Understanding the mechanisms supporting tumor-initiating cells (TIC) is vital to combat advanced-stage recurrent cancers. Here, we show that in advanced ovarian cancers NFκB signaling via the RelB transcription factor supports TIC populations by directly regulating the cancer stem-like associated enzyme aldehyde dehydrogenase (ALDH). Loss of RelB significantly inhibited spheroid formation, ALDH expression and activity, chemoresistance, and tumorigenesis in subcutaneous and intrabursal mouse xenograft models of human ovarian cancer. RelB also affected expression of the ALDH gene ALDH1A2. Interestingly, classical NFκB signaling through the RelA transcription factor was equally important for tumorigenesis in the intrabursal model, but had no effect on ALDH. In this case, classical signaling via RelA was essential for proliferating cells, whereas the alternative signaling pathway was not. Our results show how NFκB sustains diverse cancer phenotypes via distinct classical and alternative signaling pathways, with implications for improved understanding of disease recurrence and therapeutic response. Cancer Res; 77(24); 6927–40. ©2017 AACR.

Ovarian cancer is the most lethal gynecologic malignancy in the United States, resulting in over 15,000 deaths annually (1). Although most patients initially respond to platinum-based chemotherapy, over 70% of advanced-stage ovarian cancer relapses within 24 months (2). Disease recurrence is a major cause of morbidity and mortality, and mechanisms of tumor relapse and chemoresistance are unclear. Evidence suggests tumor-initiating cells (TIC) drive tumorigenesis and regulate chemoresistance and tumor reformation following first-line therapy. More recently attention is drawn to understanding the dynamic collaborations of subpopulations of cells, including TICs, that support tumor heterogeneity, therapy resistance, and disease progression (3–5).

We previously demonstrated that specific components of classical NFκB signaling are required for aggressive tumor behavior and metastatic potential in subsets of ovarian cancers (6–8). NFκB signaling is a complex network that follows both classical (canonical) and alternative (noncanonical) cascades, defined by upstream kinases and ultimate activation of either p65/p50 (classical) or RelB/p52 (alternative) transcription factors. Elevated classical NFκB signaling has been observed in TICs of prostate, breast, and ovarian tumors (9–11); however, the role of alternative NFκB signaling is not established. Both cascades are required for maintenance and promotion of breast cancer TICs (12). Classical and alternative NFκB pathways can regulate each other as well as integrate with other signaling pathways for fine-tuning functional outputs (13–16). NFκB target genes encode proteins and microRNAs important for the inflammatory response, cell survival, proliferation, adhesion, and interaction with the microenvironment. The diverse and complex roles of NFκB suggest that this transcription factor family might regulate distinct functions in different cell types within heterogeneous cell populations.

Here, we investigated the hypothesis that classical and alternative NFκB signaling support different subpopulations of ovarian cancer cells. Although both pathways are required for tumorigenesis, our data show that they maintain different cancer cell phenotypes. Classical NFκB supports proliferative cells, whereas alternative NFκB supports cells with low proliferative potential–bearing TIC markers. This study reveals a novel role for RelB in ovarian cancer and demonstrates mechanistically that NFκΒ can promote heterogeneity for efficient tumorigenesis and drug resistance. This novel insight into mechanisms of tumor heterogeneity in ovarian cancer will advance therapeutic approaches to improve clinical outcomes.

Cell lines and culture conditions

Characteristics of cell lines used in this study are provided in Supplementary Table S1. OVCAR8 (HTB-161), CAOV3 (HTB-75), and OV90 (CRL-11732) were obtained directly from the ATCC and maintained in RPMI medium supplemented with 10% heat inactivated FBS, 2 mmol/L l-glutamine and penicillin–streptomycin (Gibco) at 37°C and 5% CO2. OVCAR5 was obtained from NCI-Frederick DCTD tumor/cell line repository and maintained in RPMI with 10% FBS. ACI-23 cell line was provided by Dr. John Risinger and Memorial Health University Medical Center, Inc. and maintained in Dulbecco Modified Eagle Nutrient Mixture F-12 (DMEM/F12) medium (Gibco) supplemented with 10% FBS, 2 mmol/L l-glutamine and penicillin–streptomycin. Cell lines were authenticated via Short Tandem Repeat analysis by the Protein Expression Laboratory, Leidos, Frederick National Laboratory, and used for study within 6 months of receipt. Deidentified ascites specimens were collected as pathological waste from ovarian cancer patients and were determined to be outside of federal regulations for the protection of human subjects [45 CFR 46 (OHSRP #12727 and #12797)]. Viable tumor cells from ascites (VTCF, TGS1, TGS4) were purified using Ficoll separation and maintained in the same manner as ACI-23 cells. TIC-enriching conditions were created by culturing ovarian cancer cells in ultra-low attachment flasks (Corning) with stem cell media: DMEM:F12 (+ l-glutamine, + 15 mmol/L HEPES; Gibco) supplemented with 1% penicillin/streptomycin, 1% KnockOut serum replacement (Gibco), 0.4% BSA (Sigma-Aldrich), and 0.1% insulin-transferrin-selenium (Gibco). TIC-enriching cultures were supplemented with fresh human recombinant epidermal growth factor (EGF) and basic fibroblast growth factor (FGF) every 2 to 3 days for a final concentration of 20 and 10 ng/mL, respectively.

In vivo studies

All animal studies were approved by the NCI Animal Care and Use Committee. All mice were fed doxycycline chow (200 mg/kg) for the duration of the study. OV90 and ACI23 cells were maintained in TIC-enriching conditions for one week before injections. For intraperitoneal xenografts, 1 × 106 OV90 cells in a volume of 500 μL PBS were injected into the abdominal cavity of 8-week-old female athymic Nu/Nu mice. Tumor burden for mice injected intraperitoneally was assessed by pathology after necropsy and collection of organs. For subcutaneous xenografts, 1 × 106 ACI23 cells in 500 μL were subcutaneously injected into the right flank of 8-week-old female athymic Nu/Nu mice. For controls, 500 μL PBS was injected into the left flank of each mouse. Mice were weighed and tumors measured using calipers twice weekly. Mice were sacrificed 24 days post inoculation. Single-cell suspensions were prepared from tumors using the Human Tumor Dissociation Kit and gentleMACS Dissociator (Miltenyi Biotec, Inc.). For bursa xenografts, 2 × 104 OV90 cells in 5 μL were injected into the right bursa of 8-week-old female athymic Nu/Nu mice. For controls, 5 μL PBS was injected into the left bursa of each mouse. Mice were sacrificed 65 days post inoculation. Ovaries were removed, weighed and fixed in 4% paraformaldehyde. Sections of liver, pancreas, peritoneal wall, and diaphragm were removed and fixed in 4% paraformaldehyde to assess metastatic spread. Organ sections were stained with hematoxylin and eosin and analyzed by a pathologist.

RNA extraction and quantitative real-time PCR

Total RNA was isolated and purified using the RNeasy Mini Kit (Qiagen) as per the manufacturer's protocol and treated with DNAse. Final RNA concentration was determined with NanoDrop spectrophotometer using the 260/280 absorbance ratio. Total purified RNA was reverse transcribed with random primers using TaqMan reagents (Applied Biosystems). Resulting cDNA was used as a template for quantitative real-time PCR (qRT-PCR). Analysis of gene expression was performed on ViiA7 Real-time PCR System (Applied Biosystems) using TaqMan probe assays and GAPDH as control. Quantitation and normalization of relative gene expression were accomplished using comparative threshold cycle method or ΔΔCT. Catalog numbers for commercial primers are provided in Supplementary Table S2.

Gene expression

Total RNA was isolated from three independent cultures of ovarian cancer cells grown on traditional or ultralow attachment 75-cm cell culture flasks in presence of traditional serum-containing media or serum-free stem cell media. Samples were submitted to Laboratory of Molecular Technology, SAIC/NCI-Frederick and RNA quality checked on an Agilent Bioanalyzer; only samples having high quality score (RIN>9) were used for microarray analysis. RNA (1 μg) was reverse transcribed with T7-oligo (dT) primers and labeled with biotin using the Affymetrix One Cycle Target Labeling Kit. All replicates of experimental samples were prepared, labeled, and hybridized to Affymetrix H133 Plus 2.0 gene chips and scanned on an Affymetrix GeneChip scanner 3000 (Affymetrix). Datasets are available on the GEO repository (GSE94358).

Western blot analysis

Whole-cell protein was extracted from ovarian cancer cell lines using standard methods with NP-40 lysis buffer. BCA Protein Assay Kit (Thermo Scientific) was used to determine protein concentrations. 30 μg protein was loaded for each experiment. SDS-PAGE was performed using NuPage system (Invitrogen) and Luminata HRP Chemiluminescent Detection Reagents (Millipore). Antibodies were purchased from Cell Signaling Technology (p-p65, 3033S; RelB, 4922S), Millipore (p52/100, 05-361; GAPDH, MAB374), Santa Cruz Biotechnology (p65, SC-372), Abcam (ALDH1, ab52492; ALDH1A2, ab96060), and Miltenyi Biotec (CD133, 130-092-395). Densitometry analysis was performed using ImageJ Software 1.48v.

Immunofluorescence

Ovarian cancer cells or spheroids from traditional adherent or TIC-enriching culture conditions, respectively, were collected after 5 days in culture. Cells were pelleted, washed with PBS, and fixed at room temperature using 4% paraformaldehyde and permeabilized with 0.3% Triton X-100. Cells were sequentially stained in following order and condition: ALDH1A1 or ALDH1A2 rabbit (Abcam; 1:100 overnight at 4 degrees), Alexafluor 594 monovalent Fab fragment goat anti-rabbit secondary (Jackson Immunoresearch; 1:50 1 hour at room temperature), Alexafluor 647 Ki67 rabbit antibody (BioLegend; 1:50 overnight at 4 degrees), Alexafluor 555 RelB rabbit antibody (Bioss; 1:100 1 hour at room temperature), and Alexafluor 488 p65 rabbit antibody (Abcam; 1:100 1 hour at room temperature). Cells were washed with PBS between each antibody staining. Finally, cells were stained with DAPI (Thermo Scientific; 1:1,000) for 15 minutes at room temperature and mounted on slides using ProLong Diamond antifade mountant (Life Technologies). Images were acquired with a Zeiss LSM880 NLO laser scanning confocal microscope equipped with 34 channel spectral detector and 63x plan-apochromat (N.A. 1.4) oil immersion objective lens (Carl Zeiss Microscopy). Image acquisition details can be found in Supplementary Experimental Procedures. Further quantification and statistical analysis of fluorescence levels in populations of individual cells was performed using custom-written software in Matlab (Mathworks).

RNA interference

Silencing of RELA, RELB, GAPDH positive control, and nontargeting negative control was achieved using SMARTchoice inducible lentiviral shRNAs (GE Dharmacon) designed with the human EF1α promoter and red fluorescent protein (TurboRFP) reporter gene. OV90, ACI23 and VTCF003 cells were transduced according to the manufacturer's protocol. Briefly, cells were transduced with viral particles in serum-free transduction media for 24 hours followed by selection with 2 μg/mL puromycin for 7 days. For transient gene knockdown studies On-Target Plus Smartpool siRNAs were purchased from Dharmacon for ALDH1A1 (catalog no. L-008722-00), ALDH1A2 (catalog no. L-008118-00) and nontargeting control (catalog no. D-001810-10-20). Briefly, cells were seeded at 60% confluence and transfected with 20 nmol/L siRNA for 24 hours before proceeding with functional assays.

Flow cytometry

ALDH activity was evaluated using the Aldefluor Kit (StemCell Technologies) according to the manufacturer's instructions. Following staining procedure for ALDH, cells were incubated with APC-conjugated CD133 antibody (Miltenyi Biotec) 1:20 in the Aldefluor assay buffer for 30 minutes on ice protected from light. To evaluate proliferative populations cells were fixed and permeablilized as above and incubated with Alexafluor 647 Ki67 rabbit antibody (BioLegend) 1:100 for 30 minutes on ice. Flow cytometry data collection was performed on BD FACSCalibur or LSRFortessa using CellQuest Pro or FACSDiva acquisition software, respectively (Becton Dickinson). Data were analyzed using FlowJo software (Tree Star). Sorting of ALDH positive and negative cells was performed on a BD FACSAria Fusion sorter.

Spheroid formation assay

To generate ACI23 and OV90 spheroids, 500 cells/well in 50 μL were seeded in TIC-enriching conditions (stem cell media in ultra-low attachment flat bottom 96 well plates; Corning). After 7 days in culture, 3 fields per well were imaged at ×10 magnification using Axio Observer A1 Inverted Microscope and analyzed using ImageJ Software 1.48v. Spheroids with diameter ≥50 μm were quantified. Spheroid efficiency was calculated as spheroids ≥50 microns/cells per well).

Cell viability assay

Cell viability was assessed using CellTiter Glo luminescent reagent (Promega). Hundred μL of reagent was added to each well after analysis; 250 μL of cells with luminescent reagent was transferred to a white, flat-bottom 96-well plate (Corning). SpectraMax M5 was used to quantify cell viability using SoftMax Pro Software 5.8.3.

Chromatin immunoprecipitation-qPCR assay

The SimpleChIP Enzymatic Chromatin IP Kit (magnetic beads) was purchased from Cell Signaling Technology and assays performed according to the manufacturer's instructions. Antibodies for RelB were also from Cell Signaling Technology (catalog #10544). Promo was used to evaluate DNA sequences for transcription factor binding sites (http://alggen.lsi.upc.es/cgi-bin/promo_v3/promo/promoinit.cgi?dirDB=TF_8.3). The first 5,000 bases upstream of the transcription start site were screened for binding motifs that correspond to NFκB consensus–binding sequence. The quantification of transcription factor binding to target genes was calculated by measuring the ratio of chromatin immunoprecipitation (ChIP)-to-Input and the normal rabbit IgG antibody served as a negative control. All primers used for ChIP-PCRs are listed in Supplementary Table S2.

Statistical analysis

Raw microarray (CEL) data were processed using robust multi-array analysis (RMA) using the oligo package in Bioconductor (17). After quality control check, 2 outlier samples were removed, and differentially expressed genes were generated using Limma package taking in account batch effect (18, 19). Other statistics were performed in Prism7 (GraphPad) using data acquired from at least three biological replicates as noted. Statistical analysis of immunofluorescence images was performed in Matlab (Mathworks). P values were calculated as described in figure legends.

We previously showed that ovarian cancer cells grown in TIC-enriching conditions form spheroid structures with high ALDH activity and are more tumorigenic than the same cells grown in traditional culture conditions (20). In this study, we sought to identify gene-expression changes incurred by cells grown in TIC-enriched conditions and to characterize the role of classical and/or alternative NFκΒ in supporting cells in TIC cultures.

Differential gene expression was determined by comparing profiles from cells grown in traditional adherent culture (A) with cells grown in TIC-enriching conditions (T). Genes were selected as differentially expressed on the basis of cutoff values of fold change >2.0 and P < 0.05. TIC-enriching conditions led to downregulation of genes encoding extracellular matrix proteins, TREM1 signaling (P < 10−4), and the NFκB complex (P < 10−8; Fig. 1A and B). Conversely these conditions led to an upregulation of genes encoding SMAD 1/5/8 and ID (inhibitor of DNA binding) family transcription factors (P < 10−6; Fig. 1A and C). We confirmed that TIC culture conditions led to differential regulation of NFκB genes in two ovarian cancer cell lines (Fig. 1D). Notably, CXCL1 and CXCL2 were downregulated in both cell lines grown in TIC culture conditions. We have previously shown that these cytokines are regulated by classical NFκB signaling in ovarian cancer (6). CD44, PAX8, GAD1, and TFP12 have known NFκB–binding sites in their promoter regions (21–23). We show here that they are also differentially expressed in TIC-enriching conditions, suggesting a potential role for NFκB in regulating the transition between tumor-initiating and proliferative cell behavior.

Figure 1.

Differential NFκB signaling in TIC-enriching culture conditions. A, Up- and downregulated genes were examined in ACI23 cells cultured in TIC and adherent conditions. Heat map summarizes genes with fold change >2 and P < 0.05 in either direction (red, up; green, down) from three biological replicates. B, IPA analysis designated the NFκB complex as a top upstream regulator. NFκB predicted to be inhibited in TIC cultures relative to adherent cultures. C, The third-ranked enriched canonical pathway identified using IPA, embryonic stem cell pluripotency, indicates increased SMAD and ID signaling in TIC cultures relative to adherent cultures. D, qRT-PCR of selected NFκB–regulated genes relative to GAPDH housekeeping gene in ACI23 and OV90 ovarian cancer cell lines grown in TIC-enriching conditions for 5 days relative to cells grown in traditional adherent cultures. Data shown represent average and SEM from three independent experiments *, P < 0.05, Student t test.

Figure 1.

Differential NFκB signaling in TIC-enriching culture conditions. A, Up- and downregulated genes were examined in ACI23 cells cultured in TIC and adherent conditions. Heat map summarizes genes with fold change >2 and P < 0.05 in either direction (red, up; green, down) from three biological replicates. B, IPA analysis designated the NFκB complex as a top upstream regulator. NFκB predicted to be inhibited in TIC cultures relative to adherent cultures. C, The third-ranked enriched canonical pathway identified using IPA, embryonic stem cell pluripotency, indicates increased SMAD and ID signaling in TIC cultures relative to adherent cultures. D, qRT-PCR of selected NFκB–regulated genes relative to GAPDH housekeeping gene in ACI23 and OV90 ovarian cancer cell lines grown in TIC-enriching conditions for 5 days relative to cells grown in traditional adherent cultures. Data shown represent average and SEM from three independent experiments *, P < 0.05, Student t test.

Close modal

Next, we examined changes in mRNA and protein levels of both classical and alternative NFκB transcription factors, in a panel of ovarian cancer cell lines grown in TIC-enriching conditions compared with traditional adherent conditions. Two of six cell lines tested showed a significant increase in the transcript level of RELA, the gene encoding the classical NFκB transcription factor p65 and four of six lines showed a significant increase in the transcript level of RELB, the gene encoding the alternative NFκB transcription factor RelB (Fig. 2A). Moreover, relative to A culture, T culture led to ≥1.5-fold increase in protein levels of RelB in seven out of nine ovarian cancer lines, including three primary lines (Fig. 2B).

Figure 2.

Baseline expression of classical and alternative NFκB transcription factors. A, Gene-expression fold changes of RELA and RELB in a panel of ovarian cancer cell lines grown in TIC-enriching conditions for 5 days relative to cells grown in traditional adherent cultures. Data shown represent average and SEM from three independent experiments *, P < 0.05, Student t test. B, Western blot analysis showing baseline protein levels of classical NFκB transcription factor RelA (p65) and alternative NFκB transcription factors p100/52 and RelB in adherent (A), TIC (T), or floater (F) cultures. Densitometric analysis of protein level fold changes in F or T culture relative to A culture is shown in the heatmap. C, Confocal microscopy examining total cellular and nuclear staining of RelA and RelB in adherent (A) or TIC (T) conditions. Representative images shown at bottom. Box plots represent quantitation of at least 500 cells for each condition. Error bars, SDs. P values for cellular RelA < 3.4 × 10−5 and cellular RelB < 2.0 × 10−218 and nuclear RelA < 6.6 × 10−32 and nuclear RelB < 1.2 × 10−219 as determined by paired t tests; scale bar, 10 μm. D, Imaging flow cytometer evaluation of RelA in TIC (T) or adherent (A) conditions. Representative images of nuclear and cytoplasmic RelA shown on top. Histograms demonstrate change in localization analyzed using similarity feature. Bar graph represents quantitation of 10,000 cells per sample.

Figure 2.

Baseline expression of classical and alternative NFκB transcription factors. A, Gene-expression fold changes of RELA and RELB in a panel of ovarian cancer cell lines grown in TIC-enriching conditions for 5 days relative to cells grown in traditional adherent cultures. Data shown represent average and SEM from three independent experiments *, P < 0.05, Student t test. B, Western blot analysis showing baseline protein levels of classical NFκB transcription factor RelA (p65) and alternative NFκB transcription factors p100/52 and RelB in adherent (A), TIC (T), or floater (F) cultures. Densitometric analysis of protein level fold changes in F or T culture relative to A culture is shown in the heatmap. C, Confocal microscopy examining total cellular and nuclear staining of RelA and RelB in adherent (A) or TIC (T) conditions. Representative images shown at bottom. Box plots represent quantitation of at least 500 cells for each condition. Error bars, SDs. P values for cellular RelA < 3.4 × 10−5 and cellular RelB < 2.0 × 10−218 and nuclear RelA < 6.6 × 10−32 and nuclear RelB < 1.2 × 10−219 as determined by paired t tests; scale bar, 10 μm. D, Imaging flow cytometer evaluation of RelA in TIC (T) or adherent (A) conditions. Representative images of nuclear and cytoplasmic RelA shown on top. Histograms demonstrate change in localization analyzed using similarity feature. Bar graph represents quantitation of 10,000 cells per sample.

Close modal

We used confocal microscopy to evaluate differences in cellular intensity of both RelA and RelB in primary ovarian cancer cells (VTCF) obtained from patient ascites. These experiments show both RelA and RelB mean cellular intensities are significantly increased in the spheroids grown in TIC conditions (Fig. 2C) relative to cells grown in adherent conditions in agreement with the Western blot findings. The same trend was observed for nuclear intensities. The enhanced fluorescence intensities measured for RelB far exceeded that measured for RelA. Our expression-profiling data suggest that RelA target genes are less active in the TIC conditions relative to the adherent conditions. Using an imaging flow cytometer, we evaluated the localization of RelA as nuclear or cytoplasmic in ovarian cancer cells expressing the ALDH protein in adherent and TIC growth conditions. Costaining with DAPI, we measured the localization of RelA in ALDH-positive cells. RelA distribution was both nuclear and cytoplasmic in either growth condition, but there was a decrease in the percentage of cells with nuclear RelA in TIC conditions compared with adherent conditions (Fig. 2D). This suggests that RelA is more active in the nucleus of cells growing in adherent conditions.

These data suggest that alternative NFκB signaling may be more important for supporting cells with TIC features. To further explore this possibility and clarify the roles of classical versus alternative NFκB signaling, we stably transfected ACI23, OV90, and VTCF ovarian cancer cells with inducible shRNAs targeting RELA or RELB and assessed the functional significance of these proteins relative to nontargeting control shRNA (shNeg). We chose ACI23 and OV90 cell lines because in culture they closely resemble the morphology of primary cells obtained from patients; furthermore, they have an appreciable number of cells positive for both CD133 expression and ALDH activity, established markers for ovarian TICs (20, 24). Greater than 70% diminution of RELA or RELB transcript levels, following 72 hours exposure to 1 μg/mL doxycycline, was confirmed using qRT-PCR for each cell line (Fig. 3A). Protein level decrease was confirmed by Western blot analysis. Efficiency of protein knockdown was comparable for both A and T cultures (Fig. 3B). Loss of RELB had no effect on the levels of RelA; however, loss of RELA led to a small decrease in the levels of the alternative signaling proteins, p52 and RelB. These are known target genes for classical NFκB.

Figure 3.

RelA and RelB support ovarian cancer tumor formation. A, Gene-expression fold changes of RELA and RELB after stable transfection of ACI23 and OV90 ovarian cancer cell lines with inducible shRNAs against RELA (shRelA), RELB (shRelB), or nontargeting control transcript (shNeg). Data shown represent average fold changes and SEM relative to nontargeting negative control shRNA from three independent experiments. *, P < 0.05, Student t test. B, Whole-cell lysate Western blot analysis showing protein level decrease after 72-hour exposure to 1 μg/mL doxycycline in adherent (A) and TIC-enriched (T) cultures. C, Intraperitoneal injection of 1 × 106 OV90 cells stably transfected with shRNAs. Immunohistochemical analysis of RFP-positive cells in resected organs showing macrometastases. D, Subcutaneous injection of 1 × 106 ACI-23 cells stably transfected with shRNAs. Tumors were monitored and measured twice weekly. Mice were sacrificed 24 days following tumor cell inoculation. Shown is the average tumor volume and SEM for each group of 5 mice from three independent experiments. Representative images of resected tumors are shown; scale bar, 1 cm. At day 24, *, P = 0.0047 for RELA and 0.0001 for RELB, two-way ANOVA with Dunnett post hoc test. Differences were not significant at any other time points. E, Single-cell suspensions of resected tumors were analyzed by Western blot analysis for RelA and RelB to confirm knockdown of protein was maintained. F, Intrabursal injection of 2 × 104 OV90 cells stably transfected with shRNAs into the intrabursal sac of 10 mice. Mice were weighed twice weekly and sacrificed 60 days after tumor cell inoculation. Ovaries were resected and weighed. Shown are the average tumor weights and SEM for each group of 5 mice from two independent experiments. Representative images of xenograft ovaries are shown; scale bar, 1 cm. *, P = 0.0181 for RELA and 0.0187 for RELB, one-way ANOVA with Dunnett post hoc test. G, Hematoxylin and eosin–stained ovary sections were analyzed by a pathologist for evaluation of invasiveness. Representative images of invasive and noninvasive xenograft observed in shNeg samples and no xenograft observed in shRelA sample. Scale bar, 300 μm. H, In addition to ovaries, sections of liver, pancreas, peritoneal wall, and diaphragm were removed at necropsy and fixed in 4% paraformaldehyde. Tissues were stained with hematoxylin and eosin and analyzed by a pathologist for metastatic spread in all 10 mice for each condition. Metastases were present in contralateral ovaries, diaphragm, pancreas, peritoneal wall, and liver in the shNeg mice. No metastases were seen with shRelA mice. Only contralateral metastases were seen in one shRelB mouse *, P < 0.05, χ2 test.

Figure 3.

RelA and RelB support ovarian cancer tumor formation. A, Gene-expression fold changes of RELA and RELB after stable transfection of ACI23 and OV90 ovarian cancer cell lines with inducible shRNAs against RELA (shRelA), RELB (shRelB), or nontargeting control transcript (shNeg). Data shown represent average fold changes and SEM relative to nontargeting negative control shRNA from three independent experiments. *, P < 0.05, Student t test. B, Whole-cell lysate Western blot analysis showing protein level decrease after 72-hour exposure to 1 μg/mL doxycycline in adherent (A) and TIC-enriched (T) cultures. C, Intraperitoneal injection of 1 × 106 OV90 cells stably transfected with shRNAs. Immunohistochemical analysis of RFP-positive cells in resected organs showing macrometastases. D, Subcutaneous injection of 1 × 106 ACI-23 cells stably transfected with shRNAs. Tumors were monitored and measured twice weekly. Mice were sacrificed 24 days following tumor cell inoculation. Shown is the average tumor volume and SEM for each group of 5 mice from three independent experiments. Representative images of resected tumors are shown; scale bar, 1 cm. At day 24, *, P = 0.0047 for RELA and 0.0001 for RELB, two-way ANOVA with Dunnett post hoc test. Differences were not significant at any other time points. E, Single-cell suspensions of resected tumors were analyzed by Western blot analysis for RelA and RelB to confirm knockdown of protein was maintained. F, Intrabursal injection of 2 × 104 OV90 cells stably transfected with shRNAs into the intrabursal sac of 10 mice. Mice were weighed twice weekly and sacrificed 60 days after tumor cell inoculation. Ovaries were resected and weighed. Shown are the average tumor weights and SEM for each group of 5 mice from two independent experiments. Representative images of xenograft ovaries are shown; scale bar, 1 cm. *, P = 0.0181 for RELA and 0.0187 for RELB, one-way ANOVA with Dunnett post hoc test. G, Hematoxylin and eosin–stained ovary sections were analyzed by a pathologist for evaluation of invasiveness. Representative images of invasive and noninvasive xenograft observed in shNeg samples and no xenograft observed in shRelA sample. Scale bar, 300 μm. H, In addition to ovaries, sections of liver, pancreas, peritoneal wall, and diaphragm were removed at necropsy and fixed in 4% paraformaldehyde. Tissues were stained with hematoxylin and eosin and analyzed by a pathologist for metastatic spread in all 10 mice for each condition. Metastases were present in contralateral ovaries, diaphragm, pancreas, peritoneal wall, and liver in the shNeg mice. No metastases were seen with shRelA mice. Only contralateral metastases were seen in one shRelB mouse *, P < 0.05, χ2 test.

Close modal

Our prior studies implicate NFκB in ovarian cancer aggressiveness and show that RelA and RelB are coexpressed in ovarian cancer tumor specimens (6, 7). We measured the role of RELA or RELB in tumor development in athymic nude mice by injecting ACI23 or OV90 cells stably transfected with inducible shRNAs (cultured in T conditions) by intraperitoneal, subcutaneous, or intrabursal routes. All mice were fed doxycycline chow for two days before inoculation and throughout the remainder of the study to ensure persistent knockdown of RELA or RELB. The shRNA constructs coexpress RFP also under the inducible promoter, so that induction can be identified by immune-staining for this protein. When transduced cells were injected intraperitoneally, dramatically fewer RFP-positive cells were present in mice inoculated with cells containing shRelB construct compared with mice injected with cells containing shRelA or shNeg constructs (Fig. 3C). The intraperitoneal model suggests that RelB is required for tumor growth in the “unattached” or suspension setting. We next used the ACI23 cells to monitor subcutaneous tumor formation as this model was previously shown to give enhanced tumorigenic activity in T cultures compared with A cultures (20). Tumor growth was significantly diminished with knockdown of RelB, although the effect was less than in the intraperitoneal model (Fig. 3D). To verify shRNA activity was maintained we resected tumors and assessed protein levels via Western blot analysis (Fig. 3E).

We next used OV90 cells to monitor intrabursal tumor growth as we recently found these cells to be highly tumorigenic in this model (24). This model most closely mimics human ovarian cancer, where tumors begin in the ovary or fallopian tube and then metastasize throughout the abdomen. Compared with mice that received cells containing the negative control shRNA (shNeg), mice implanted with cells containing either the RELA or RELB shRNAs gained significantly less tumor burden as measured using tumor weights (Fig. 3F). Invasive xenograft cells were only present in the contralateral PBS-injected control ovaries of mice inoculated with shNeg cells, and not in mice that received cells depleted of either RELA or RELB (Fig. 3G). Moreover, the ability to metastasize to abdominal organs was significantly reduced by knockdown of either RELA or RELB (Fig. 3H).

Tumor cell viability was significantly decreased in vitro for both ACI23 and OV90 cells with RELA knockdown but not with RELB knockdown (Fig. 4A). These data are intriguing given the significant decrease in tumorigenesis observed in vivo with loss of either transcription factor in both cell lines. We investigated whether either of these transcription factors supports features associated with TICs. Spheroid formation efficiency was significantly decreased with loss of either RELA or RELB in both ovarian cancer cell lines (Fig. 4B). Our previous experiment showed, however, that there was no change in viability or proliferation with RELB knockdown. Therefore, these data suggest that the absence of RELB prevents efficient formation of spheroids, even though the same number of viable cells is present.

Figure 4.

RelA and RelB differentially support viability but equally support spheroid formation. A, ACI23 and OV90 cells were grown in TIC conditions for 72 hours. A total of 2 × 103 viable cells were seeded in ultralow attachment 96-well plates and allowed to grow under doxycycline induction for 72 hours. Viability was calculated relative to the negative control shRNA. *, P = 0.0017 for RELA in ACI23 and 0.0001 for RELA in OV90, one-way ANOVA with Dunnett post hoc test. B, A total of 5 × 102 cells were seeded in ultralow attachment 96-well plates with TIC culture media and cultured for 7 days. Three image fields were obtained for each well. *, P = 0.0116 for RelA and 0.0005 for RelB in ACI23; P = 0.0007 and 0.0251 for RELB in OV90, one-way ANOVA with Dunnett post hoc test. C, ACI23 and OV90 cells grown in TIC conditions for 72 hours before exposure to doxycycline for an additional 72 hours. Cells were fixed, permeabilized, and stained with Ki67 antibody conjugated to Alexa 647 fluorophore and quantitated by flow cytometry. *, P = 0.0178 for RELA in OV90, one-way ANOVA with Dunnett post hoc test. D, Fold change in geometric mean Ki67 fluorescence intensity in shRNA-expressing cells. Representative histograms are shown on left. *, P = 0.191 for ACI23 and 0.0149 for OV90 cells, Kruskal–Wallis test. Data shown for all panels represent average and SEM from five independent experiments. ns, nonsignificant.

Figure 4.

RelA and RelB differentially support viability but equally support spheroid formation. A, ACI23 and OV90 cells were grown in TIC conditions for 72 hours. A total of 2 × 103 viable cells were seeded in ultralow attachment 96-well plates and allowed to grow under doxycycline induction for 72 hours. Viability was calculated relative to the negative control shRNA. *, P = 0.0017 for RELA in ACI23 and 0.0001 for RELA in OV90, one-way ANOVA with Dunnett post hoc test. B, A total of 5 × 102 cells were seeded in ultralow attachment 96-well plates with TIC culture media and cultured for 7 days. Three image fields were obtained for each well. *, P = 0.0116 for RelA and 0.0005 for RelB in ACI23; P = 0.0007 and 0.0251 for RELB in OV90, one-way ANOVA with Dunnett post hoc test. C, ACI23 and OV90 cells grown in TIC conditions for 72 hours before exposure to doxycycline for an additional 72 hours. Cells were fixed, permeabilized, and stained with Ki67 antibody conjugated to Alexa 647 fluorophore and quantitated by flow cytometry. *, P = 0.0178 for RELA in OV90, one-way ANOVA with Dunnett post hoc test. D, Fold change in geometric mean Ki67 fluorescence intensity in shRNA-expressing cells. Representative histograms are shown on left. *, P = 0.191 for ACI23 and 0.0149 for OV90 cells, Kruskal–Wallis test. Data shown for all panels represent average and SEM from five independent experiments. ns, nonsignificant.

Close modal

We hypothesized that classical and alternative NFκB signaling might be relevant in different subpopulations of ovarian cancer cells in TIC-enriched cultures. It has been established that classical NFκB signaling supports an antiapoptotic, proliferative phenotype that sustains tumorigenesis (6, 25–27); however, the mechanism of alternative NFκB signaling in supporting tumor growth is unclear. We tested whether RELA might support a more proliferative subpopulation as would be expected given the traditional role of RELA in tumorigenesis. Cells positive for the proliferation marker Ki67 were quantified by flow cytometry in each of the knockdown conditions. OV90 cells showed a significant loss of Ki67-positive cells when RELA was depleted, but not with depletion of RELB (Fig. 4C, right). A similar trend was observed in ACI23 cells that approached statistical significance (Fig. 4C, left). We measured the mean fluorescence intensity of Ki67 and found that both cell lines had a significant decrease in intensity specifically with depletion of RELA relative to negative control shRNA (Fig. 4D).

Chemotherapy resistance is another feature commonly associated with TICs (11, 28–30). We first verified that TIC culture conditions enriched for cells with carboplatin resistance by measuring viability across a range of concentrations in ACI23 and OV90 cells (Fig. 5A). IC50 was calculated using CompuSyn software. Culture in TIC conditions caused greater than 2-fold increase in carboplatin IC50 for both cell lines.

Figure 5.

RelB supports resistance to carboplatin. A, A total of 2 × 103 viable cells were seeded in adherent (A) or ultralow attachment (T) 96-well plates with traditional or stem cell media, respectively. The next day, a range of carboplatin concentrations was added and cells were cultured for 72 hours. Viability was calculated relative to vehicle control. B, A total of 1 × 106 ACI23 and OV90 cells were seeded in adherent plates and allowed to attach overnight. The next day, 30 μmol/L of carboplatin was added and cells were cultured an additional 48 hours. CD133 expression and ALDH activity was quantitated by flow cytometry and assessed relative to vehicle control. *, P < 0.0001 for ALDH+CD133+ in ACI23; P < 0.0231 for ALDH+ in OV90, one-way ANOVA with Dunnett post hoc test. C, Fold change in geometric mean of ALDH fluorescence with carboplatin treatment significantly increased relative to vehicle control. *, P = 0.008 for ACI23; P = 0.002 for OV90, Mann–Whitney test. D, Top, cells were seeded in adherent conditions and allowed to attach overnight. Cells were then treated with 100 μmol/L of carboplatin for 6 hours, followed by whole-cell lysate preparation and Western blot analysis. Bottom, densitometric analysis relative to vehicle for three independent blots. E, A total of 2 × 103 viable cells grown in TIC conditions for 72 hours were seeded in ultralow attachment 96-well plates with doxycycline. The next day, 30 μmol/L of carboplatin was added and cells were cultured for 72 hours. Viability was calculated relative to vehicle control for each shRNA group. *, P < 0.03 for RelB in ACI23; *, P < 0.008 for RELB in OV90; #, P < 0.016 for RELA in OV90, one-way ANOVA with Dunnett post hoc test. Bar graphs for all panels represent average and SEM obtained from at least three independent experiments.

Figure 5.

RelB supports resistance to carboplatin. A, A total of 2 × 103 viable cells were seeded in adherent (A) or ultralow attachment (T) 96-well plates with traditional or stem cell media, respectively. The next day, a range of carboplatin concentrations was added and cells were cultured for 72 hours. Viability was calculated relative to vehicle control. B, A total of 1 × 106 ACI23 and OV90 cells were seeded in adherent plates and allowed to attach overnight. The next day, 30 μmol/L of carboplatin was added and cells were cultured an additional 48 hours. CD133 expression and ALDH activity was quantitated by flow cytometry and assessed relative to vehicle control. *, P < 0.0001 for ALDH+CD133+ in ACI23; P < 0.0231 for ALDH+ in OV90, one-way ANOVA with Dunnett post hoc test. C, Fold change in geometric mean of ALDH fluorescence with carboplatin treatment significantly increased relative to vehicle control. *, P = 0.008 for ACI23; P = 0.002 for OV90, Mann–Whitney test. D, Top, cells were seeded in adherent conditions and allowed to attach overnight. Cells were then treated with 100 μmol/L of carboplatin for 6 hours, followed by whole-cell lysate preparation and Western blot analysis. Bottom, densitometric analysis relative to vehicle for three independent blots. E, A total of 2 × 103 viable cells grown in TIC conditions for 72 hours were seeded in ultralow attachment 96-well plates with doxycycline. The next day, 30 μmol/L of carboplatin was added and cells were cultured for 72 hours. Viability was calculated relative to vehicle control for each shRNA group. *, P < 0.03 for RelB in ACI23; *, P < 0.008 for RELB in OV90; #, P < 0.016 for RELA in OV90, one-way ANOVA with Dunnett post hoc test. Bar graphs for all panels represent average and SEM obtained from at least three independent experiments.

Close modal

CD133 expression and ALDH activity have been the most consistent markers of ovarian cancer stem cells (31–35). For this reason, we proceeded to measure changes in these markers upon exposure to carboplatin. Carboplatin treatment led to a significant increase in the percentage of cells containing both ALDH activity and CD133 expression in ACI23 cells but had little effect on the CD133-positive cells relative to vehicle treatment (Fig. 5B). Carboplatin treatment of OV90 cells led to a significant increase in the percentage of cells with ALDH activity and no effect on CD133-positive cells (Fig. 5B). Quantitation of ALDH fluorescence intensity was significantly increased with carboplatin treatment in both cell lines (Fig. 5C). Taken together, these data suggest that ALDH activity in our system more reliably associates with chemotherapy resistance, and both are increased in conditions that enrich for tumorigenesis.

We analyzed the influence of RELA and RELB in supporting carboplatin resistance. Cells remaining after 6-hour exposure to 100 μmol/L carboplatin were collected, and the lysates analyzed for protein changes. RelB increased in both cell lines after this treatment, whereas no obvious change was evident in total or phosphorylated RelA (Fig. 5D). Viability assays demonstrate that loss of RELB was relatively better at restoring sensitivity to carboplatin (Fig. 5E), again highlighting a key function of RelB in promoting the chemoresistant phenotype of TICs.

Flow-cytometry experiments confirmed that loss of RELB significantly reduced the ALDH+CD133+ population in vitro for both ACI23 and OV90 cells (Fig. 6A and B). Interestingly, depletion of either RELA or RELB led to a statistically significant diminution of TICs in cells derived from in vivo tumors (Fig. 6C). It is important to note that the duration of knockdown in the in vivo tumors was substantially longer than the in vitro experiments (6 weeks vs. 6 days). Given the known dynamic interplay between classical and alternative NFκB signaling, it is difficult to pinpoint an initial effect in the long-term setting.

Figure 6.

RelB supports CD133+ALDH+ ovarian cancer cells. A, ACI23 cells were grown in TIC conditions for 72 hours before addition of doxycycline for an additional 72 hours. Cells were analyzed using flow cytometry by first gating on the RFP channel to select the shRNA-expressing cells and subsequent quantitation of CD133 expression and ALDH activity. Data shown are average and SEM percent double-positive cells relative to negative control shRNA from five independent experiments. *, P = 0.0244, one-way ANOVA with Dunnett post hoc. B, OV90 cells were treated and analyzed as in A. *, P = 0.0070, one-way ANOVA with Dunnett post hoc. C, Single-cell suspensions of ACI23 xenograft tumors were freshly prepared and analyzed by flow cytometry as in A. Data shown represent average and SEM obtained from tumors resected from three mice in each group. *, P = 0.0382 for RELA and 0.0152 for RELB, one-way ANOVA with Dunnett post hoc. D, Fold change in geometric mean of ALDH fluorescence in shRNA-expressing cells. ALDH activity significantly decreased with loss of RELB relative to control shRNA. Representative histograms shown for ACI23 and OV90 cells. Bar graphs represent average and SEM obtained from at least three independent experiments. *, P = 0.007 for ACI23; P = 0.029 for OV90; P = 0.019 for ACI23 xenograft–derived cells; #, P = 0.140 for VTCF, Kruskal–Wallis test. ns, nonsignificant.

Figure 6.

RelB supports CD133+ALDH+ ovarian cancer cells. A, ACI23 cells were grown in TIC conditions for 72 hours before addition of doxycycline for an additional 72 hours. Cells were analyzed using flow cytometry by first gating on the RFP channel to select the shRNA-expressing cells and subsequent quantitation of CD133 expression and ALDH activity. Data shown are average and SEM percent double-positive cells relative to negative control shRNA from five independent experiments. *, P = 0.0244, one-way ANOVA with Dunnett post hoc. B, OV90 cells were treated and analyzed as in A. *, P = 0.0070, one-way ANOVA with Dunnett post hoc. C, Single-cell suspensions of ACI23 xenograft tumors were freshly prepared and analyzed by flow cytometry as in A. Data shown represent average and SEM obtained from tumors resected from three mice in each group. *, P = 0.0382 for RELA and 0.0152 for RELB, one-way ANOVA with Dunnett post hoc. D, Fold change in geometric mean of ALDH fluorescence in shRNA-expressing cells. ALDH activity significantly decreased with loss of RELB relative to control shRNA. Representative histograms shown for ACI23 and OV90 cells. Bar graphs represent average and SEM obtained from at least three independent experiments. *, P = 0.007 for ACI23; P = 0.029 for OV90; P = 0.019 for ACI23 xenograft–derived cells; #, P = 0.140 for VTCF, Kruskal–Wallis test. ns, nonsignificant.

Close modal

Because ALDH is associated with chemotherapy resistance in our model and we showed RelB sensitizes ovarian cancer cells to chemotherapy, we analyzed ALDH fluorescence intensity after knockdown. Loss of RELB, but not RELA, led to a significant decrease in ALDH fluorescence intensity in cells obtained from both in vitro and in vivo conditions (Fig. 6D). The VTCF primary ovarian cancer cells have decreased ALDH intensity with loss of RELB that approaches significance relative to the negative control shRNA. Even with the limited expansion imposed by primary cells in culture, and short-term selection and induction periods, the decrease in ALDH activity with RELB loss was notable, indicating an important biological effect of this pathway.

Taken together our data suggest that both RELA and RELB support cells with TIC phenotypes, and point toward a greater influence of RELB on spheroid formation, chemoresistance and ALDH activity. We questioned whether the transcription factor RelB regulated expression of genes linked to stem cell functions. We first quantitated gene-expression levels of two ALDH isozymes and three stem cell transcription factors in cells grown in adherent or TIC-enriching conditions. There was significantly enriched expression of ALDH1A2, OCT4, and NANOG in OV90 cells grown in TIC conditions (Fig. 7A). Gene expression in cells sorted for ALDH activity (ALDH+) show a significant enrichment in RELB, ALDH1A1, and ALDH1A2 expression in both cell lines relative to cells lacking ALDH activity (ALDH; Fig. 7B). Differences in gene expression after knockdown of RELA or RELB further implicate RELB in regulating ALDH1A2 gene expression (Fig. 7C). Loss of RELA significantly decreased NANOG expression; however, loss of RELB significantly decreased both NANOG and ALDH1A2, and neither affected ALDH1A1. Furthermore, RELB knockdown significantly decreased expression of putative TIC gene CD44, consistent with previously published data showing that this gene is regulated by RelB in other cell types (22). Western blot analysis confirmed that RelB knockdown led to a decrease in ALDH1A2 protein levels (Fig. 7D).

Figure 7.

RelB regulates expression of ALDH1A2. A, qRT-PCR of selected stem cell genes in ACI23 and OV90 ovarian cancer cell lines cultured in TIC-enriching conditions for 5 days relative to cells cultured in adherent conditions. *, P < 0.05, Student t test. B, qRT-PCR for RELA, RELB, and ALDH genes in ALDH-positive cells. Data shown are average fold enrichment and SEM relative to ALDH-negative cells for three independent experiments. *, P < 0.05, Student t test. C, OV90 cells were grown in TIC conditions for 72 hours before addition of doxycycline for an additional 72 hours. RNA was collected and gene expression analyzed by qRT-PCR. Data shown are average fold enrichment and SEM relative to negative shRNA. *, P < 0.05, one-way ANOVA with Dunnett post hoc. D, Whole-cell lysate Western blot analysis showing protein level decrease after 72-hour exposure to doxycycline. The Western blot shown from Fig. 2 was reprobed using ALDH antibodies. E, Left, TIC conditions significantly enhanced binding of RelB to specific NFκB consensus sites in the ALDH1A2 promoter as assessed by ChIP-PCR. X-axis represents distance in base pairs from transcription start site. There is no significant binding enrichment under adherent conditions. Enrichment is defined as the percentage of input. Data shown are average fold enrichment relative to the negative control IgG antibody pull down. Right, TIC conditions did not significantly enrich binding of RelB to any NFκB consensus sites in the ALDH1A1 promoter. *, P < 0.002, Student t test. F, OV90 cells transfected with siRNAs targeting ALDH1A1, ALDH1A2, or a nontargeting negative control sequence were analyzed by flow cytometry for ALDH activity. *, P < 0.05, one-way ANOVA with Dunnett post hoc. G, 2 × 103 OV90 cells transfected with indicated siRNAs were seeded in 96-well plates with 30 μmol/L of carboplatin. After 72 hours, cell death was calculated relative to vehicle control for each siRNA group. *, P < 0.05 one-way ANOVA with Dunnett post hoc. H, A total of 5 × 102 cells transfected with indicated siRNAs were seeded in ultralow attachment 96-well plates with TIC culture media and cultured for 5 days. Three image fields were obtained for each well. *, P < 0.05, one-way ANOVA with Dunnett post hoc test. I, Schematic illustrating the NFκB signaling mechanisms supporting ovarian cancer spheroids. Data shown for all panels represent average and SEM for three independent experiments.

Figure 7.

RelB regulates expression of ALDH1A2. A, qRT-PCR of selected stem cell genes in ACI23 and OV90 ovarian cancer cell lines cultured in TIC-enriching conditions for 5 days relative to cells cultured in adherent conditions. *, P < 0.05, Student t test. B, qRT-PCR for RELA, RELB, and ALDH genes in ALDH-positive cells. Data shown are average fold enrichment and SEM relative to ALDH-negative cells for three independent experiments. *, P < 0.05, Student t test. C, OV90 cells were grown in TIC conditions for 72 hours before addition of doxycycline for an additional 72 hours. RNA was collected and gene expression analyzed by qRT-PCR. Data shown are average fold enrichment and SEM relative to negative shRNA. *, P < 0.05, one-way ANOVA with Dunnett post hoc. D, Whole-cell lysate Western blot analysis showing protein level decrease after 72-hour exposure to doxycycline. The Western blot shown from Fig. 2 was reprobed using ALDH antibodies. E, Left, TIC conditions significantly enhanced binding of RelB to specific NFκB consensus sites in the ALDH1A2 promoter as assessed by ChIP-PCR. X-axis represents distance in base pairs from transcription start site. There is no significant binding enrichment under adherent conditions. Enrichment is defined as the percentage of input. Data shown are average fold enrichment relative to the negative control IgG antibody pull down. Right, TIC conditions did not significantly enrich binding of RelB to any NFκB consensus sites in the ALDH1A1 promoter. *, P < 0.002, Student t test. F, OV90 cells transfected with siRNAs targeting ALDH1A1, ALDH1A2, or a nontargeting negative control sequence were analyzed by flow cytometry for ALDH activity. *, P < 0.05, one-way ANOVA with Dunnett post hoc. G, 2 × 103 OV90 cells transfected with indicated siRNAs were seeded in 96-well plates with 30 μmol/L of carboplatin. After 72 hours, cell death was calculated relative to vehicle control for each siRNA group. *, P < 0.05 one-way ANOVA with Dunnett post hoc. H, A total of 5 × 102 cells transfected with indicated siRNAs were seeded in ultralow attachment 96-well plates with TIC culture media and cultured for 5 days. Three image fields were obtained for each well. *, P < 0.05, one-way ANOVA with Dunnett post hoc test. I, Schematic illustrating the NFκB signaling mechanisms supporting ovarian cancer spheroids. Data shown for all panels represent average and SEM for three independent experiments.

Close modal

We next inquired whether alternative NF-κB signaling through the RelB transcription factor directly binds to DNA in the ALDH1A2 promoter region. Chromatin immunoprecipitation was performed to compare RelA- and RelB-binding enrichment at the ALDH1A1 or ALDH1A2 promoter in TIC cultures relative to adherent cultures. Indeed, almost 3-fold enrichment of RelB binding occurred in a region approximately 2,000 base pairs upstream of the transcription start site of ALDH1A2 that contains NF-κB consensus sequence (Fig. 7E). As expected on the basis of qPCR results, enrichment of RelB at the promoter of ALDH1A1 was not significantly different from adherent cultures (Fig. 7E). No enrichment of RelA was observed at the promoter of either ALDH1A1 or ALDH1A2 (Supplementary Fig. S1). These data support a divergent role for alternative NFκB signaling in the regulation of ALDH1A2 in ovarian cancer TICs.

We confirmed the role of ALDH1A2 in supporting TIC features in ovarian cancer cells. Treatment with siRNA directed against ALDH1A1 or ALDH1A2 resulted in a significant decrease in the percentage of cells with ALDH activity relative to cells treated with nontargeting control siRNA (Fig. 7F). Sensitization to carboplatin was significantly enhanced with knockdown of ALDH1A1 and nearly significantly with knockdown of ALDH1A2 (Fig. 7G). Spheroid formation efficiency was significantly decreased with loss of ALDH1A2, whereas spheroid area was significantly decreased with knockdown of either ALDH1A1 or ALDH1A2 (Fig. 7H).

Ovarian cancer is unique in that it spreads early through the shedding of malignant cells into the abdominal cavity, rather than invasion into blood vessels or lymphatics. In many cases, initial treatment of the disease appears successful, but then cancer recurs months to years after an apparent complete remission. This clinical pattern suggests that chemotherapy resistant cells are present at the onset, and capable of regenerating tumors. TICs are defined as cells that resist chemotherapy and can repopulate tumors after a period of apparent remission. Many investigators have attempted to replicate and study these putative TICs in the laboratory. The early shedding of ovarian cancer suggests that culturing cells in low attachment conditions may more accurately mimic the in vivo conditions for ovarian cancer growth, dispersion and heterogeneity. We previously showed that such conditions resulted in formation of spheroid structures in vitro, and enhanced the ability of ovarian cancer cells to form tumors in mice. In this study, we used low attachment growth conditions to enrich TIC spheroids and examined the functional consequences of classical and alternative NFκB transcription factor signaling in supporting TIC features.

We discovered that both classical and alternative NFκB pathways are important in sustaining TIC spheroids, with distinct roles. Classical NFκB signaling appears to promote proliferation and expansion of the spheroid, which is likely to be crucial for repopulation of the tumor recurrence. Classical NFκB signaling has been shown to support ovarian TICs (identified by expression of cell surface markers CD133 and CD44) through mechanisms, including chemoresistance, epithelial-to-mesenchymal transition, invasion, and migration potential (36–39). A recent study also linked classical NFκB activity to lipid metabolism in cancer spheroids (40). Our data complement and expand on this paradigm by indicating that the alternative NFκB pathway supports self-renewal properties of the spheroid and directly or indirectly regulates genes linked to stem cell features such as ALDH1A2. In this study, we sought to understand the role of NFκB in regulating TICs defined by expression of CD133 and ALDH activity, a combination we and others have used to identify tumorigenic ovarian cancer subpopulations (20, 31, 32, 41). This study further defines NFκB signaling in ovarian cancer by highlighting a novel role for alternative NFκB signaling in supporting tumorigenesis via regulation of ALDH+ chemoresistant subpopulations. Our data also underscore a role for the ALDH1A2 enzyme in supporting a TIC phenotype comparable with the ALDH1A1 enzyme already established in ovarian cancer (40–42). These findings expand on recent studies implicating ALDH1A2 in drug resistance and pathogenesis of ovarian cancer and reveals the diversity of ALDH enzymes in supporting cancer cells (43–46). We show that RelB regulates ALDH1A2 expression that is enhanced in culture conditions that enrich TICs. Moreover, our work suggests that RELA more strongly supports proliferative subpopulations compared to RELB. Most likely, NFκB signaling is nuanced in different subpopulations of cells. This idea is intriguing given the increasing evidence of intratumor heterogeneity and the dynamic relationships between cellular subclones that cooperate to develop malignancies, resist therapy, and repopulate tumors (3).

TICs must be eliminated with initial treatment to prevent relapse of ovarian cancer. NFκB may be a key target in TICs. Our study suggests that both classical and alternative NFκB signaling is present in the heterogeneous spheroid (Fig. 7I). Specific inhibition of downstream effectors of NFκB in spheroids may eliminate TICs. ALDH enzyme activity is a common marker of cancer stem cells in a variety of tissues, and is associated with high tumor-initiating capacity of ovarian cancer cells and poor outcomes for patients with ovarian cancer (32, 47–49). Selective inhibitors of ALDH are in development and may provide a therapeutic benefit for tumor cells dependent on ALDH activity (42, 48, 50), particularly if used in combination with traditional chemotherapy. This study highlights how differential NFκB signaling can maintain diverse phenotypes within the cancer spheroid. Further exploration into the dynamic biology of NFκB in ovarian cancer spheroids will increase our understanding of the complex nature of these tumor-initiating structures, with the hopes of identifying additional biology to influence a long-term cure for women with this disease.

No potential conflicts of interest were disclosed.

Conception and design: C.D. House, C.M. Annunziata

Development of methodology: C.D. House, M. Kim

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C.D. House, L. Hernandez, M. Ozaki, J.M. James, M.J. Kruhlak

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C.D. House, E. Jordan, L. Hernandez, J.M. James, M. Kim, M.J. Kruhlak, E. Batchelor, F. Elloumi, M.C. Cam, C.M. Annunziata

Writing, review, and/or revision of the manuscript: C.D. House, E. Jordan, M.C. Cam, C.M. Annunziata

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C.D. House, E. Jordan, J.M. James

Study supervision: C.M. Annunziata

We thank Karen Wolcott of the CCR FACS core facility, Elena Kuznetsova of the CCR mouse core facility, Brian Hall of Millipore Sigma, and Dr. Tobias Eggert for their excellent technical assistance.

This study was supported by the NCI intramural research program ZIA BC 011054 (to C.M. Annunziata) and 1K99CA204727 (to C.D. House).

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.

1.
Siegel
R
,
Naishadham
D
,
Jemal
A
. 
Cancer statistics, 2013
.
CA Cancer J Clin
2013
;
63
:
11
30
.
2.
Ushijima
K
. 
Treatment for recurrent ovarian cancer-at first relapse
.
J Oncol
2010
;
2010
:
497429
.
3.
Tabassum
DP
,
Polyak
K
. 
Tumorigenesis: it takes a village
.
Nat Rev Cancer
2015
;
15
:
473
83
.
4.
Polyak
K
,
Marusyk
A
. 
Cancer: Clonal cooperation
.
Nature
2014
;
508
:
52
3
.
5.
Blagden
SP
. 
Harnessing pandemonium: the clinical implications of tumor heterogeneity in ovarian cancer
.
Front Oncol
2015
;
5
:
149
.
6.
Hernandez
L
,
Hsu
SC
,
Davidson
B
,
Birrer
MJ
,
Kohn
EC
,
Annunziata
CM
. 
Activation of NF-kappaB signaling by inhibitor of NF-kappaB kinase beta increases aggressiveness of ovarian cancer
.
Cancer Res
2010
;
70
:
4005
14
.
7.
Annunziata
CM
,
Stavnes
HT
,
Kleinberg
L
,
Berner
A
,
Hernandez
LF
,
Birrer
MJ
, et al
Nuclear factor kappaB transcription factors are coexpressed and convey a poor outcome in ovarian cancer
.
Cancer
2010
;
116
:
3276
84
.
8.
Hsu
S
,
Kim
M
,
Hernandez
L
,
Grajales
V
,
Noonan
A
,
Anver
M
, et al
IKK-ϵ coordinates invasion and metastasis of ovarian cancer
.
Cancer Res
2012
;
72
:
5494
504
.
9.
Rajasekhar
VK
,
Studer
L
,
Gerald
W
,
Socci
ND
,
Scher
HI
. 
Tumour-initiating stem-like cells in human prostate cancer exhibit increased NF-κB signalling
.
Nat Commun
2011
;
2
:
162
.
10.
Glinka
Y
,
Mohammed
N
,
Subramaniam
V
,
Jothy
S
,
Prud'homme
GJ
. 
Neuropilin-1 is expressed by breast cancer stem-like cells and is linked to NF-κB activation and tumor sphere formation
.
Biochem Biophys Res Commun
2012
;
425
:
775
80
.
11.
Alvero
AB
,
Chen
R
,
Fu
HH
,
Montagna
M
,
Schwartz
PE
,
Rutherford
T
, et al
Molecular phenotyping of human ovarian cancer stem cells unravels the mechanisms for repair and chemoresistance
.
Cell Cycle
2009
;
8
:
158
66
.
12.
Kendellen
MF
,
Bradford
JW
,
Lawrence
CL
,
Clark
KS
,
Baldwin
AS
. 
Canonical and non-canonical NF-κB signaling promotes breast cancer tumor-initiating cells
.
Oncogene
2014
;
33
:
1297
305
.
13.
Perkins
ND
. 
The diverse and complex roles of NF-κB subunits in cancer
.
Nat Rev Cancer
2012
;
12
:
121
32
.
14.
Gray
CM
,
Remouchamps
C
,
McCorkell
KA
,
Solt
LA
,
Dejardin
E
,
Orange
JS
, et al
Noncanonical NF-κB signaling is limited by classical NF-κB activity
.
Sci Signal
2014
;
7
:
ra13
.
15.
Oeckinghaus
A
,
Hayden
MS
,
Ghosh
S
. 
Crosstalk in NF-κB signaling pathways
.
Nat Immunol
2011
;
12
:
695
708
.
16.
Hayden
MS
,
Ghosh
S
. 
Signaling to NF-kappaB
.
Genes Dev
2004
;
18
:
2195
224
.
17.
Carvalho
BS
,
Irizarry
RA
. 
A framework for oligonucleotide microarray preprocessing
.
Bioinformatics
2010
;
26
:
2363
7
.
18.
Kauffmann
A
,
Gentleman
R
,
Huber
W
. 
arrayQualityMetrics–a bioconductor package for quality assessment of microarray data
.
Bioinformatics
2009
;
25
:
415
6
.
19.
Ritchie
ME
,
Phipson
B
,
Wu
D
,
Hu
Y
,
Law
CW
,
Shi
W
, et al
limma powers differential expression analyses for RNA-sequencing and microarray studies
.
Nucleic Acids Res
2015
;
43
:
e47
.
20.
House
CD
,
Hernandez
L
,
Annunziata
CM
. 
In vitro enrichment of ovarian cancer tumor-initiating cells
.
J Vis Exp
2015
.
doi: 10.3791/52446.
21.
Pahl
HL
. 
Activators and target genes of Rel/NF-kappaB transcription factors
.
Oncogene
1999
;
18
:
6853
66
.
22.
Zhang
J
,
Yamada
O
,
Kida
S
,
Matsushita
Y
,
Yamaoka
S
,
Chagan-Yasutan
H
, et al
Identification of CD44 as a downstream target of noncanonical NF-κB pathway activated by human T-cell leukemia virus type 1-encoded Tax protein
.
Virology
2011
;
413
:
244
52
.
23.
Wang
S
,
Xiao
X
,
Zhou
X
,
Huang
T
,
Du
C
,
Yu
N
, et al
TFPI-2 is a putative tumor suppressor gene frequently inactivated by promoter hypermethylation in nasopharyngeal carcinoma
.
BMC Cancer
2010
;
10
:
617
.
24.
Hernandez
L
,
Kim
MK
,
Lyle
LT
,
Bunch
KP
,
House
CD
,
Ning
F
, et al
Characterization of ovarian cancer cell lines as in vivo models for preclinical studies
.
Gynecol Oncol
2016
;
142
:
332
40
.
25.
Pan
X
,
Arumugam
T
,
Yamamoto
T
,
Levin
PA
,
Ramachandran
V
,
Ji
B
, et al
Nuclear factor-kappaB p65/relA silencing induces apoptosis and increases gemcitabine effectiveness in a subset of pancreatic cancer cells
.
Clin Cancer Res
2008
;
14
:
8143
51
.
26.
Yang
G
,
Xiao
X
,
Rosen
DG
,
Cheng
X
,
Wu
X
,
Chang
B
, et al
The biphasic role of NF-kappaB in progression and chemoresistance of ovarian cancer
.
Clin Cancer Res
2011
;
17
:
2181
94
.
27.
Arora
R
,
Yates
C
,
Gary
BD
,
McClellan
S
,
Tan
M
,
Xi
Y
, et al
Panepoxydone targets NF-kB and FOXM1 to inhibit proliferation, induce apoptosis and reverse epithelial to mesenchymal transition in breast cancer
.
PLoS ONE
2014
;
9
:
e98370
.
28.
Zhang
S
,
Balch
C
,
Chan
MW
,
Lai
HC
,
Matei
D
,
Schilder
JM
, et al
Identification and characterization of ovarian cancer-initiating cells from primary human tumors
.
Cancer Res
2008
;
68
:
4311
20
.
29.
Auffinger
B
,
Tobias
AL
,
Han
Y
,
Lee
G
,
Guo
D
,
Dey
M
, et al
Conversion of differentiated cancer cells into cancer stem-like cells in a glioblastoma model after primary chemotherapy
.
Cell Death Differ
2014
;
21
:
1119
31
.
30.
Kulkarni-Datar
K
,
Orsulic
S
,
Foster
R
,
Rueda
BR
. 
Ovarian tumor initiating cell populations persist following paclitaxel and carboplatin chemotherapy treatment in vivo
.
Cancer Lett
2013
;
339
:
237
46
.
31.
Kryczek
I
,
Liu
S
,
Roh
M
,
Vatan
L
,
Szeliga
W
,
Wei
S
, et al
Expression of aldehyde dehydrogenase and CD133 defines ovarian cancer stem cells
.
Int J Cancer
2012
;
130
:
29
39
.
32.
Silva
IA
,
Bai
S
,
McLean
K
,
Yang
K
,
Griffith
K
,
Thomas
D
, et al
Aldehyde dehydrogenase in combination with CD133 defines angiogenic ovarian cancer stem cells that portend poor patient survival
.
Cancer Res
2011
;
71
:
3991
4001
.
33.
Curley
MD
,
Therrien
VA
,
Cummings
CL
,
Sergent
PA
,
Koulouris
CR
,
Friel
AM
, et al
CD133 expression defines a tumor initiating cell population in primary human ovarian cancer
.
Stem Cells
2009
;
27
:
2875
83
.
34.
Baba
T
,
Convery
PA
,
Matsumura
N
,
Whitaker
RS
,
Kondoh
E
,
Perry
T
, et al
Epigenetic regulation of CD133 and tumorigenicity of CD133+ ovarian cancer cells
.
Oncogene
2009
;
28
:
209
18
.
35.
Wang
Y
,
Cardenas
H
,
Fang
F
,
Condello
S
,
Taverna
P
,
Segar
M
, et al
Epigenetic targeting of ovarian cancer stem cells
.
Cancer Res
2014
;
74
:
4922
36
.
36.
Leizer
AL
,
Alvero
AB
,
Fu
HH
,
Holmberg
JC
,
Cheng
YC
,
Silasi
DA
, et al
Regulation of inflammation by the NF-κB pathway in ovarian cancer stem cells
.
Am J Reprod Immunol
2011
;
65
:
438
47
.
37.
Chefetz
I
,
Alvero
AB
,
Holmberg
JC
,
Lebowitz
N
,
Craveiro
V
,
Yang-Hartwich
Y
, et al
TLR2 enhances ovarian cancer stem cell self-renewal and promotes tumor repair and recurrence
.
Cell Cycle
2013
;
12
:
511
21
.
38.
Xiang
T
,
Long
H
,
He
L
,
Han
X
,
Lin
K
,
Liang
Z
, et al
Interleukin-17 produced by tumor microenvironment promotes self-renewal of CD133+ cancer stem-like cells in ovarian cancer
.
Oncogene
2015
;
34
:
165
76
.
39.
Long
H
,
Xie
R
,
Xiang
T
,
Zhao
Z
,
Lin
S
,
Liang
Z
, et al
Autocrine CCL5 signaling promotes invasion and migration of CD133+ ovarian cancer stem-like cells via NF-κB-mediated MMP-9 upregulation
.
Stem Cells
2012
;
30
:
2309
19
.
40.
Li
J
,
Condello
S
,
Thomes-Pepin
J
,
Ma
X
,
Xia
Y
,
Hurley
TD
, et al
Lipid desaturation is a metabolic marker and therapeutic target of ovarian cancer stem cells
.
Cell Stem Cell
2017
;
20
:
303
314
e5
.
41.
Ishiguro
T
,
Sato
A
,
Ohata
H
,
Ikarashi
Y
,
Takahashi
RU
,
Ochiya
T
, et al
Establishment and characterization of an in vitro model of ovarian cancer stem-like cells with an enhanced proliferative capacity
.
Cancer Res
2016
;
76
:
150
60
.
42.
Yokoyama
Y
,
Zhu
H
,
Lee
JH
,
Kossenkov
AV
,
Wu
SY
,
Wickramasinghe
JM
, et al
BET inhibitors suppress ALDH activity by targeting ALDH1A1 super-enhancer in ovarian cancer
.
Cancer Res
2016
;
76
:
6320
30
.
43.
Liu
X
,
Gao
Y
,
Zhao
B
,
Li
X
,
Lu
Y
,
Zhang
J
, et al
Discovery of microarray-identified genes associated with ovarian cancer progression
.
Int J Oncol
2015
;
46
:
2467
78
.
44.
Shi
C
,
Zhang
Z
. 
Screening of potentially crucial genes and regulatory factors involved in epithelial ovarian cancer using microarray analysis
.
Oncol Lett
2017
;
14
:
725
32
.
45.
Ma
YM
,
Zhao
S
. 
Prognostic values of aldehyde dehydrogenase 1 isoenzymes in ovarian cancer
.
Onco Targets Ther
2016
;
9
:
1981
8
.
46.
Auersperg
N
. 
The stem-cell profile of ovarian surface epithelium is reproduced in the oviductal fimbriae, with increased stem-cell marker density in distal parts of the fimbriae
.
Int J Gynecol Pathol
2013
;
32
:
444
53
.
47.
Liu
S
,
Liu
C
,
Min
X
,
Ji
Y
,
Wang
N
,
Liu
D
, et al
Prognostic value of cancer stem cell marker aldehyde dehydrogenase in ovarian cancer: a meta-analysis
.
PLoS One
2013
;
8
:
e81050
.
48.
Ma
I
,
Allan
AL
. 
The role of human aldehyde dehydrogenase in normal and cancer stem cells
.
Stem Cell Rev
2011
;
7
:
292
306
.
49.
Bai
S
,
Ingram
P
,
Chen
YC
,
Deng
N
,
Pearson
A
,
Niknafs
Y
, et al
EGFL6 regulates the asymmetric division, maintenance, and metastasis of ALDH+ ovarian cancer cells
.
Cancer Res
2016
;
76
:
6396
409
.
50.
Yang
SM
,
Yasgar
A
,
Miller
B
,
Lal-Nag
M
,
Brimacombe
K
,
Hu
X
, et al
Discovery of NCT-501, a potent and selective theophylline-based inhibitor of aldehyde dehydrogenase 1A1 (ALDH1A1)
.
J Med Chem
2015
;
58
:
5967
78
.