Author contributions: C.L.: conception and design, collection and assembly of data, data analysis and interpretation, and manuscript writing; E.V.: conception and design and data analysis and interpretation; L.D.D.: conception and design; C.D.: collection and assembly of data; F.P.: conception and design, data analysis and interpretation, manuscript writing, final approval of manuscript, and financial support.
Disclosure of potential conflicts of interest is found at the end of this article.
First published online in STEM CELLSEXPRESS February 10, 2012.
Breast cancers are thought to be organized hierarchically with a small number of breast cancer stem cells (BCSCs) able to regrow a tumor while their progeny lack this ability. Recently, several groups reported enrichment for BCSCs when breast cancers were subjected to classic anticancer treatment. However, the underlying mechanisms leading to this enrichment are incompletely understood. Using non-BCSCs sorted from patient samples, we found that ionizing radiation reprogrammed differentiated breast cancer cells into induced BCSCs (iBCSCs). iBCSCs showed increased mammosphere formation, increased tumorigenicity, and expressed the same stemness-related genes as BCSCs from nonirradiated samples. Reprogramming occurred in a polyploid subpopulation of cells, coincided with re-expression of the transcription factors Oct4, sex determining region Y-box 2, Nanog, and Klf4, and could be partially prevented by Notch inhibition. We conclude that radiation may induce a BCSC phenotype in differentiated breast cancer cells and that this mechanism contributes to increased BCSC numbers seen after classic anticancer treatment. STEM CELLS2012;30:833–844
Recent clinical and preclinical data support the view that many solid cancers, including breast cancers, are organized hierarchically with a small number of cancer stem cells (CSCs) able to regrow a tumor while their progeny lack this feature [1, 2]. Clinically, CSCs have been associated with higher rates of recurrence and metastasis [3, 4]. Importantly, CSCs in breast cancer and glioma have been found to be relatively resistant to radiation and chemotherapy compared with their nontumorigenic progeny [5–7]. Consistent with these reports, several groups reported enrichment for CSCs when solid cancers were subjected to classic anticancer treatments [5, 6, 8].
Using an in vitro system, we quantified the number of breast CSCs (BCSCs) surviving after radiation treatment in patient samples as well as in several breast cancer lines. When we compared the absolute number of BCSCs that survived radiation treatment to the number of BCSCs expected to survive, we found a profound enrichment in BCSCs after exposure to ionizing radiation, and such a drastic increase in numbers could not easily be explained by differences in radiation sensitivity and/or by active repopulation. Here, we report that ionizing radiation induced a BCSC phenotype in previously nontumorigenic cells. This transition was Notch dependent and coincided with upregulation of transcription factors used to generate induced pluripotent cells from differentiated normal cells.
Human SUM159PT breast cancer cell lines were purchased from Asterand (Asterand, Inc., MI, http://asterand.com). Human MCF-7 and T47D breast cancer cell lines were purchased from American Type Culture Collection (Manassas, VA, ATCC.org). SUM159PT-ZsGreen-cODC, MCF-7-ZsGreen-cODC, and T47D-ZsGreen-cODC were obtained as described in Vlashi et al. . SUM159PT cells were cultured in log-growth phase in F12 Medium (Invitrogen, Carlsbad, CA, http://www.invitrogen.com) supplemented with 5% fetal bovine serum (Sigma-Aldrich, St. Louis, MO, http://www.sigmaaldrich.com), penicillin (100 units/ml) and streptomycin (100 μg/ml) (both Invitrogen), and insulin (5 μg/ml) and hydrocortisone (1 μg/ml). MCF-7 and T47D cells were cultured in log-growth phase in Dulbecco's modified Eagle medium (DMEM) (Invitrogen) supplemented with 10% fetal bovine serum, penicillin, and streptomycin. All cells were grown in a humidified incubator at 37°C with 5% CO2.
Cells grown as monolayers were irradiated at room temperature using an experimental x-ray irradiator (Gulmay Medical Inc., Atlanta, GA, xstrahl.com) at a dose rate of 2.789 Gy/minute for the time required to apply a prescribed dose. Corresponding controls were sham irradiated. Assessment of cell proliferation, the number of BCSCs, and sphere-forming assays were performed 5 days after radiation.
BCSCs were identified based on their low proteasome activity [9, 10] using the ZsGreen-cODC reporter system. Five days after radiation, cells were trypsinized and ZsGreen-cODC expression was assessed by flow cytometry (MACSQuant Analyzer, Miltenyi, www.miltenyibiotec.com). Cells were defined as “ZsGreen-cODC positive” if the fluorescence in the FL-1H channel exceeded the fluorescence level of 99.9% of the empty vector-transfected control cells.
Aldefluor Assay and Separation of the ALDH1-Negative Population by Fluorescence-Activated Cell Sorting
Ginestier et al. previously reported that BCSCs could be isolated based on their high ALDH1 activity . The ALDEFLUOR kit (StemCell Technologies, Durham, NC, http://www.stemcell.com/) was used to isolate the population with no ALDH1 enzymatic activity. Cells obtained from breast cancer monolayer (SUM159PT and T47D) were suspended in ALDEFLUOR assay buffer containing ALDH1 substrate (BAAA, 1 μmol/l per 1 × 106 cells) and incubated for 40 minutes at 37°C. As negative control for each sample of cells, an aliquot was treated with 50 mmol/l diethylaminobenzaldehyde (DEAB), a specific ALDH1 inhibitor. The sorting gates were established using ALDEFLUOR-stained cells treated with DEAB as negative controls.
CD24/CD44 Staining and Separation of the CD24+/high/CD44− Population by Fluorescence-Activated Cell Sorting
MCF-7 and T47D cells growing as monolayer cultures were stained for CD24 and CD44 expression as described previously . Briefly, cells were incubated with trypsin–EDTA, dissociated, and passed through a 40 μm sieve. Cells were pelleted by centrifugation at 500g for 5 minutes at 4°C, resuspended in 100 μl of monoclonal mouse anti-human CD24–fluorescein isothiocyanate (FITC) antibody (BD Pharmingen, San Jose, CA, http://www.bdbiosciences.com) and a monoclonal mouse anti-human CD44–phytoerythrin (PE) antibody (BD Pharmingen), and incubated for 20 minutes at 4°C. The sorting gates were established using cells stained with isotype controls (isotype control FITC-conjugated antibodies [BD pharmingen] and isotype control PE-conjugated antibodies [BD pharmingen], respectively).
After irradiation, cells were trypsinized and plated in mammosphere media (DMEM-F12, 0.4% BSA [Sigma], 10 ml/500 ml B27 [Invitrogen], 5 μg/ml bovine insulin [Sigma], 4 μg/ml heparin [Sigma], 20 ng/ml basic fibroblast growth factor 2 [bFGF-2, Sigma], and 20 ng/ml epidermal growth factor [EGF, Sigma]) into 96-well ultra-low adhesion plates, ranging from 1 to 256 cells per well. Growth factors, EGF and bFGF, were added every 3 days, and the cells were allowed to form spheres for 20 days. The number of spheres formed per well was then counted and expressed as a percentage of the initial number of cells plated. Cells were also plated in mammosphere media into 100 mm suspension dishes at 10,000 cells per microliter and allowed to form spheres for 15 days, these cells were used for secondary sphere-forming experiments. Three independent experiments were performed.
Quantitative Reverse Transcription-PCR
Total RNA was isolated using TRIZOL Reagent (Invitrogen). cDNA synthesis was carried out using the SuperScript Reverse Transcription III (Invitrogen). Quantitative PCR was performed in the My iQ thermal cycler (Bio-Rad, Hercules, CA) using the ×2 iQ SYBR Green Supermix (Bio-Rad, http://www.bio-rad.com/). Ct for each gene was determined after normalization to GAPDH or RPLP0, and ΔΔCt was calculated relative to the designated reference sample. Gene expression values were then set equal to 2, as described by the manufacturer of the kit (Applied Biosystems, Carlsbad, CA, http://www.appliedbiosystems.com). All PCR primers were synthesized by Invitrogen and designed for the human sequences of Oct4, sex determining region Y (SRY)-box 2 (Sox2), Nanog, Klf4, and c-Myc. Primers for the customized stem cell gene expression array were synthesized by Real Time Primers LLC (Elkins Park, PA, http://www.realtimeprimers.com).
Two Channel Flow Cytometry for OCT4/Sox2/Nanog/Klf4/c-Myc and DNA Content
Cells were harvested at relevant time points, washed in cold TBS, and fixed overnight in cold (−20°C) 70% ethanol. After two washes in TBS, cells were permeabilized with TBS/4% BSA/0.1% Triton X-100 for 10 minutes at RT. Samples were incubated with a rabbit polyclonal anti-Oct4 antibody (Cell Signaling, Danvers, MA, http://www.cellsignal.com/), a monoclonal mouse anti-Sox2 (R&D Systems, Minneapolis, MN, http://www.rndsystems.com/), a rabbit polyclonal anti-Nanog (Abcam, Cambridge, MA, http://www.abcam.com/), a monoclonal mouse anti-Klf4 (Abgen), monoclonal mouse anti-c-Myc (Abcam), or corresponding isotype control (Biolegend, San Diego, CA, http://www.biolegend.com) in TBS/4% BSA/0.1% Triton X-100 for 1 hour at RT. Following three washes in TBS, cells were incubated with goat anti-rabbit Alexa Fluor 750-APC (Invitrogen) or goat anti-mouse PE-Cy7 (Sigma) antibodies in TBS/4% BSA/0.1% Triton X-100, 1:200 for 1 hour in the dark. DNA was counterstained with 10 μg/ml propidium iodide solution in PBS, containing 200 μg/ml RNase (Sigma), assessed by flow cytometry using a MACSQuant Analyzer (Miltenyi Biotec), and analyzed using the FlowJo Software (version 9.3.1).
Nude (nu/nu), 6–8-week-old female mice, originally obtained from The Jackson Laboratories (Bar Harbor, ME, jax.org) were rederived, bred, and maintained in a pathogen-free environment in the American Association of Laboratory Animal Care-accredited Animal Facilities of Department of Radiation Oncology, UCLA (Los Angeles, CA) in accordance to all local and national guidelines for the care of animals.
SUM159PT-ZsGreen-cODC-negative cells derived from monolayer cultures and sorted by fluorescence-activated cell sorting (FACS) were plated in F12 media supplemented with 5% fetal bovine serum, penicillin (100 units/ml), streptomycin (100 μg/ml) cocktail, insulin (5 μg/ml), and hydrocortisone (1 μg/ml). The following day, cells were irradiated with 0, 4, or 8 Gy. Five days after irradiation, cells were injected subcutaneously into the thighs and shoulders of 6-week-old female Nu/Nu mice (106, 105, 104, 103, or 102 cells per inoculum, n = 8) within Matrigel (BD Biosciences). Tumor growth was assessed on a weekly basis, and the mice were sacrificed when the tumor size reached tumor diameters requiring euthanasia. The experiment was terminated after 13 weeks. Data were fitted using a sigmoidal regression model (Y = a × X∧b/(c + X∧b), Graphpad Prism 5.0). The number of cells needed to obtain tumors in 50% of the animals (TD50) was calculated for each radiation dose.
Human Breast Cancer Primary Specimens
Primary tumor specimens were obtained under a protocol approved by the UCLA, Los Angeles Institutional Review Boards through the Translational Pathology Core Laboratory at UCLA (IRB# 02-02-057-22).
The tumor specimens were digested and cells were expanded ex vivo for 2–3 weeks. ALDH-negative cells were isolated using FACS.
Notch1-4 and Sox2/Nanog siRNA Transfection
SUM159PT cells were subjected to transfection with Notch1, Notch2, Notch3, Notch4, Sox2, or Nanog-specific siRNA (Sigma-Aldrich). MISSION siRNA universal negative control (Sigma-Aldrich) was used as a transfection control. Briefly, siRNA (100 ng total) and lipofectamine (Invitrogen) were diluted in OptiMEM I reduced serum media (Invitrogen), mixed and incubated for 20 minutes as described by the manufacturer. Cells were rinsed with PBS ×1, twice, and incubated in 750 μl of OptiMEM I. siRNA/lipofectamine mix was added on the top of the cells and incubated at 37°C, 5% CO2 for 6 hours. After incubation, media were removed and 2 ml of serum containing F12 medium was added. Twenty-four hours after transfection, cells were plated for a sphere-forming capacity assay, or cells were sorted and treated as previously described.
Noscapine Treatment and Cell Cycle Analysis
In order to identify the contribution of polyploidy to CSCs generation, nontumorigenic MCF-7, T47D, SUM159PT, and two patient samples (2 and 3) were treated with Noscapine . Nontumorigenic cells (CD24+/high/CD44−, ALDH1-negative, or ZsGreen-cODC-negative) were sorted and plated as monolayer cultures. Each following day, MCF-7 and T47D cells were treated with noscapine (0, 10, or 25 μM). SUM159PT cells and patient samples 1 and 2 were treated with noscapine at 0, 25, or 50 μM (Sigma-Aldrich). At day 5, the presence of CD24−/low/CD44high, ALDH1-positive, or ZsGreen-cODC-positive cells was analyzed by flow cytometry. In parallel, cells were analyzed for polyploidy as described above.
All results are expressed as mean values. A p value of ≤.05 in a paired two-sided Student's t test was considered to indicate statistically significant differences. The test was applied to normalized data to compensate for the variance of measurements between biologically independent replicates of the same experiments.
Radiation Induces a BCSCs Phenotype in Previously Nontumorigenic Cells
Consistent with a radioresistant phenotype for BCSCs, we and others previously reported that ionizing radiation increased the number of BCSCs in the overall breast cancer cell population [5, 6, 10]. This has been ascribed to selective killing of nontumorigenic cells and/or a switch from an asymmetric to symmetric type of cell division of BCSCs that gives rise to two identical daughter BCSCs and thus, leading to a relative and absolute increase in BCSCs.
In this study, we used single-cell suspensions from fresh human breast specimen and stained them for ALDH1 activity, a recently described marker for BCSCs [2, 3]. Using FACS, we isolated non-BCSCs (ALDH1-negative cells) from these specimens after purging BCSCs (ALDH1-positive cells). Purified non-BCSCs (ALDH1-negative cells) were then plated and irradiated the following day with 0, 4, or 8 Gy. Five days after irradiation, we assessed the number of BCSCs arising within the non-BCSC population. We found that radiation led to a dose-dependent increase in the number of ALDH1-positive cells (Fig. 1A; patient sample 1: 0 Gy, 0.92%, 4 Gy, 1.92%, 8 Gy, 4.05%; patient sample 2: 0 Gy, 1.46%, 4 Gy, 3.56%, 8 Gy, 8.59%; patient sample 3: 0 Gy, 1.24%, 4 Gy, 2.9%, 8 Gy, 3.45%).
From previous experience with FACS-purified cell population, we were aware of the fact that in practice these purified cell population are not 100% pure . Indeed, a closer look at the FACS-purified population of ALDH1-negative cells revealed that they always contained a very small population of contaminating ALDH1-positive cells. However, calculations based on the number of contaminating cells, the long doubling time of BCSCs (∼72 hours), which is not affected by irradiation , and the surviving fraction of BCSCs for each dose point (Supporting Information Fig. S1), suggested that contaminating BCSCs present at the time of irradiation were unlikely to be the sole source of the absolute increase in BCSC numbers seen 5 days after irradiation. We therefore chose to test an alternative explanation, namely that nontumorigenic breast cancer cells acquire a BCSC phenotype in response to ionizing radiation, thus contributing to the enrichment in BCSCs seen after radiation treatment.
To test this alternative hypothesis, we used a panel of three established and widely used breast cancer cell lines (SUM159PT, MCF-7, and T47D). This panel allowed for the necessary number of experimental repeats in a large number of different assays. In addition to the Aldefluor test (Fig. 1B), we used CD24 and CD44 to identify BCSCs and made use of a previously described imaging system for CSCs in breast cancer and glioma. This latter system is based on a fusion protein between the green fluorescent protein ZsGreen and the C-terminal degron of murine ornithine decarboxylase (cODC) and reports for 26S proteasome activity (Fig. 1D, 1E), a protease activity that was found to be very low in CSCs . Cells with low proteasome activity accumulate the fluorescent fusion protein while in cells with high proteasome activity the cODC portion of the fusion protein directs ZsGreen to ubiquitin-independent degradation by the 26S proteasome. Importantly, in breast cancer, cells with intrinsically low proteasome activity overlapped with the CD24−/low/CD44high cell population  and with cells positive for ALDH1 (Supporting Information Fig. S2A/2B).
Next, we confirmed that radiation-induced increases in the number of BCSCs did not only occur in non-BCSCs from patient samples but also occur in non-BCSCs from established cells lines. Therefore, sorted ALDH1-negative cells from SUM159PT (Fig. 1B) and T47D (Supporting Information Fig. S3A) cell lines were plated as monolayer cultures and irradiated the following day with 0, 4, or 8 Gy. Like in patient-derived samples, we observed a significant dose-dependent increase in the number of ALDH1-positive cells, 5 days after irradiation (Fig. 1B, SUM159PT: 0 Gy: 0.3%; 4 Gy, 1.76%, p = .008; 8 Gy, 5.74%, p = .002; two-sided Student's t test). To confirm this observation, we used the CD24high/CD44low marker combination to isolate nontumorigenic MCF-7 and T47D breast cancer cells. Again, 5 days after irradiation, we observed the induction of CD24−/low/CD44high from previously nontumorigenic CD24high/CD44low cells (MCF-7: 0 Gy: 0.37%: 4 Gy: 2.35%, p = .016; 8 Gy: 6.33%, p = .221. T47D: 0 Gy: 0.08%: 4 Gy: 0.38%, p = .025; 8 Gy: 1.38%, p = .006). In parallel experiments, we used SUM159PT, MCF-7, and T47D cells transfected with the ZsGreen-cODC reporter construct for proteasome activity. Using this third marker for BCSCs, we also observed a dose-dependent, radiation-induced increase of ZsGreen-cODC-positive (low proteasome activity) cell numbers (Figs. 1D/1E, 2; Supporting Information Tables S1 and S2). Importantly, cells positive for BCSC markers could not only be generated from differentiated cells in monolayer cultures but also be generated from differentiating cells sorted from mammospheres (Fig. 2, Supporting Information Fig. S3B).
Previous studies reported a role for the Notch signaling pathway in maintaining a stem cell phenotype in mammary epithelial cells . We decided to repeat the above radiation experiments in the presence of a γ-secretase inhibitor to block Notch signaling. Inhibition of Notch signaling attenuated the generation of cells positive for the BCSC marker (ZsGreen-cODC-positive) but did not completely abrogate the ability of nontumorigenic cells (ZsGreen-cODC-negative) to give rise to cells with intrinsically low proteasome activity (Fig. 2A--2C; Supporting Information Tables S1 and S2). To further test the involvement of the Notch pathway in the induction of BCSCs, we inhibited the expression of Notch receptors by transfecting cells with targeting siRNA. SUM159PT-ZsGreen-cODC cells were transfected with specific siRNA targeting the Notch1, Notch2, Notch3, or Notch4 receptors using lipofectamine. Twenty-four hours after transfection, SUM159PT-ZsGreen-cODC-negative cells were isolated and plated as monolayer cultures. The following day, cells were irradiated with 0, 4, or 8 Gy. Five days after irradiation, we observed a decrease in the number of BCSCs generated if compared with cells transfected with a scrambled siRNA control. Downregulation of Notch2, Notch3, and Notch4 did not prevent induction of BCSCs (Fig. 2D, 8 Gy: siCTL: 3.04%, siNotch1: 1.54%, p = .013). This indicated that signaling through the Notch1 receptor was either involved in inducing BCSCs or in maintaining the stem cell phenotype of newly generated BCSCs.
In order to study whether nontumorigenic cells could also generate BCSCs in the presence of pre-existing BCSCs, we reconstituted a mixed population of nontumorigenic cells and StrawberryRed-labeled BCSCs at different dilutions (0, 2, and 10% BCSCs). After sorting, SUM159PT-ZsGreen-cODC-negative nontumorigenic cells and SUM159PT-ZsGreen-cODC-positive/StrawberryRed-expressing BCSCs were mixed and plated as monolayer cultures. The following day, cells were irradiated with 0, 4, or 8 Gy. When we analyzed the cells 5 days after irradiation by flow cytometry, we found that radiation-induced generation of BCSCs (ZsGreen-cODC-positive/StrawberryRed-negative) was decreased in the presence of pre-existing BCSCs (Fig. 2E) suggesting a negative feedback loop of existing BCSC in the generation of induced BCSCs (iBCSCs).
Using operational means, we next sought to test whether the occurrence of ALDH1-positive cells and ZsGreen-cODC-positive cells with low proteasome activity reflected only an induction of BCSC marker expression or whether it was truly the induction of a BCSC phenotype. In order to assess the self-renewal capacity of iBCSC, we used a sphere-forming assay in which cells depleted from BCSCs with low proteasome activity were seeded at clonal densities into ultra-low adhesion plates in the absence of fetal calf serum to allow formation of mammospheres from single cells. In breast cancer, mammosphere formation is a measure of in vitro BCSC self-renewal capacity and correlates closely with tumorigenicity . In order to test whether radiation induced a BCSC phenotype on the nontumorigenic cells, we compared the sphere-forming capacity of irradiated samples and the nonirradiated control. Furthermore, the number of spheres formed for each radiation dose was compared with the hypothetical number of mammospheres expected. The expected number of mammospheres at each dose was calculated based on (a) the number of contaminating ZsGreen-cODC-positive cells after sorting, (b) a doubling time of 72 hours during the 5 days of culture, and (c) on their clonogenic surviving fraction at each dose point (Supporting Information Fig. S1).
In primary spheres, the self-renewal capacity of irradiated cells remained either the same or exceeded that of corresponding nonirradiated control cells. With the exception of T47D, this effect was also observed in secondary mammospheres. However, when the observed number of mammospheres was compared with the expected number of mammospheres at each dose point, mammosphere formation of irradiated cells significantly exceeded the number of mammospheres expected to be formed (Fig. 3A and Supporting Information Fig. 3B).
Next, we sorted SUM159PT cells based on ZsGreen-cODC expression and irradiated non-BCSCs (ZsGreen-cODC-negative) with 0, 4, and 8 Gy. After 5 days, cells were injected into the thighs and shoulders of 6-week-old female Nu/Nu mice in a limiting dilution assay (106, 105, 104, 103, or 102 cells per inoculum, n = 8 per injection). Thirteen weeks after injection, the number of cells required to initiate tumor growth in 50% of the animals (TD50) was calculated. As expected, TD50 values of nonirradiated non-BCSCs were high (1.15 × 105 cells) consistent with a small number of contaminating BCSCs after FACS. However, TD50 values were reduced 32-fold after a single dose of 4 Gy (3.6 × 103) and ninefold (1.26 × 104) after a single dose of 8 Gy suggesting a radiation-induced relative and absolute increase in BCSC numbers (Fig. 3B and Supporting Information Table S3). Taken together, these data indicated that ionizing irradiation not only induced expression of BCSC markers in nontumorigenic BCSCs but also led to the acquisition of CSC traits.
Finally, we decided to compare the expression profile of 86 genes associated with stem cell traits between nonirradiated ZsGreen-cODC-negative cells, nonirradiated ZsGreen-cODC-positive BCSCs, and ZsGreen-cODC-positive iBCSCs induced by 8 Gy. Ten genes were consistently and significantly upregulated in ZsGreen-cODC-positive BCSCs and iBCSCs generated by 8 Gy (Fig. 4A and Supporting Information Fig. S4). Genes included key elements of the Notch, Wnt, Shh, and FGF signaling pathways as well as genes involved in cell cycle regulation, cell adhesion, and cell-to-cell contact. The comparable gene expression profiles of ZsGreen-cODC-positive iBCSCs after a single radiation dose of 8 Gy given to FACS-sorted ZsGreen-cODC-negative non-BCSC and pre-existing ZsGreen-cODC-positive BCSCs suggested that iBCSCs are driven by the same set of stem cell-related genes. However, some differences in that expression of SRY-box 1 (Sox1), Sox2, S100 calcium binding protein B (S100B), Par-6 partitioning defective six homolog alpha (PARP6A), Deltex homolog 1, and delta-like 1 were significantly higher in iBCSCs, while expression of frizzled homolog 1 and collagen type II alpha 1 were significantly lower (Fig. 6B). Overall, the expression profile of stemness-related genes in iBCSCs reassembled the expression profile found in BCSCs much closer than the expression profile found in the nonirradiated non-BCSC population they originated from.
Radiation Induces Re-Expression of Oct4, Sox2, Nanog, and Klf4
Acquisition of a stem cell phenotype has been described for nonmalignant-differentiated cells after overexpression of Oct4, Sox2, Nanog, Klf4, and c-Myc. These transcription factors are now routinely used to generate induced pluripotent stem (iPS) cells from differentiated somatic cells [14, 15] and have also been shown to maintain the BCSCs phenotype . Interestingly, Oct4, Sox2, Nanog, and Klf4 are known substrates of the 26S proteasome [17–20], and therefore expected to be stabilized in cells with low proteasome activity. To determine whether these transcription factors were reactivated in non-BCSCs population after irradiation, we analyzed their expression levels 5 days after 0, 4, or 8 Gy of radiation, the time point at which we observed absolute increases in BCSC numbers. As expected, we found a significant radiation dose-dependent increase in Oct4, Sox2, Nanog, and Klf4 mRNA expression levels that matched the expression levels for these transcription factors in intrinsically occurring, nonirradiated BCSCs (Fig. 5).
To test whether re-expression of the above four transcription factors occurred randomly or in a specific subset of cells, we analyzed the expression of Oct4, Sox2, Nanog, and Klf4 protein levels and correlated it to the DNA content of the cells. In irradiated cells, we observed increased number of polyploid cells (SUM159PT: 0 Gy: 0.74%, 4 Gy: 4.63%, p < .0001; 8 Gy: 6.33%, p < .0001; patient sample 2: 0 Gy: 12.6%, 4 Gy: 14.4%, p < .032; 8 Gy: 22.5%, p < .0003; patient sample 3: 0 Gy: 6.41%, 4 Gy: 12.6%, p < .012; 8 Gy: 15.4%, p < .0004; MCF-7: 0 Gy: 3.60%, 4 Gy: 6.96%, p < .004; 8 Gy: 24.7%, p < .0006; T47D: 0 Gy: 12.0%, 4 Gy: 17.6%, p < .007; 8 Gy: 47.23%, p < .003; Fig. 6A and Supporting Information Fig. S5A/S5B) in which Oct4, Sox2, Nanog, and Klf4 proteins were upregulated (SUM159PT: Oct4: 0 Gy, 2.71%, 4 Gy: 17.02%, p = .041; 8 Gy: 20.1%, p = .003; Sox2: 0 Gy, 4.52%, 4 Gy: 14.5%, p = .0002; 8 Gy: 48.98%, p < .0001; Nanog: 0 Gy, 0.39%, 4 Gy: 18.9%, p = .223; 8 Gy: 27.58%, p = .222: Klf4: 0 Gy, 1.89%, 4 Gy: 4.74%, p = .031; 8 Gy: 9.86%, p = .084; Fig. 6B/6C and Supporting Information Fig. 5C/5D) and which were also highly enriched for ZsGreen-cODC-positive cells with low proteasome activity (total population: 0 Gy, 0.64%, 4 Gy: 1.84%, p < .0001; 8 Gy: 9.09%, p < .0001; polyploid: 0 Gy, 0.08%, 4 Gy: 0.94%, p = .01; 8 Gy: 2.69%, p = .005; polyploid/Oct4+: 0 Gy, 0%, 4 Gy: 3.97%, p < .0001; 8 Gy: 12.14%, p = .031; polyploid/Sox2+: 0 Gy, 0%, 4 Gy: 3.84%, p = .041; 8 Gy: 10.19%, p = .042; polyploid/Nanog+: 0 Gy, 0.45%, 4 Gy: 1.14%, p = .107; 8 Gy: 2.79%, p = .029; Fig. 6D and Supporting Information Fig. 5B). c-Myc levels were not increased (Figs. 5D, 6B/6D). Consistent with a previous report , inhibition of Notch activation attenuated the induction of polyploidy (4 Gy: 4.87%; 4 Gy + γ-secretase inhibitor: 2.72, p = .03; 8 Gy: 6.32%; 8 Gy + γ-secretase inhibitor: 3.42, p = .028; Supporting Information Fig. S5B) and the induction of Oct4, Sox2, Nanog, and Klf4 expressing polyploid cells (Supporting Information Fig. S5D). This again supported involvement of Notch signaling in radiation reprogramming.
In order to identify the role of these transcription factors in the generation of iBCSCs, we used siRNA to target the expression of Sox2 and Nanog. Nontumorigenic cells transfected with siRNA-Sox2 and siRNA-Nanog were sorted, plated, and irradiated with 0, 4, and 8 Gy the following day. Downregulation of either Sox2 or Nanog alone had no effect on the induction of iBCSCs. However, generation of iBCSCs was significantly reduced when Sox2 and Nanog were downregulated simultaneously (0 Gy: siRNA control: 0.37%, siRNA-Sox2/Nanog: 0.57%, p = .505; 4 Gy: siRNA control: 1.36%, siRNA-Sox2/Nanog: 0.49%, p = .008; 8 Gy: siRNA control: 3.04%, siRNA-Sox2/Nanog: 1.63%, p = .047; Fig. 6E).
Induction of Polyploidy-Induced Generation of iBCSCs
In differentiated cells, Sox2 and Oct4 are epigenetically silenced. The suppression of gene expression for both transcription factors is incomplete but still sufficient to maintain protein levels below a critical threshold. We hypothesized that in polyploid cells found after irradiation, multiple copies of partially silenced Oct4 and Sox2 genes might be sufficient to drive gene expression beyond this threshold, thereby inducing a BCSC phenotype. To test this hypothesis, we explored whether the effect of radiation could be mimicked by pharmacological induction of polyploidy. Treatment of breast cancer cell lines and patient samples with noscapine caused a substantial increase in the number of polyploid cells on day 5 after addition of the drug. The observed increase was comparable to the amount of polyploidy found after irradiation (Fig. 7A and Supporting Information Fig. S7A/S7B). Next we tested whether the number of BCSCs was also increased. In both patient samples, noscapine treatment increased the number of ALDH1-positive cells significantly (Fig. 7B). Increased number of BCSCs with low proteasome activity was also found when MCF-7, T47D, or SUM159PT ZsGreen-cODC expressing cells were analyzed for ZsGreen-cODC-positive cells (Fig. 7C and Supporting Information Fig. S7C) and when MCF-7 and T47D cells were analyzed for the number of CD24low/−/CD44high cells (Fig. 7D). These results suggested that increased gene dose in polyploid cells could indeed be one of the mechanisms leading to acquisition of a CSC phenotype in breast cancer in response to radiation treatment.
Finally, we confirmed that BCSCs rely on Notch signaling using specific siRNAs targeting the Notch1, Notch2, Notch3, or Notch4 receptor. Cells were transfected with lipofectamine, 24 hours after transfection cells were plated in sphere media for a sphere-forming capacity assay. As expected, downregulation of Notch receptor expression reduced the ability of the cells to form mammospheres (Fig. 7E). Furthermore, in order to demonstrate that BCSCs rely on the transcription factors induced by radiation, we downregulated Sox2 and Nanog using specific siRNAs. Cells were transfected with lipofectamine, 24 hours after transfection cells were plated and tested for sphere formation. As expected, downregulation of Sox2 and Nanog expression reduced self-renewal capacity of cells from established breast cancer cell lines and primary breast cancer samples, and Sox2, which acts upstream of Nanog and has targets in addition to Nanog, was more efficient in Figure 7F.
Treatment gaps in radiation therapy have long been known to worsen the outcome for patients suffering from epithelial cancers including cancers of the head and neck region and the breast [22, 23]. The underlying mechanisms are incompletely understood but in general are attributed to accelerated repopulation, a phenomenon that refers to the increased growth rates of cancers during treatment gaps that far exceed their initial growth rates. It is thought that during accelerated repopulation, CSCs switch from an asymmetric type of cell division, which leads to one daughter CSC and one differentiating cell, to a symmetric type of cell division, which yields two identical daughter CSCs. Our data suggest that in addition to the classic view of accelerated repopulation in which CSCs switch from an asymmetric type of cell division that leads to one daughter CSC and one differentiating cells to a symmetric type of cell division that yields in two daughter CSCs, differentiated cancer cells may also be able to acquire stem cell traits under certain conditions of tumor microenvironmental stress, including stress induced by ionizing radiation. Acquisition of stem cell traits by CD133-negative, nontumorigenic glioma cells was previously reported under hypoxic conditions  and in response to nitric oxide-induced notch signaling , suggesting that CSC plasticity may be a common response to multiple stimuli including cancer therapies.
Our observation that ionizing radiation reactivated the same transcription factors in differentiated breast cancer cells that reprogram differentiated somatic cells into iPS cells is provoking. However, it is in line with recent reports that baseline levels of Sox2, Oct4, and Nanog expression can be detected in breast cancers [26, 27] and that ectopic overexpression of Oct4 in normal mammary epithelial cells induces a BCSCs phenotype . Our data further indicate that an increase number of gene copies of Oct4 and Sox2 in polyploid cells could be one possible mechanism behind radiation-induced reprogramming. This was supported by our data showing that downregulation of Sox2 prevented mammosphere formation. Downregulation of the Sox2 downstream target Nanog was less efficient, indicating that multiple genes downstream of Sox2 contribute to the acquisition of a CSCs phenotype. Similar observations have been reported for lymphoma cells . Furthermore, this is in accordance with previous reports on Notch-dependent induction of polyploidy  and our data showing that inhibition of Notch signaling partially prevented the occurrence of iBCSCs (Supporting Information Fig. S5B), suggesting that targeting Notch signaling might enhance local control after radiation therapy.
The CSC hypothesis was formulated more than a century ago . However, until recently prospective identification of CSCs was impossible. The discovery of marker combinations that identify CSCs has resulted in novel insights into the biology of cancer. Still, the CSC hypothesis has been challenged and some experimental data support a model of clonal evolution as an alternative organizational structure of tumors  in which every cancer cell may acquire stem cell traits at some point.
Our study unites the competing models of clonal evolution and hierarchical organization of cancers , as it suggests that undisturbed growing tumors indeed maintain a low number of CSCs. However, if challenged by various stressors including ionizing radiation, iCSCs are generated, which may together with the surviving CSCs repopulate a tumor. These findings have implications for the design of novel treatment protocols that target CSCs, including radiation therapy. The curability of a cancer may not only be dependent on the intrinsic radiosensitivity of CSCs but also on the radiosensitivity of induced CSCs and the rate at which they are generated. Controlling the radio resistance of BCSCs and the generation of new iBCSCs during radiation treatment may ultimately improve curability and may allow for de-escalation of the total radiation doses currently given to breast cancer patients thereby reducing acute and long-term adverse effects.
CONCLUSIONS AND SUMMARY
In summary, our study shows that ionizing radiation reactivated the expression of Oct4 and Sox2 and induced a CSC phenotype in previously nontumorigenic breast cancer cells. The phenomenon was dependent on the induction of polyploidy and Notch signaling. We conclude that a detailed understanding of the underlying pathways could lead to novel combination therapies that will potentially enhance the efficacy of radiation treatment.
F.P. was supported by grants from the California Breast Cancer Research Program (15NB-0153), the Department of Defense (W81XWH-07-1-0065), and the National Cancer Institute (5RO1CA137110). C.L. was supported by an award from the Ward Family Foundation.