antibodies were from Molecular Probes (Invitrogen). As NF‐κB inhibitor, we used N‐Oleyoldopamine (ALEXIS, Lonza, UK) (Sancho et al, 2004).
Mammalian expression vectors coding for human Bax, human IκBα and human RIP1 were from Origene (USA). Plasmids for mitoNEET's, Smac/DIABLO's and MCR1's mitochondria localisation sequences were obtained from Geneservice (UK), and amplified sequences were fused with IκBα using recombinant PCR. ANT1‐HA, VDAC1‐HA, Caspase‐8‐HA, tBID‐HA as well as the rest of the plasmids were cloned into pcDNA3 (Invitrogen). The Super‐Repressor‐IκBα (SR‐IκBα) and IκBα‐110A3 were kind gifts from L. Schmitz, Giessen, Germany. The IκBα‐110A3 protein contains alanine substitutions at leucine 115, leucine 117 and isoleucine 120 in IκBα. The reporter plasmid pNFκB‐hrGFP was purchased from Stratagene (UK).
Mitochondria isolation and treatment
Mitochondria were isolated with an established protocol (Gogvadze et al, 2003), re‐suspended in the protocol's mitochondria buffer and kept at −80°C. For greater mitochondria purity but smaller yield, a commercial kit from Pierce (Thermo Fisher, UK) was also used. 25 ng/ml of proteinase K (Sigma) was applied to isolated mitochondria for 30 min at 4°C, and PMSF was used to stop the reaction. Cleavage of outer mitochondria membrane (OMM) proteins was observed on Western blots. Recombinant Bax was purchased from ProSpec‐Tany TechnoGene LTD (Israel), diluted with mitochondria buffer and incubated with isolated mitochondria for 30 min at 25°C. Transfected mitochondria with IκBα were used or in vitro‐translated (IVT) proteins using the TNT Quick Coupled Transcription–Translation (Promega, UK). IVT proteins were run on an SDS–PAGE for their integrity determination, and equal amounts were applied on isolated mitochondria for 30 min at 25°C prior to addition of recombinant Bax (Sukumaran et al, 2010). Association of IVT proteins to isolated mitochondria and the cytochrome c release induced by recombinant Bax protein from isolated mitochondria were analysed as described previously (Kim et al, 2006).
Quantification of cell death
Mitochondria dyes TMRE (tetramethylrhodamine, ethyl ester) and DiOC6 (3,3′‐dihexyloxacarbocyanine iodide) were used in conjunction with propidium iodide to quantify apoptosis (Majewski et al, 2004; Chiara et al, 2008). In short, floating and adherent cells were harvested and centrifuged at 2,100 rpm. Supernatant was discarded and cells were re‐suspended in 150 μl of PBS where DiOC6/PI mix was added. Cells were incubated in total for 90 min and were then analysed using FACS (BD Biosciences) with the CellQuest programme (BD Biosciences). Where indicated, PI was used after permeabilisation of the cells with lysis buffer (0.1% sodium citrate, 0.1% Triton X‐100 in PBS) and sub‐G1/G0 DNA content was measured by FACS to quantify cell death. FlowJo (TreeStar Inc.) was used for data analysis. For qualitative analysis of apoptosis, PARP cleavage was measured.
Cells, seeded on coverslips (CS), were transfected by mitoCFP, mitoGFP or mitoYFP with Effectene (Qiagen) and grown for 24 h, then fixed in 4% paraformaldehyde in PBS for 20 min at 4°C. After a blocking step with 3% BSA in Tris–HCl 50 mM, NaCl 155 mM, pH 7.6 and 0.02% saponine (TBSS) for 30 min at room temperature (RT), cells were incubated with the primary antibody for 1 h at RT. Cells were rinsed five times in TBSS and then incubated with the secondary anti‐rabbit Alexa Fluor 594 antibody at a dilution of 1/500 in TBSS. Cells were washed five times in TBSS and incubated 30 min in DAPI at 2 μg/ml (Invitrogen) at RT, before being mounted in polyvinyl alcohol mounting medium with DABCO (Sigma). Cells were examined with confocal laser‐scanning microscope (Leica) with a 63× objective and analysed with LAS AF software.
Immunoprecipitation and Western blotting
Treated or untreated cells were lysed in RIPA buffer [150 mM NaCl, 50 mM Tris pH 8.0, 1% NP‐40, 0.5% deoxycholate, 0.1% SDS, 2× proteases inhibitor mixture (Sigma)]. Cell lysates were incubated on ice for 30 min and cleared by centrifugation at 4°C for 30 min at 13,000 rpm followed by two incubations of 30 min protein G magnetic beads (Millipore, UK). Cells lysates were incubated overnight at 4°C with 10 μg of anti‐IκBα goat polyclonal Ab. Immune complexes were precipitated with magnetic protein G beads, which were washed with RIPA buffer. For VDAC1–HKII interaction experiments, VDAC1 antibodies (D‐16 or N‐18) were used to immunoprecipitate complexes from isolated mitochondria following the protocol above. For FLAG immunoprecipitations, a monoclonal antibody 4C5 from Origene, Rockville, MD, USA was used. 1–2% of the total lysate was loaded for the ‘Input’ lanes. Immunoprecipitated proteins were separated in a 10% or 12% SDS–PAGE and transferred onto a PVDF membrane (Millipore). Membranes were blocked with 3% BSA in TBS containing 0.1% Tween‐20 and incubated with the appropriate antibodies. Specific signals were revealed by the ECL detection reagent (Pierce, UK).
Virulent media were produced in HEK293FT cells (Invitrogen) using the following optimised protocol for lentiviral production. For each 10‐cm dish, plasmids encoding pVSV‐G (5.4 μg, Invitrogen), pRev (3.8 μg, Invitrogen), pGag.Pol (7.8 μg, Invitrogen), pAdVantage (9 μg, Promega), and pLKO.1‐shG.O.I. Vector (20 μg) or pLenti7.3‐G.O.I. vectors (20 μg, Invitrogen) were mixed, replenished to 437.5 μl with water, and 62.5 μl of 2 M CaCl2 was added. After 5 min, the DNA–calcium mix was transferred dropwise to 2× HBS (500 μl) and was added immediately to the cells. Following overnight incubation, the medium was replaced with DMEM containing 1 mM sodium butyrate (Sigma). The supernatant was collected 48 h post‐transfection, centrifuged and filtered (0.45‐μm pore size, Millipore). 3T3 IκBα−/− cells were transduced for 6 h in the presence of polybrene (6 μg/ml, Sigma) for the reconstitution with IκBα variants (pLenti7.3‐G.O.I.). After infection, cells were sorted through a BD FACSAriaIIu fluorescence‐activated cell sorter for eGFP since the pLenti7.3 vector allows the expression of a fluorescent marker along with the gene of interest through a bi‐cistronic vector system.
IκΒα and VDAC1 downregulation by shRNA
Commercial shRNA constructs for IκBα from Sigma were tested with Western blots for sufficient downregulation of IκBα and two constructs were validated. The VDAC1 construct (also from Sigma) was validated with the same technique. The appropriate controls were also purchased from Sigma. The vectors were packaged into lentiviruses using the same approach as described above. Transductions of cells with constitutively active NF‐κB activity (MDA‐MD‐231, PC3) and non‐constitutively active (HeLa) were performed with the help of polybrene, and stable cell lines were established. The sequences of the synthetic oligonucleotides (Invitrogen) used for IκBα shRNA constructs and VDAC1 were the following:
sh‐IκBα (NFKBIA)‐42 (targeting the 3′ UTR):
sh‐IκBa‐(NFKBIA)‐88 (targeting the CDS):
sh‐VDAC1 (targeting the CDS):
sh‐VDAC1 (targeting CDS):
sh‐mVDAC1 (targeting the CDS):
sh‐NFKBIB‐ IKBbeta (targeting CDS):
scrambled shRNA sequence:
Quantitative real‐time PCR
To validate gene expression changes observed in cells with reduced IκΒα, we performed quantitative real‐time PCR on total RNA isolated from MDA‐MB‐231 and PC3 cells. RNA was extracted from treated cells using the RNeasy extraction kit (Qiagen), and cDNA synthesis was performed as described. Quantitative PCR (Q‐PCR) was performed on the QuantStudio™ 12K Flex Real‐Time QPCR System using Taqman probes (Applied Biosystems, Life Technologies) specific to the coding regions of each of the genes assessed. Samples were run in duplicate, and GAPDH expression levels were used as an internal control for normalisation of cDNA content. The QuantStudio 12K Flex V1.1.2 software was used to analyse Q‐PCR data.
Six‐week‐old female Balb/c nu/nu mice were injected with 2 × 106 of MDA‐MB‐231/SC and MDA‐MB‐231/shIκBα cells bilaterally and in the mouse fat pad using an insulin syringe. Matrigel was used and mixed with the cells at a 1:1 ratio. Animals were anaesthetised using isoflurane. A pilot study indicated small difference in the ability of cells to grow with the MDA‐MB‐231/shIκBα cells growing faster 3 weeks post‐inoculation. There was no difference in growth of both cell lines in vitro. The treatment groups were not balanced in respect to the tumour volumes as each mouse received sc and KD cells bilaterally. Three weeks post‐inoculation and irrespective of tumour growth treatment with doxorubicin or PBS started by administration through the i.v. route. Mice were mock‐ or doxorubicin‐treated once weekly, and tumour measurements were taken twice weekly using an electronic calliper. Tumour sizes were calculated using the formula π/6× length × width × height. Graphs represent average sizes of the tumour from a cohort of six mice each, and error bars show the standard error of the means. Mice with unilateral growth were excluded. Animals were treated under the 70/7113 project license following the ‘Animals’ (Scientific Procedures) Act 1986.
NF‐κB activity assay in 293T and 3T3 IκBα‐reconstituted cells
293T cells were transfected with the reporter plasmid pNFκB‐hrGFP along with expression constructs for ANT1, Bax and Tax (Munoz & Israel, 1995). The pan‐caspase inhibitor zVAD (MPBio, UK) was applied at 20 μM to avoid late degradation of proteins through apoptosis by Bax and ANT1. 24 h post‐transfection, 293T cells were collected for FACS analysis. For the rest of the cell lines, a non‐radioactive biotinylated‐oligo pull‐down assay was used to quantify the level of NF‐κB activity. In short, sense and antisense oligonucleotides encoding tandem κB sites (sense strand, AGCTTACAAGGGACTTTCCGCTGGGGACTTTCCAGGG) were annealed after denaturing and slow cooling in a buffered solution. The sense strand was biotinylated in the 5′ position, while a non‐biotinylated oligonucleotide was used as control. Nuclear extracts were prepared with standard protocols and incubated with annealed oligonucleotides for 60 min at 4°C. The oligonucleotides were subsequently precipitated by streptavidin agarose beads (Invitrogen) and rinsed three times before proceeding to Western blot analysis.
Statistical analysis was performed using the unpaired Student's t‐test. Data were obtained from n > 3 experiments for every figure and were regarded as statistically significant if P < 0.05 (*), 0.01 (**) and 0.001 (***) based on Student's t‐test.
EP designed and performed experiments and analysed the data. A‐LM‐M, CD, WC, M‐SH, JK, DQ and FO performed experiments. AA‐R, AA‐A, NDM, EOA and SG analysed the data. EP and SG wrote the manuscript. Stefan Grimm passed away unexpectedly after acceptance of this paper for publication. The EMBO Journal editorial team offers our condolences and sympathies to his family and coworkers on his untimely death.
Conflict of interest
The authors declare that they have no conflict of interest.
The IκΒα−/− cells were a kind gift of David Baltimore/Shengli Hao (California Institute of Technology). E.P. was supported by Cancer Research UK, A‐L.M‐M. and C.D by Breast Cancer Campaign. F.O. was supported by the Wellcome Trust, W.C. by the Development and Promotion of Science and Technology Talents Project (DPST), Royal Thai Government, Thailand, D.Q. by a stipend from the University of Dammam and M‐S.H. by a stipend from AstraZeneca Ldt. We thank Joel Abraham for in vivo work.
FundingCancer Research UK
- © 2014 The Authors