The primary goal of chemotherapy is to cause cancer cell death. However, a side effect of many commonly used chemotherapeutic drugs is the activation of nuclear factor-κB (NF-κB), a potent inducer of antiapoptotic genes, which may blunt the therapeutic efficacy of these compounds. We have assessed the effect of doxorubicin, an anthracycline in widespread clinical use, on NF-κB activation and expression of antiapoptotic genes in breast cancer cells. We show that doxorubicin treatment activates NF-κB signaling and produces NF-κB complexes that are competent for NF-κB binding in vitro. Surprisingly, these NF-κB complexes suppress, rather than activate, constitutive- and cytokine-induced NF-κB–dependent transcription. We show that doxorubicin treatment produces RelA, which is deficient in phosphorylation and acetylation and which blocks NF-κB signaling in a histone deacetylase–independent manner, and we show that NF-κB activated by doxorubicin does not remain stably bound to κB elements in vivo. Together these data show that NF-κB signaling induced by doxorubicin reduces expression of NF-κB–dependent genes in cancer cells.
Nuclear factor-κB (NF-κB) controls the expression of numerous gene products that play crucial roles in cell survival, cell cycle, immune responses, angiogenesis, and nervous system function. In most cell types, NF-κB is maintained in a latent form in the cytoplasm through an interaction with one of the inhibitory IκB proteins. Various inflammatory and cytokine stimuli activate a kinase cascade that converge on IKK1 and IKK2, related catalytic kinase subunits that, together with IKKγ/NEMO, form a complex which phosphorylates IκB family members and thus targets them for ubiquitination and proteosomal degradation (reviewed in ref. 1). This frees NF-κB from cytosolic retention, allowing translocation to the nucleus and activation of NF-κB target genes. There are five distinct mammalian NF-κB subunits [RelA (p65), NF-κB1 (p52/p100), NF-κB2 (p50/p105), RelB, and c-Rel] and heterodimers consisting of RelA and p50 are the most prevalent form of NF-κB in nonimmune cells.
Genes that encode antiapoptotic factors are important physiologic targets of NF-κB. The role of NF-κB as a physiologically relevant prosurvival transcription factor was initially shown in RelA null mice, which die at E15 due to massive apoptosis of hepatocytes (2). Subsequent studies have established that NF-κB plays a crucial role in the regulation of antiapoptotic genes that include cIAP1 and cIAP2, XIAP, and Bcl-Xl (3–5). Because NF-κB is a potent regulator of antiapoptotic genes, its disregulation can have profound pathophysiologic consequences. Indeed, constitutive NF-κB activity has been observed in several types of cancer where it is believed to induce gene products that allow these cells to evade apoptosis; inhibiting constitutive activation NF-κB in cancer cells can enhance cell death and NF-κB has therefore emerged as an attractive therapeutic target for cancer treatment (6, 7).
The primary goal of chemotherapeutic therapy is to induce apoptosis in cancer cells. However, topoisomerase inhibitors such as doxorubicin, camptothecin, and etoposide have been reported to potently activate NF-κB (8, 9). NF-κB–dependent activation of prosurvival gene transcription will block apoptosis and this side effect of chemotherapeutic drugs may therefore blunt the therapeutic efficacy of these compounds (10). The precise mechanisms that allow chemotherapeutic drugs to induce NF-κB activation have been slow to emerge, but recent findings indicate that single- or double-stranded DNA breaks produced by chemotherapeutic topoisomerase inhibitors are detected by the ataxia telangiectasia–mutated protein (11, 12), which then initiates receptor interacting protein–dependent activation of the IκB kinase (IKK) complex and NF-κB activation (13).
Doxorubicin is an anthracycline compound in widespread use in humans as a chemotherapeutic agent. In this study, we have assessed the effect of doxorubicin on NF-κB activation and on expression of antiapoptotic genes in breast cancer cells. Consistent with previous studies, our data shows that doxorubicin treatment results in IκBα degradation and the production of NF-κB complexes that are competent for DNA binding in vitro. However, our data show that doxorubicin treatment does not produce transcriptionally active NF-κB but rather reduces constitutive- and cytokine-induced NF-κB–dependent transcription. We show that NF-κB induced by doxorubicin is deficient in the phosphorylation and acetylation of RelA and show that NF-κB produced by doxorubicin does not facilitate formation of an active repressor complex but rather, forms NF-κB complexes that are deficient in stable binding to endogenous target loci. Together, these data show that activation of the NF-κB signaling pathway by doxorubicin results in the production of the defective form of NF-κB that serves to reduce, rather than activate, transcription.
Materials and Methods
Cell lines, adenovirus, and reagents. Cell culture reagents were obtained from Hyclone (Logan, UT). MDA-MB-231, MDA-MB-435s, and HEK293 cells were purchased from American Type Culture Collection (Manassas, VA) and maintained in 5% CO2 at 37°C in DMEM supplemented with 10% FCS, 2 mmol/L l-glutamine, and 100 μg/mL penicillin/streptomycin. IκBα (C-21, catalogue #sc-371), p50 (catalogue #sc-114), and RelA (C-20, catalogue #sc-372; for immunoblots) antibodies were purchased from Santa Cruz Biotechnology (Santa Cruz, CA); phospho-Ser276 (catalogue #3037) and phospho-Ser536 (catalogue #3036) -RelA antibodies were from Cell Signaling Technology (Beverly, MA); acetylated lysine (catalogue #05-515) and RelA (catalogue #06-418; for ChIP assay) antibodies were purchased from Upstate (Lake Placid, NY); actin antibody (catalogue #69100) was from ICN (Costa Mesa, CA). The cIAP1 and cIAP2 antibody was a generous gift from Dr. Robert Korneluk (University of Ottawa, Canada). Secondary antibodies were obtained from Jackson Immuno- Research Laboratories (West Grove, PA). Doxorubicin and trichostatin A were obtained from Calbiochem (La Jolla, CA) and recombinant murine tumor necrosis factor TNF-α was from R&D Systems (Minneapolis, MN). SMARTPOOL RNAi directed against RelA was purchased from Dharmacon RNA Technologies (Lafayette, CO). To validate the RNAi, HEK293, and MDA-MB-231 cells plated in 24-well plates were transfected with varying concentrations of the RNAi using LipofectAMINE 2000 (Invitrogen, San Diego, CA). Cell lysates were harvested 24 hours after transfection and RelA levels determined by immunoblotting. Twenty picomoles of RelA RNAi were sufficient for >95% knockdown (data not shown) and this concentration was therefore used in subsequent experiments. Recombinant adenovirus expressing β-galactosidase (LacZ) or FLAG-epitope tagged IKK2 were prepared as previously described (14).
Immunoblotting and immunoprecipitation. Cells were lysed in NP40 lysis buffer [10 mmol/L Tris (pH 8.0), 150 mmol/L NaCl, 1% Nonidet P-40, 10% glycerol, 1 μg/mL aprotinin, 1 μg/mL leupeptin, 1 mmol/L sodium orthovanadate, and 500 μmol/L phenylmethylsulfonyl fluoride (PMSF)] and protein content was analyzed using a bicinchoninic acid assay (Pierce, Rockford, IL). Samples with equivalent protein contents were separated by SDS-PAGE, and proteins were transferred onto nitrocellulose membrane. Membranes were blocked in TBS/Tween [10 mmol/L Tris (pH 8.0), 150 mmol/L NaCl, 0.2% Tween 20] and 5% (w/v) dried skim milk powder was added. The primary and secondary antibody incubations were carried out in the blocking solution (except for phospho-RelA antibodies in which a 5% bovine serum albumin solution was used). Immunoreactive bands were detected using the enhanced chemiluminescence solution (Perkin-Elmer Life Sciences, Norwalk, CT). For immunoprecipitation experiments, samples with an equal amount of proteins were incubated with RelA Ab (C-20) at 4°C overnight. The immunocomplexes were then incubated with protein A agarose for 1 hour at 4°C. Immunocomplexes were washed with lysis buffer, eluted with sample buffer and analyzed by SDS-PAGE as described above.
Electrophoretic mobility shift assay. Nuclear extracts were prepared from cells treated with doxorubicin for 0 to 6 hours. Briefly, cells were collected in cold TBS and centrifuged at 1,500 rpm for 5 minutes. Cells were then washed in buffer A [10 mmol/L Hepes (pH 7.9), 10 mmol/L KCl, 1.5 mmol/L MgCl2, 0.5 mmol/L DTT and 0.5 mmol/L PMSF], lysed in the same buffer, now supplemented with 0.1% Nonidet-P40, and centrifuged at 13,000 rpm for 10 minutes. The supernatant was removed as the cytosolic component, and the pellet was resuspended in buffer B [20 mmol/L Hepes (pH 7.9), 25% glycerol, 420 mmol/L NaCl, 1.5 mmol/L MgCl2, 0.2 mmol/L EDTA, 0.5 mmol/L DTT, 0.5 mmol/L PMSF, 5 μg/mL aprotinin, 5 μg/mL leupeptin, 5 μg/mL pepstatin, 0.5 mmol/L spermidine, 0.15 mmol/L spermine, and 100 μmol/L sodium orthovanadate] and centrifuged at 13,000 rpm for 10 minutes. The supernatant was then added to five volumes of buffer C [20 mmol/L Hepes (pH 7.9), 20% glycerol, 0.2 mmol/L EDTA, 50 mmol/L KCl, 0.5 mmol/L DTT, and 0.5 mmol/L PMSF]. An equivalent amount of protein (20 μg) from each sample was used in the DNA-binding assay. Poly dI-dC (1.5 μg) was added as a nonspecific competitor with an incubation of 10 minutes, followed by an incubation of 20 minutes with 0.2 ng oligonucleotides (AGGGACTTTCCGCTGGGGACTTTCC) end-labeled with [32P]deoxynucleotide triphosphates. The samples were then separated on 5% polyacrylamide gels. The gel was dried and analyzed by autoradiography.
Transcriptional assay. MDA-MB-231 or HEK293 cells were transfected using LipofectAMINE 2000 (Invitrogen) with the reporter plasmid pNF-κB-Luc (Stratagene, La Jolla, CA) in which the luciferase gene was controlled by a promoter containing five binding sites for NF-κB. Doxorubicin and TNF-α were added to the cells the next day and cells were harvested 48 hours after transfection. Transcriptional assay was done using a luciferase assay system purchased from Promega (Madison, WI) by following the manufacturer's instructions. Briefly, cells were washed with cold PBS and lysed in 100 μL of reporter lysis buffer. Samples were collected and centrifuged at 13,000 rpm for 2 minutes. Samples (20 μL) were read on luminometer programmed to inject 100 μL luciferase assay reagent per well and to perform a 2-second measurement delay and a 10-second reading. The protein content of each sample was analyzed and then used to normalize the luciferase reading obtained from the luminometer.
RT-PCR. MDA-MB-231 and HEK293 cells were treated with doxorubicin and/or TNF-α with the designated concentrations for 4 or 18 hours. After, 24 hours, mRNA was isolated using RNeasy Mini kits according to the manufacturer's instructions (Qiagen, Chatsworth, CA) and cDNA was generated using the Omniscript RT kit (Qiagen) with random hexamers (Roche, Nutley, NJ) as primers. PCR was done using the following primer pairs:
cIAP1 sense, 5′-GCACATTCATTATCTCCCACCTTG
cIAP2 sense, 5′-CCTCTCAGCCTACTTTTCCTTCTTC
XIAP sense, 5′-TCAGCATCAACACTGGCACGAG
Bcl-Xl sense, 5′-ACTGAATCGGAGATGGAGACCC
IL-8 sense, 5′-AAGAGCCAGGAAGAAACCACCG
IκBα sense, 5′-ACTCCATCCTGAAGGCTACCAAC
actin sense, 5′-CACCACTTTCTACAATGAGC
ChIP assay. ChIP assay was as described in ref. (15). Briefly, cells were treated with 5 μmol/L doxorubicin and/or 20 ng/mL TNF-α for 4 hours. Formaldehyde was added to the medium (1% final concentration) for 10 minutes to cross-link cellular proteins and DNA. Cells were then washed with cold PBS, collected and centrifuged at 1,500 rpm for 5 minutes. Cells were lysed with 0.1% Nonidet-P buffer [50 mmol/L Tris (pH 8.0), 2 mmol/L EDTA and 10% glycerol] and centrifuged at 3,000 rpm for 5 minutes. The nuclei were resuspended in 1% SDS buffer [in 50 mmol/L Tris (pH 8.0) and 5 mmol/L EDTA] and sonicated using a Branson Sonifier 450 (three pulses at 12 seconds each) and precleared with salmon sperm DNA/protein A agarose (UBI) at 4°C for 30 minutes. Immunoprecipitation was carried out using RelA Ab (UBI) at 4°C overnight. Salmon sperm DNA/protein A agarose was added to the immunocomplexes for incubation at 4°C for 30 minutes. Complexes were then washed thrice with washing buffer [20 mmol/L Tris (pH 8.0), 2 mmol/L EDTA, 0.1% SDS, 1% Nonidet-P, and 500 mmol/L NaCl], twice with LiCl buffer (0.25 mol/L LiCl, 1% IGEPAL-CA 630, 1% deoxycholic acid, 1 mmol/L EDTA, and 10 mmol/L Tris pH 8.0), and thrice with TE buffer [10 mmol/L Tris (pH 8.0), and 1 mmol/L EDTA]. Complexes were eluted with extraction buffer [2% SDS, 10 mmol/L Tris (pH 8.0), and 1 mmol/L EDTA]. Protein/DNA cross-linking was reversed by incubation at 65°C overnight and DNA was purified using the PCR purification kit (Qiagen). The purified DNA and input genomic DNA were then subjected to PCR using the following primers:
cIAP2 sense, 5′-ACCTTTTCCAGGCAGGCTAAGC
Doxorubicin activates the nuclear factor-κB signaling pathway but does not induce nuclear factor-κB transcription. MDA-MB-231 breast cancer cells were treated with 5 μmol/L doxorubicin or with 20 ng/mL TNF-α for periods ranging from 1 to 6 hours, and nuclear lysates were assayed for NF-κB DNA binding activity by electrophoretic mobility shift assay (EMSA). Figure 1A shows that untreated MDA-MB-231 cells had low levels of constitutive κB binding activity and that 1 hour of TNF treatment, or 3-hour doxorubicin treatment, elicited a large increase in NF-κB binding activity. Constitutive κB binding activity was not present in HEK293 cells and both TNF and doxorubicin induced robust activation of NF-κB DNA binding after 1 or 3 hours treatment, respectively (Fig. 1B).
The accumulation of NF-κB DNA binding activity correlated with the disappearance of IκBα in MDA-MB-231 cells and in HEK293 cells (Fig. 2A,-B and C-D, respectively). IκBα loss induced by TNF was transient and returned to baseline levels within 60 minutes but doxorubicin-induced loss of IκBα was maintained for at least 18 hours in both cell types. In the clinic, doxorubicin is often supplied as a bolus that achieves plasma concentrations of 5 to 10 μmol/L (16). To establish a dose-response for doxorubicin-mediated IκBα degradation, MDA-MB-231 and HEK293 cells were treated with increasing amounts of doxorubicin for 3 hours; in both cell types, doxorubicin caused detectable IκBα degradation at 1 μmol/L, with almost total loss of IκBα at 5 μmol/L (Fig. 2E). To determine if doxorubicin-induced IκBα degradation occurred through an IKK-dependent pathway, MDA-MB-231 cells were infected with recombinant adenoviruses encoding either dominant-negative IKK2, or as a negative control, with recombinant adenoviruses encoding LacZ. Cells were then treated with either TNF, doxorubicin, or the two together. Figure 2F shows that both TNF- and doxorubicin-induced IκBα degradation was attenuated in cells expressing dominant-negative IKK2, demonstrating that IKK2 activation was required for this effect.
These findings indicate that clinically relevant concentrations of doxorubicin activate the NF-κB signaling pathway and increase NF-κB DNA binding activity, consistent with previous findings. To assess the effect of doxorubicin on NF-κB transcriptional activity, transcriptional reporter assays were done in MDA-MB-231 and HEK293 cells (Fig. 3). The MDA-MB-231 line had a relatively high level of constitutive NF-κB transcriptional activity and this was increased modestly by TNF treatment (∼2-fold; compare columns 1 and 5). Treatment with doxorubicin alone did not enhance NF-κB transcriptional activity but instead strongly reduced both constitutive and TNF-induced NF-κB transcriptional activity (Fig. 3A). To determine if this was a cell-specific effect, similar experiments were done in HEK293 cells. The low basal NF-κB activity present in HEK293 cells was not altered significantly by doxorubicin treatment but the robust transcriptional response elicited by TNF-α (∼30-fold) was sharply reduced by cotreatment with doxorubicin (Fig. 3B). 3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide survival assays and lactate dehydrogenase release assays done on each cell line revealed that neither TNF nor doxorubicin caused significant cell death over this assay period (data not shown). Thus, although doxorubicin treatment induces IκBα degradation and produces NF-κB, which is capable of binding κB elements in vitro, it reduces NF-κB driven transcription in reporter assays.
Several prosurvival genes, including cIAP2, Bcl-Xl, and XIAP, have been identified as NF-κB transcriptional targets. We therefore used RT-PCR to assess the effect of doxorubicin on the expression of these and other NF-κB target genes in MDA-MB-231 and HEK293 cells. Figure 4A shows that in MDA-MB-231 cells, treatment with TNF for 4 hours substantially increased mRNA levels of cIAP2, IL-8, and IκBα, whereas doxorubicin exposure invariably reduced levels of these same gene products. When doxorubicin and TNF were combined, cIAP2, IκBα, and IL-8 mRNAs levels were close to control values, suggesting that the effects of TNF and doxorubicin are mutually antagonistic. Levels of several other genes, including cIAP1, XIAP, and Bcl-Xl were not significantly altered by either treatment in this cell type.
In HEK293 cells, IκBα mRNA levels were strongly induced by TNF and this effect was blocked in the presence of doxorubicin. cIAP2 and IL-8 mRNA levels were below our detection limit (data not shown) but XIAP and Bcl-Xl mRNAs were present and were modestly increased by TNF, albeit only after 18 hours of exposure (Fig. 4B). Interestingly, this increase was blocked in cells cotreated with doxorubicin. To exclude the possibility that the effects of doxorubicin were due to a general toxic effect on transcription or mRNA stability, actin mRNA levels were also assessed and these were unaffected by any of the treatments used (Fig. 4A -C, bottom).
To correlate the doxorubicin dose-response observed using NF-κB reporter plasmids (Fig. 2) with the effect on endogenous gene products, we assessed the effect of increasing amounts of doxorubicin on cIAP2, IL-8, and IκBα mRNA levels in the MDA-MB-231 line. Figure 4C shows that treatment with 1 μmol/L doxorubicin reduced cIAP2 mRNA levels in the presence and absence of TNF, whereas higher concentrations are required to reduce IL-8 and IκBα levels, suggesting that cIAP2 transcription is particularly sensitive to the effects of doxorubicin. We next examined cIAP1 and cIAP2 protein levels in MDA-MB-231 cells exposed to TNF and/or increasing amounts of doxorubicin for 4 or 18 hours. TNF exposure increased levels of cIAP2 but had no effect on cIAP1 (Fig. 4D), consistent with the effects of TNF on their respective mRNA levels. Doxorubicin exposure strongly reduced cIAP2 levels in both the absence and presence of TNF; at a later time point, even low doxorubicin concentrations (500 nmol/L) dramatically reduced cIAP2 levels. cIAP1 levels are not reduced by short-term doxorubicin treatment but longer-term exposure did produce a modest attenuation of cIAP1 protein. Levels of actin were not altered by TNF or doxorubicin treatment at either time point. Together, these data indicate that doxorubicin potently suppresses basal and TNF-induced expression of a number of NF-κB-dependent genes.
Nuclear factor-κB activated by doxorubicin is deficient in phosphorylation and acetylation. Recent findings have shown that production of transcriptionally active NF-κB requires not only that NF-κB is released from IκB but also involves post-translational modification of NF-κB subunits. This has been best characterized for RelA, in which phosphorylation and acetylation seem to play a key role. Several phosphorylation sites have been identified in RelA, and Ser276 and Ser536 have emerged as important regulatory sites. We therefore compared doxorubicin and TNF for their ability to induce RelA phosphorylation at these residues. MDA-MB-231 cells were treated with doxorubicin, TNF, or both for periods ranging from 10 to 180 minutes, and lysates were analyzed on immunoblots using phosphospecific antibodies directed against RelA Ser276 and Ser536. Figure 5A shows that the phosphostatus of Ser276 was not altered by either TNF or doxorubicin, added alone or together. In contrast, TNF increases phosphorylation of Ser536 in RelA within 10 minutes and this was maintained for at least 3 hours. In contrast, doxorubicin treatment caused no detectable change in Ser536 phosphorylation over this same period. TNF and doxorubicin exposure both resulted in IκBα degradation; therefore, these stimuli seem to share some, but not all, NF-κB activation variables. To determine if long-term doxorubicin exposure altered RelA phosphostatus, cells were exposed to doxorubicin, TNF or the two together for up to 24 hours. Figure 5B shows that TNF-induced Ser536 phosphorylation was maintained at 24 hours but doxorubicin did not increase Ser536 phosphorylation at any time point. Interestingly, when combined, doxorubicin did not alter the TNF-induced Ser536 phosphorylation at early time points, but after extended exposure, Ser536 phosphorylation was lost in cells exposed to both agents.
Recent studies have suggested that phosphorylation of RelA may be a prerequisite for binding and subsequent acetylation by CBP and p300. To examine its acetylation status, RelA was immunoprecipitated and examined on immunoblots using antibodies directed against acetylated lysine. Figure 5C shows that acetylated RelA could be detected in MDA-MB-231 cells that were pretreated with trichostatin A, a histone deacetylase inhibitor; under these conditions, TNF treatment clearly induced RelA acetylation, whereas doxorubicin exposure did not. Furthermore, when added together, doxorubicin blocked the RelA acetylation induced by TNF. Thus, RelA induced by doxorubicin shows deficits in phosphorylation and acetylation.
Nuclear factor-κB activated by doxorubicin represses transcription through a histone deacetylase-independent mechanism. Our data indicate that NF-κB released from IκB by doxorubicin treatment is deficient in post-translational events required for NF-κB transcriptional activity. In addition, we have shown that NF-κB produced by doxorubicin binds NF-κB elements in EMSAs and reduces NF-κB–dependent transcriptional events in vivo. One possible mechanism for this effect, consistent with the recent study by Campbell et al. (28), is that NF-κB induced by doxorubicin binds endogenous loci and assembles a histone deacetylase-dependent transcriptional repressor complex. We took two approaches to examine this possibility. In the first, we reasoned that if doxorubicin treatment produces an active repressor complex, reductions in cellular RelA levels should attenuate doxorubicin-induced repression of NF-κB responsive loci. To address this, RNA interference was used to reduce levels of RelA and the effect of doxorubicin on expression of cIAP2, IL-8, and IκBα was determined. Transfection of short interfering RNAs directed against RelA reduced RelA mRNA and protein expression dramatically (Fig. 6A and B, respectively) and resulted in a profound reduction in basal expression of cIAP2, IL-8, and IκBα mRNA and cIAP2 protein. Transfection of control scrambled RNA oligonucleotides had no effect on the expression of any of these gene products, indicating that constitutively active NF-κB normally plays a crucial role in the expression of these loci in MDA-MB-231 cells. Interestingly, the effects of TNF and doxorubicin in cells with attenuated RelA levels were reduced in magnitude but were qualitatively similar to the effects produced on normal MDA-MB-231 cells. It is therefore likely that the low amounts of RelA remaining after RNAi treatment are sufficient to mediate the effects of TNF and doxorubicin on these loci and these results show that reducing RelA levels does not relieve doxorubicin-induced repression from these NF-κB responsive loci.
As an alternative approach, we reasoned that if RelA induced by doxorubicin is acting as an active repressor that involves recruitment of deacetylases, its effects should be blocked using deacetylase inhibitors such as trichostatin A. To address this, we first employed NF-κB reporter assays on MDA-MB-231 cells treated with doxorubicin or TNF in the presence and absence of trichostatin A. Figure 7 shows that trichostatin A treatment enhanced basal (Fig. 7A) and TNF-mediated (Fig. 7B) transcription from the reporter construct but did not block the effect of doxorubicin, suggesting that histone deacetylases' action does not play a significant role in doxorubicin-induced repression. To confirm this result on endogenous NF-κB–regulated gene products, we used RT-PCR to assess levels of cIAP2, IL-8, and IκBα in cells treated with TNF, doxorubicin, or both, in the presence and absence of trichostatin A (Fig. 7C). Trichostatin A had no effect on cIAP2 or IL-8 mRNA expression but did enhance levels of IκBα in both the absence and presence of TNF. However, when added with doxorubicin, trichostatin A had no effect on doxorubicin-mediated transcriptional repression on any of these loci, indicating that histone deacetylase action is not required for the transcriptional repression induced by doxorubicin in these cancer cells.
Nuclear factor-κB activated by doxorubicin is not stably associated with κB sites in vivo. Post-translational modifications of RelA can alter its ability to stably interact with κB sites in DNA and we therefore assessed the interaction of RelA with the cIAP2 promoter with chromatin immunoprecipitation assays (Fig. 7D). In untreated MDA-MB-231 cells, the cIAP2 promoter was bound by RelA and, as expected, treatment with TNF increased this interaction substantially. However, the interaction of RelA with the cIAP2 promoter was reduced below control levels in cells treated with doxorubicin both in the presence and absence of TNF. Thus, NF-κB complexes produced in the presence of doxorubicin lack post-translational modifications necessary for assembly of an active transcriptional complex and do not stably associate with NF-κB binding elements in vivo.
Our data shows that doxorubicin treatment activates the NF-κB signaling pathway and results in the degradation of IκBα and production of NF-κB complexes that are capable of binding κB elements in vitro. However, doxorubicin treatment suppresses, rather than activates, transcription from NF-κB reporter constructs and from several endogenous NF-κB–dependent loci. RelA activated by doxorubicin is not phosphorylated at Ser536, a key regulatory residue, and is not acetylated. We show that reductions in NF-κB–dependent transcription induced by doxorubicin do not require histone deacetylase activity and that RelA binding to an endogenous promoter is reduced in cells treated with doxorubicin.
Several previous studies have established that chemotherapeutic topoisomerase inhibitors that include doxorubicin, etoposide, and camptothecin act as potent inducers of the NF-κB pathway (8, 9). The time course and dose-response of IκBα degradation and the EMSA profiles that we obtained from the cell lines examined in our studies are consistent with this earlier work. The observation that these compounds activate the NF-κB pathway and induce NF-κB, which is competent to bind DNA, has given rise to the concept that these drugs will induce NF-κB transcriptional events and lead to the production of antiapoptotic gene products that can contribute to chemotherapeutic resistance (10). In this study, we have examined the transcriptional responses mediated by doxorubicin at several endogenous NF-κB–dependent loci and, contrary to expectations, found that doxorubicin treatment invariably resulted in gene repression rather than activation.
The traditional view of NF-κB is that, once released from its inhibitory IκB subunits and bound to appropriate cis elements in responsive promoters, RelA/p50 homodimers invariably activate transcription. There is increasing evidence that post-translational modifications of NF-κB subunits play a critical role in transcriptional activation (reviewed in ref. 17). Most studies have focused on RelA and, of several phosphorylation sites identified, Ser276 and Ser536 have emerged as key regulatory residues. Ser276 is a substrate for both PKA (18) and MSK1 (19), and phosphorylation at this site enhances binding of CBP and p300 to RelA (18). Ser536 is a substrate of IKK1 (20–22) and IKK2 (23, 24); whether this phosphorylated residue facilitates binding to acetylases or other components of a functional transcriptional complex remains uncertain. Our data show that doxorubicin induces IκBα degradation and produces NF-κB, which is capable of binding κB elements but it reduces, rather than activates, NF-κB–dependent transcription. We show that in MDA-MB-231 cells, Ser276 residue is constitutively phosphorylated and not altered by treatment with TNF or by doxorubicin, whereas Ser536 becomes robustly phosphorylated in response to TNF but not doxorubicin. This differential effect spurred us to examine the acetylation status of RelA induced by TNF and doxorubicin, which revealed that TNF induced the acetylation of RelA, whereas doxorubicin did not. This raises the possibility that phosphorylation of Ser536 may be a prerequisite for subsequent RelA acetylation and shows that RelA released from IκB by doxorubicin lacks key post-translational modifications required for its role as a transcriptional activator. Interestingly, when combined with TNF, doxorubicin blocked TNF-induced RelA acetylation and, after several hours' exposure, blocked the effect of TNF on Ser536 phosphorylation and acetylation. It is likely that the inhibitory effect of doxorubicin on TNF-induced transcriptional events are a result of these altered post-translational modifications.
The transactivation function of RelA can be negatively regulated via direct and indirect association with histone deacetylase-1, -2, and -3 corepressors (18, 25, 26). We hypothesized that doxorubicin treatment might produce RelA that was capable of binding histone deacetylases and form an active transcriptional repressor complex. However, trichostatin A, a potent histone deacetylase inhibitor, had no effect on the doxorubicin-induced repression in MDA-MB-231 cells or in HEK293 cells (data not shown), even though it did enhance transcription from NF-κB reporter constructs and endogenous loci in untreated cells.
We also tested whether RNAi-mediated knockdown of RelA would “rescue” the suppressive effect conferred by doxorubicin. The RNA interference studies showed that in MDA-MB-231 cells, maintenance of cIAP2, IL-8, and IκBα levels is strongly dependent on the presence of RelA. NF-κB is activated in a wide variety of tumors and cancer cell lines (reviewed in ref. 6) and these findings are consistent with the hypothesis that constitutive activation of NF-κB in cancer cells plays a crucial role in the expression of antiapoptotic molecules and cytokines. However, although RelA knockdown dramatically decreased basal decrease in expression from these genes, the effect of doxorubicin was qualitatively similar to cells with a normal complement of RelA. Together, these data indicate that formation of an active histone deacetylase–dependent repressor complex is not required for doxorubicin-induced gene repression.
Our ChIP assays revealed that doxorubicin treatment strongly reduced the association of RelA with the cIAP2 promoter, indicating that RelA released from IκBα by doxorubicin treatment is defective either in binding or maintaining an association with κB elements. The fact that free NF-κB generated by doxorubicin readily binds κB elements in EMSA assays suggests that its affinity for DNA is not dramatically altered. However, additional constraints regulate the association of NF-κB with κB sites in vivo, and it is possible that deficits in the post-translational modifications of RelA may reduce its ability to bind or assemble factors necessary for stable DNA association (26, 27). Alternatively, a recent study (15) has shown that RelA bound to κB elements in vivo is a target for proteosomal activity; it is possible that RelA generated by doxorubicin is rapidly degraded after binding κB elements.
Recent work from the Perkins group (28) has also provided evidence for repressive forms of NF-κB. When UV light and daunorubicin were used to activate the NF-κB signaling pathway in mouse embryonic fibroblasts, the RelA released was reported to act as a transcriptional repressor at Bcl-Xl and XIAP loci, similar to our own findings. However, in the mouse embryonic fibroblast system, repression of Bcl-Xl transcription was reduced by trichostatin A and a protein complex containing RelA and histone deacetylase-3 was identified, suggesting that RelA-dependent transcriptional repression requires histone deacetylase activity. In contrast, we have shown that in MDA-MB-231 cells, doxorubicin-mediated repression of cIAP2, IL-8, and IκBα gene expression is histone deacetylase–independent. Our attempts to identify a complex containing RelA and histone deacetylase-1 or -3 following doxorubicin treatment of MDA-MB-231 cells have not been successful (data not shown). We believe that there are several possible reasons for the differences between the two studies. First, doxorubicin and daunorubicin are similar but not identical compounds that could exert differential effects on NF-κB components. Second, the regulation of NF-κB components in cancer cells and mouse embryonic fibroblasts is distinct, with constitutively active NF-κB a prominent feature of cancer cells but not mouse embryonic fibroblasts. In an unstimulated mouse embryonic fibroblast, it is likely that little NF-κB is bound to endogenous loci and therefore, transcription from these genes will reflect contributions of other transcription factors. In this setting, production of an actively repressive NF-κB complex that contains histone deacetylases may be required to reduce transcription. However, the reductions in transcription induced by doxorubicin in cancer cells could occur in the absence of active repression. It is notable in this regard that the dramatic effects of doxorubicin on cIAP2, IL-8, and IκB expression in the MDA-MB-231 line are qualitatively identical to that evoked by RelA knockdown. Third, the transcriptional complexes assembled at the various NF-κB-sensitive loci examined in these two studies are likely distinct and subject to different regulatory constraints. Indeed, although Bcl-Xl and XIAP are induced by NF-κB in mouse embryonic fibroblasts, their NF-κB–dependent regulation is very modest in the cell lines we have examined and RNAi-mediated RelA knockdown has little effect on XIAP and Bcl-Xl mRNA levels. This suggests that other transcriptional mechanisms may have superceded NF-κB as primary regulators of these genes in the breast cancer line. Taken together, our studies and those of Campbell et al. (28) indicate that activation of the NF-κB signaling pathway by several distinct stimuli can produce repressive forms of NF-κB and suggest that the precise mechanism of NF-κB repression induced by various agents will differ between loci and cell types.
In closing, we have established that doxorubicin treatment results in production of NF-κB that is deficient in post-translational modifications and which suppresses transcription of NF-κB-dependent loci in cancer cells. These suggest that the hypothesis that chemotherapy-induced NF-κB–dependent activation induces prosurvival gene transcription may require re-evaluation.
Grant support: Canadian Institute of Health Research grant #MOP62827. W.C. Ho was supported by a Jean Timmins Costello Foundation Fellowship, and P.A. Barker is an Investigator of the Canadian Institutes of Health Research.
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.
We thank Robert Korneluk for providing antibodies against cIAP1 and cIAP2, and Gioacchino Natoli for providing a helpful ChIP protocol.