The human colorectal carcinoma cell line HCT15 was cultured at 37ºC, 5% CO₂ in RPMI-1640 medium (Gibco BRL, Gaithersburg, MD, USA), containing 10% FCS. This cell line endogenously expresses high-level ABCB1 and possesses a strong MDR phenotype (33). Authentification of the cell line was performed by STR DNA typing (DSMZ, Braunschweig, Germany).

Briefly, 5×10⁴ cells were seeded into 24-well plates and grown for 24 h. Then, cells were treated with 30 ng TNF-α/ml (Invitrogen, Carlsbad, CA, USA) for 5–120 min in short-term incubations and 24–72 h in long-term incubations. The cells were harvested for further analyses at indicated time points (5, 10, 15, 30, 60 and 120 min; 24, 48 and 72 h). In the long-term incubations in addition cells were harvested shortly after TNF-α application at the respective days, indicated as 24, 48 and 72 h+.

The blocking of TNF-receptors was performed in cells treated for 72 h with TNF-α. For this, 5×10⁴ cells were seeded into 24-well plates and grown for 24 h. One hour before each treatment with 30 ng TNF-α/ml (Invitrogen) cells were incubated with 8 or 20 μg mouse anti-TNFR1 monoclonal antibody (R&D Systems, Inc., Minneapolis, MN, USA) and 4 or 12 μg mouse anti-TNFR2 monoclonal antibody (R&D Systems, Inc.) respectively, to specifically block the receptor (as recommended by manufacturer). The control cells were treated with 30 ng TNF-α/ml only, or remained untreated.

Total RNA from cells was isolated using the TRIzol™ method (Invitrogen). Reverse transcriptase (RT) reaction was performed with 50 ng of total RNA (MuLV reverse transcriptase; Perkin-Elmer, Weiterstadt, Germany). Each quantitative real-time PCR (95ºC for 10 sec, 45 cycles of 95ºC 10 sec, 60ºC 30 sec and 72ºC 1 sec) was done using the LightCycler (LightCycler Fast Start DNA master hybridization probes kit; Roche Diagnostics GmbH, Mannheim, Germany). Expression of human ABCB1, NF-κB/p65, IκBα, IκBβ and of the housekeeping gene glucose-6-phosphate dehydrogenase (G6PDH) was determined in parallel from the same RT-reaction, each done in duplicate per sample. For ABCB1 a 167 bp amplicon (forward, 5′-CCCATCATTGCAATAGCAGG-3′; FITC-labeled probe forward, 5′-CACTGAAAGATAAGAAAGAACTAGAAGGTGCT-3′; LCRed640-labeled probe forward, 5′-GGAAGATCGCTACTGAAGCAATAGAAAACT-3′ and reverse, 5′-GTTCAAACTTCTGCTCCTGA-3′); for NF-κB/p65 a 365 bp amplicon (forward, 5′-AGATCAATGGCTACACAGGA-3′ and reverse, 5′-GATGGGATGAGAAAGGACA-3′); for IκBα a 354 bp amlicon (forward, 5′-CCGAGACTTTCGAGGAAAT-3′ and reverse, 5′-GTGAGCTGGTAGGGAGAATA-3′); for IκBβ a 417 bp amplicon (forward, 5′-AGTACATGGACCTGCAGAAT-3′ and reverse, 5′-GGACCATCTCCACATCTTTg-3′), and for G6PDH a 123 bp amplicon was produced, which were detected by gene-specific fluorescein- and LCRed640-labeled hybridization probes [(primers for ABCB1, NF-κB, IκBα, IκBβ; BioTeZ, Berlin, Germany); (probes for ABCB1; TIB Molbiol, Berlin, Germany); (primers and probes for G6PDH; Roche Diagnostics GmbH)]. The calibrator cDNA, derived from the human ABCB1 expressing cell line HCT15, was employed in serial dilutions (in duplicate) simultaneously in each run.

Cells were lysed in lysis buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1% NP-40, 0.5% sodium deoxycholate, 1X complete protease inhibitor cocktail (Roche Diagnostics GmbH). Cytoplasmic and nuclear extracts were prepared from cells by using the NE-PER extraction kit according to manufacturer's instructions (Pierce Biotechnology, Inc., Rockford, IL, USA). The protein content was quantified by using the Coomassie Plus protein assay according to manufacturer's instructions (Pierce Biotechnology, Inc.). Precast NuPAGE 4–12% gradient polyacrylamide gels (Invitrogen) were loaded with 50 μg protein of either total cell lysates, cytoplasmic extracts, or 10 μg of nuclear extracts and electrophorezed at 200 V, 60 min. The gels were blotted onto nitrocellulose filter (Hybond-C Extra; Amersham, Freiburg, Germany). The filters were blocked 1 h at room temperature in TBS blocking buffer (50 mM Tris, 150 mM NaCl, pH 7.5, 5% fat-free dry milk) and washed in TBST (0.05% Tween-20 in TBS buffer) at RT.

For detection anti-human-TNFR1 mouse IgG monoclonal antibody (1:100), anti-human-TNFR2 mouse IgG monoclonal antibody (1:250) (both from Santa Cruz Biotechnology, Inc., Santa Cruz, CA, USA), anti-human PgP monoclonal mouse IgG-antibody C219 (1:100; Calbiochem, San Diego, CA, USA), and anti-human NF-κB/p65 mouse monoclonal IgG-antibody (1:500), anti-human IκBα mouse monoclonal IgG-antibody (1:500), anti-human IκBβ mouse monoclonal IgG-antibody (1:100) (all from Santa Cruz Biotechnology, Inc.), anti-human β-tubulin mouse monoclonal IgM antibody (1:500; BD Pharmingen, Heidelberg, Germany), anti-human nuclear matrix protein (NMP) p84 mouse monoclonal IgG antibody (1:2,000; Abcam, Cambridge, UK) and anti-human GAPDH goat polyclonal antibody (1:1,000; Santa Cruz Biotechnology, Inc.) were used. As secondary antibodies HRP-labeled goat anti-mouse IgG-antibody (1:6,000; Pierce Biotechnology, Inc.), goat anti-mouse IgM-antibody (1:5,000; Sigma-Aldrich, St. Louis, MO, USA) or mouse anti-goat antibody (1:5,000; Santa Cruz Biotechnology, Inc.) were used. All antibodies were diluted in TBST containing 5% BSA. After incubation with the respective primary and secondary antibodies, the filters were washed in TBST. After washing, the respective protein was detected using ECL-solution (Amersham) and exposed to Kodak X-Omat AR film.

The nuclear extracts were prepared from cells by using the NE-PER extraction kit according to manufacturer's instructions (Pierce Biotechnology, Inc.). The EMSA was performed with 3 μg of nuclear extract using biotin end-labeled double-stranded oligonucleotides harboring the NF-κB/p65-binding site of the wild-type human ABCB1 promoter (sense, 5′-GCTGCTCTGGCCGCGATGGGCACTGCAGGGGCTTTCCTGTGCGCGGGGTCTCCAGCATCT-3′ and antisense, 5′-AGATGCTGGAGACCCCGCGCACAGGAAAGCCCCTGCAGTGCCCATCGCGGCCAGAGCAGC-3′); or mutant (sense, 5′-GCTGCTCTGGCCGCGATGGGCACTGCACTCGCTTTCCTGTGCGCGGGGTCTCCAGCATCT-3′ and antisense, 5′-AGATGCTGGAGACCCCGCGCACAGGAAAGCGAGTGCAGTGCCCATCGCGGCCAGAGCAGC-3′). In competition experiments unlabeled double-stranded oligonucleotides were used to control binding specificity of NF-κB/p65. For the supershift experiments, the NF-κB/p65 mouse monoclonal IgG-antibody (0.5 or 1.0 μg; Santa Cruz Biotechnology, Inc.) was used. For the EMSA 6% TBS pre-cast gels (Invitrogen) were used. Gels were blotted onto nitrocellulose filters (Amersham). After UV cross-linking shifted and supershifted bands were detected by the LightShift kit (Pierce Biotechnology, Inc.) according to manufacturer's instructions and exposed to Kodak X-Omat AR film.

After the pretreatment with 30 ng TNF-α/ml (Invitrogen) for 2 or 72 h, respectively, cells were incubated with doxorubicin (50 μM; Sigma, Taufkirchen, Germany) in phenol red-free RPMI-1640 medium for 3 h at 37ºC, and were washed with phenol red-free medium and kept on ice. Fluorescence intensity of 1×10⁴ cells was measured in duplicate per sample by using the FACSCalibur (Cell Quest program; Becton-Dickinson, San Diego, CA, USA). The drug uptake is expressed as fold-increase compared to untreated cells, which did not receive TNF-α.

For the cytotoxicity assay, 5×10³ cells were seeded into 96-well plates. The short- or long-term TNF-α effects on drug sensitivity of cells was determined by 2 or 72 h pretreatment with 30 ng TNF-α/ml. Thereafter cells were washed and treated with 50, 100, 200, 400, 1,000, 1,250 and 1,500 ng doxorubicin/ml (Sigma) for 72 h at 37ºC. Then, MTT (5 mg/ml; Sigma) was added and absorbance was measured in triplicates at 560 nm in a microplate reader (Tecan Group Ltd., Männedorf, Switzerland). Values are expressed as percent of untreated controls. Cells treated 2 or 72 h with TNF-α only served as additional control.

To determine TNF-α mediated short- and long-term induction via NF-κB/p65, we used the Cignal NF-κB/p65 Renilla/firefly dual luciferase reporter assay kit (SABiosciences, Frederick, MD, USA). For each well of a 96-well plate 100 ng Cignal NF-κB reporter or Cignal negative control plasmid-DNA, respectively, were reverse transfected with FuGENE HD (Roche Diagnostics GmbH) into 1×10⁴ cells and seeded in serum-free conditions, as recommended by the manufacturer. Then, 24 h after transfection medium was replaced with RPMI-1640 medium +10% FBS and treatment with 30 ng TNF-α/ml (Invitrogen) was started. For short-term TNF-α affected reporter expression cells were incubated 2, 4, 6, 8, 10 and 12 h, washed with PBS and lysed in passive lysis buffer (Promega). For evaluation of TNF-α affected reporter expression in long-term treatment, transfected cells were incubated for 8, 24, 48 and 72 h with 30 ng TNF-α/ml. In addition, time points of 8 h after the respective TNF-α treatment were included, indicated as 24, 48 and 72 h+. Cells were washed with PBS and lysed in passive lysis buffer. The luciferase assay was performed in triplicates with Dual-Luciferase Reporter Assay system (Promega) using 3 μl cell lysate in triplicates and using the Centro LB 960 luminometer (Berthold Technologies GmbH & Co., Bad Wildbad, Germany).

To analyze the short- and long-term effects of TNF-α on the human ABCB1 promoter activity, the NF-κB/p65 consensus harboring 990 bp promoter sequence was PCR-cloned (24). This sequence spans the first exon and intron, and was used to drive the pGL3 plasmid luciferase reporter expression (Promega).

For the assay 150 ng ABCB1 promoter luciferase or promotorless pGL3 plasmid-DNA as negative control respectively were reverse transfected in 1×10⁴ cells in each well of a 96-well plate with LTX-RG (Invitrogen) in serum-free conditions. Medium was replaced with RPMI-1640 medium and 10% FBS 24 h after transfection and TNF-α treatment was started. For determination of short-term TNF-α affected reporter expression cells were incubated with 30 ng TNF-α/ml (Biosource) for 2, 4, 6, 8, 10 and 12 h, washed with PBS and lysed passive lysis buffer (Promega). For evaluation of TNF-α-affected reporter expression in long-term treatment, transfected cells were incubated for 8, 24, 48 and 72 h with TNF-α, washed with PBS and lysed in passive lysis buffer. In addition, cells were harvested 8 h after the respective TNF-α application in the long-term treatment, indicated as 24, 48 and 72 h+. The luciferase activity was quantified by luciferase reporter assay in triplicates by the Centro LB 960 luminometer.

Analyses for statistical significance were performed with GraphPad Prism version 5 (GraphPad Software Inc., La Jolla, CA, USA). Comparison of several groups was done by one-way analysis of variance (ANOVA) and Bonferroni post hoc multiple comparison. Statistical significance was set at P-values <0.05.

The TNF-α mediated effects on ABCB1 expression were analyzed in HCT15 human colon carcinoma cells, which is an intrinsically high expressor of ABCB1. First, we evaluated the effects of human TNF-α on ABCB1 expression and ABCB1-mediated resistance. The treatment with 30 ng TNF-α/ml significantly reduced ABCB1 expression at mRNA level starting from 24–72 h (Fig. 1A). This long-term treatment caused 1.9-fold reduction in mRNA expression after 24 h, 3.8-fold reduction after 48 h and 6-fold reduction after 72 h, whereas short-term treatment (0.5–2 h) did not reduce ABCB1 expression. Similarly, western blot analysis revealed reduced PgP-expression at 24 h up to 96 h of TNF-α treatment (Fig. 1B). The short-term treatment (5 min to 2 h) however, did not alter PgP-expression. In a next step we determined possible functional effects of TNF-α mediated ABCB1 downregulation. For this, the uptake of the fluorescent drug doxorubicin as PgP substrate was determined (Fig. 1C). TNF-α treatment for 72 h leads to significant up to 3.8-fold increase in doxorubicin uptake, whereas 2 h TNF-α treatment caused only 1.4-fold increase in drug uptake. We next asked if the increased drug uptake after long-term TNF-α treatment has impact on doxorubicin cytotoxicity. Long-term treatment did increase cytotoxicity leading to 64% growth inhibition at highest dose of 1,500 ng/ml (Fig. 1E). By comparison, 2 h TNF-α treatment did not exert any effect on doxorubicin cytotoxicity even at highest drug doses (Fig. 1D). Taken together, we demonstrated in our model that long-term TNF-α treatment downregulated ABCB1 expression at mRNA and protein levels in colorectal cancer cells. This leads to high drug accumulation and sensitization towards doxorubicin, reflected by increased cytotoxicity.

Next we determined, whether TNFR1/p55 or TNFR2/p75 is essential in mediating the observed TNF-α effects on ABCB1 expression. Western blot analysis revealed, that both receptors are present in the colorectal cancer cells (Fig. 2A). For blocking the TNF-α binding to either TNFR1 or TNFR2 we pre-treated the cells with anti-TNFR1 or anti-TNFR2 antibody (Fig. 2B). As shown, addition of anti-TNFR1 antibody prevented PgP downregulation of the long-term 72 h TNF-α treatment, whereas addition of the anti-TNFR2 antibody could not prevent downregulation of PgP. In control experiments sole addition of the antibodies, however, did not alter PgP expression. This excluded any antibody mediated unspecific effect on PgP expression (Fig. 2C). Thus, binding of TNF-α to TNFR1 is essential for TNF-α mediated downregulation of ABCB1 expression after long-term TNF-α treatment.

TNFR1 mediates signaling via the NF-κB pathway. Therefore, we determined the effect of short- and long-term TNF-α on NF-κB/p65 and on IκBα, IκBβ as regulators of NF-κB/p65 (Fig. 3A and B). The short-term (5–120 min) TNF-α mediated effects show strong cytoplasmatic reduction in IκBα levels after 10 min of TNF-α treatment, whereas, no detectable alteration for cytoplasmic NF-κB level was observed (Fig. 3A). As early as 10 min of TNF-α treatment the NF-κB/p65 protein starts to accumulate in the cell nuclei. This differs from what is observed for long-term (24–72 h) TNF-α mediated effects. Here IκBα levels remain at very low level and NF-κB shuttling to the nucleus is reduced. This is paralleled by reduction in PgP levels in the same experiment (Fig. 3B). Therefore, long-term TNF-α treatment reduces shuttling of NF-κB/p65 to the nucleus and prevents re-appearance of IκBα to levels of control cells in the cytoplasm. The analysis of IκBα mRNA expression revealed that this reduction in IκBα for long-term TNF-α treatment is the result of significant up to 3-fold reduced transcription (Fig. 3C). The cytoplasmic NF-κB level, however, seems to be unaffected by long-term TNF-α action. Even if analyzing the time point of 30 min after re-application of TNF-α (indicated as 24, 48 and 72 h+), no increase in nuclear accumulation of NF-κB/p65 is seen. For NF-κB/p65 shows only slight and insignificant reduction in mRNA expression by the long-term TNF-α treatment as seen in Fig. 3C.

Furthermore, we analyzed the fate of IκBβ in long-term TNF-α treatment where IκBβ protein persists in the cytoplasm, however, for 24–72 h TNF-α treatment at slightly lower level (Fig. 3C). The expression analysis also revealed an up to 2.5-fold reduction of IκBβ mRNA for the long-term TNF-α treatment, which correlates with the reduction at protein level (Fig. 3B and C).

Since long-term TNF-α reduced nuclear NF-κB/p65 we determined if this is associated with reduced binding of NF-κB/p65 to its consensus sequence in the human ABCB1 promoter. In the EMSA we first analyzed, if binding to the NF-κB consensus sequence of the ABCB1 promoter is specific (Fig. 4A). Under non-stimulated conditions there is binding of the nuclear extract, which by addition of excess nonlabeled oligo disappers due to competition. The mutated oligo however shows no binding. After treatment with TNF-α for 30 min the mutated oligo shows binding of nuclear extract. However, addition of anti-NF-κB/p65 antibody does not lead to supershift. By contrast, use of the wild-type oligo shows binding of the nuclear extract, and supershift bands after addition of the anti-p65 antibody. This indicates, that the short-term TNF-α treatment induces NF-κB binding to the consensus sequence. Next we analyzed the impact of long-term TNF-α action on NF-κB binding to its consensus sequence within the human ABCB1 promoter by supershift experiment using the anti-p65 antibody. As positive control, the 30 min TNF-α stimulation again induced NF-κB binding, verified by appearace of a supershift band (Fig. 4B). This supershift disappeared when TNF-α was added for 24, 48 and 72 h, respectively. Interestingly, this is also seen for the samples, collected 30 min after each re-application of TNF-α, indicated as 24, 48 and 72 h+. This suggests that long-term TNF-α treatment desensitizes the colorectal cancer cells to TNF-α mediated NF-κB signaling and prevents NF-κB from binding to its consensus sequence within the ABCB1 promoter. This in turn leads to downregulation of ABCB1 expression.

Since long-term TNF-α action prevents nuclear translocation and binding to consensus sequence of NF-κB, determination of functional consequences of this is important. First, we analyzed the effects of short- and long-term TNF-α treatment using the Cignal NF-κB/p65 Renilla/firefly dual luciferase reporter assay, which harbors tandem repeats of NF-κB transcriptional response element (Fig. 5). In this assay short-term application of TNF-α induces reporter expression peaking at 8 h with a 1.7-fold increase compared to the control (Fig. 5A). The long-term effects of TNF-α dramatically differ from this. Here, persistent TNF-α application leads to significant downregulation of reporter expression compared to the short-term TNF-α effects (Fig. 5B). This is reflected by the 3- to 5-fold decreases in promoter activity at 24–72 h of TNF-α application. Even at the time points of 8 h after each respective re-application of TNF-α (indicated as 24, 48 and 72 h+) the reporter system remains unresponsive towards TNF-α mediated induction. To further validate this observation, we used the NF-κB consensus sequence containing human ABCB1 promoter for luciferase expression. These assays revealed that short-term TNF-α application slightly reduced basal promoter activity in the colorectal cancer cells (Fig. 5C). The picture again drastically changes for long-term TNF-α treatment. The results show significant 2- to 9-fold decreases in ABCB1 promoter activity (Fig. 5D). At 8 h after respective TNF-α re-application (indicated as 24, 48 and 72 h+) this reporter system remains also unresponsive for TNF-α mediated induction and the reporter expression is even further reduced. Taken together, these analyses support that long-term TNF-α inhibits ABCB1 promoter activity. This results in reduced ABCB1 transcription and PgP expression (Fig. 1) leading to MDR reversal in the resistant cells.