Role of CD133/NRF2 Axis in the Development of Colon Cancer Stem Cell-Like Properties

Cancer stem cells (CSCs) exhibit intrinsic therapy/stress resistance, which often cause cancer recurrence after therapy. In this study, we investigated the potential relationship between the cluster of differentiation (CD)-133, a CSC marker of colon cancer, and nuclear factor erythroid 2-like 2 (NFE2L2; NRF2), a master transcription factor for the regulation of multiple antioxidant genes. In the first model of CSC, a sphere culture of the colorectal cell line HCT116, showed increased levels of CD133 and NRF2. Silencing of CD133 reduced the levels of CSC markers, such as Kruppel-like factor 4 (KLF4) and ATP-binding cassette subfamily G member 2 (ABCG2), and further suppressed the expression levels of NRF2 and its target genes. As a potential molecular link, CD133-mediated activation of phosphoinositide 3-kinase/serine-threonine kinase (PI3K/AKT) signaling appears to increase the NRF2 protein levels via phosphorylation and the consequent inhibition of glycogen synthase kinase (GSK)-3β. Additionally, NRF2-silenced HCT116 cells showed attenuated sphere formation capacity and reduced CSC markers expression, indicating the critical role of the NRF2 pathway in the development of CSC-like properties. As a second model of CSC, the CD133high cell population was isolated from HCT116 cells. CSC-like properties, including sphere formation, motility, migration, colony formation, and anticancer resistance, were enhanced in the CD133high population compared to CD133low HCT116 cells. Levels of NRF2, which were elevated in CD133high HCT116, were suppressed by CD133-silencing. In line with these, the analysis of The Cancer Genome Atlas (TCGA) database showed that high levels of CD133 expression are correlated with increased NRF2 signaling, and alterations in CD133 gene or expression are associated with unfavorable clinical outcome in colorectal carcinoma patients. These results indicate that the CD133/NRF2 axis contributes to the development of CSC-like properties in colon cancer cells, and that PI3K/AKT signaling activation is involved in CD133-mediated NRF2 activation.


INTRODUCTION
Cancer stem cell (CSC) is a subpopulation of tumor cells, which is known to account for 1-2% of tumors. Initially, Dick and colleagues identified the leukemia-initiating cluster of differentiation (CD)-34 + CD38cells from acute myeloid leukemia (AML) and showed that these cell fractions have differentiation and self-renewal capacities using serial transplantation in immunodeficient mice (1). CSCs share several common characteristics with normal stem cells, including self-renewal capacity, asymmetric division, and differentiation potential (2)(3)(4). In addition, CSCs attribute core characteristics to aggressive cancers due to their intrinsic resistance to anticancer treatment. Upregulation of drug efflux transporters, increased expression of the reactive oxygen species (ROS) scavenging system, and promotion of DNA damage repair are observed in CSCs, which enhance their survival in response to chemo-and radiotherapy (5)(6)(7)(8)(9). Several cell surface molecules, such as CD44, CD133, and ATP-binding cassette subfamily G member 2 (ABCG2), and transcription factors, such as Kruppellike factor 4 (KLF4) and octamer-binding transcription factor 4 (OCT4), have been used to isolate and characterize CSCs (2,8,10,11) CD133 (Prominin-1) is a transmembrane penta-span glycoprotein localized in cholesterol-based lipid rafts in the plasma membrane (12). It has two large extracellular loops, an N-terminal extracellular domain and a C-terminal intracellular domain, and eight glycosylation sites. Since its first identification in human hematopoietic stem cells (13), CD133 has been recognized as a marker of CSCs (14). The CD133-positive population was identified as 2.5% of the total tumor cells from colon cancer tissues and reproduced original tumors in immunodeficient mice (15). In an animal model of renal capsule transplantation, all colon cancer-initiating cells were CD133-positive, while CD133-negative cells, which comprised majority of cancer specimens, were not able to initiate tumorigenesis (16). High CD133 expression in colorectal cancer is correlated with low survival of patients with cancer (17). Furthermore, CD133 expression is associated with an aggressive cancer phenotype. CD133 overexpression in pancreatic cancer cells induced epithelial-mesenchymal transition (EMT) and enhanced cancer metastasis in athymic mice (18). CD133-positive cells from primary non-small cell lung cancer (NSCLC) specimens exhibited higher levels of genes associated with stemness, migration, and drug efflux than CD133-negative cells (19). Additionally, cisplatin treatment in primary tumor xenografts showed that the CD133-positive population survived after therapy. Suppression of ABCG2 in CD133-positive colon cancer cells enhances their apoptotic response to chemotherapy (20).
Although NRF2 shows protective roles in normal cells under stressful conditions, elevated levels of NRF2 in cancers promote cancer cell survival and facilitate tumor growth, cancer progression, and development of resistance to therapy (31)(32)(33). In particular, there is evidence that NRF2 signaling is upregulated in several types of CSC models, such as tumor spheres, CD44 high cells, and aldehyde dehydrogenase (ALDH) high cancer cells. Additionally, this upregulation was responsible for the development of CSC-like properties, including therapy resistance, spheroid growth, enhanced migration capacity, and facilitated tumor growth (34)(35)(36)(37)(38)(39). In the present study, we investigated the potential relationship between CSC markers, CD133 and NRF2, in colon cancer cells and demonstrated the role of the CD133/NRF2 axis in the development of CSC-like properties using two CSC models of spheroid culture system and CD133 high subpopulation system.

Sphere Culture
Cells were plated at a density of 1 × 10 5 cells/mL in ultralow attachment 100 mm plates or 6-well plates. The cells were grown in serum-free DMEM and Nutrient Mixture F-12 medium supplemented with B27 (Life Technologies), 20 ng/mL epithelial growth factor, 20 ng/mL basic fibroblast growth factor (R&D Systems, Minneapolis, MN, USA), 5 mg/mL bovine insulin (Cell Applications Inc., San Diego, CA, USA), 0.5 mg/mL hydrocortisone (Sigma-Aldrich), and penicillin/ streptomycin. In sphere culture conditions, HCT116 cells were grown for 3-6 d and then harvested as described previously (37). The sphere number and size were counted using ToupView software (ToupTek, Hangzhou, Zhejiang, China).

Immunoblotting Analysis
Whole lysates were prepared by adding 5X sample buffer containing 250 mM Tris-HCl (pH 6.8), 10% sodium dodecyl sulfate (SDS; Biosesang, Gyeonggi-do, Korea), 30% glycerol, 0.25% bromophenol blue, and 5% b-mercaptoethanol (Sigma-Aldrich, Co.). Protein samples were separated in 8-10% SDSpolyacrylamide gels and then transferred to nitrocellulose membranes (Whatman GmbH, Dassel, Germany). The membranes were blocked with 5% skim milk for 1 h and then incubated with the primary antibody in 3% bovine serum albumin (BSA) overnight. After incubation with the secondary antibody, chemiluminescent images were detected using a LAS-4000 mini-imager (GE Healthcare Life Sciences, Piscataway, NJ, USA). The loading control was detected after antibodies removal using stripping buffer (Restore Western blot stripping buffer; Thermo Fischer Scientific Inc., Waltham, MA, USA) followed by membrane blocking, primary and secondary antibodies incubation, and chemiluminescent detection as described previously.

Flow Cytometry and Cell Sorting
Approximately 1×10 6 HCT116 cells were harvested and incubated in 2 mL CD133/1-APC staining dye (Miltenyi Biotec) with 98 mL buffer containing 2 mM ethylenediaminetetraacetic acid (EDTA) and 2% FBS for 40 min. IgG control samples were incubated with 2 mL REA control (S)-APC staining dye (Miltenyibiotec Korea) with 98 mL of buffer. The fluorescence intensity of the stained cells was analyzed using an FACS Aria III cell sorter flow cytometer (BD Biosciences, Franklin Lakes, NJ, USA), and CD133 high and CD133 low subpopulations were sorted as described previously (35).

MTT Assay
Cells were plated at a density of 3×10 3 cells/well in a 96-well plate and incubated with doxorubicin for 24 h. After the addition of MTT solution (2 mg/mL), the cells were further incubated for 4 h. The MTT solution was removed, 100 mL/well of dimethyl sulfoxide (DMSO; Sigma-Aldrich Co.) was added, and the absorbance was measured at 540 nm using SpectraMax (Molecular Devices, San Jose, CA, USA) (41).

Wound Healing Assay
To determine cell motility, CD133 high and CD133 low cells were plated in a 12-well plate at a density of 2.5×10 5 cells/well. When 95-100% confluency was achieved, a straight scratch was made on the surface using a pipette tip. Then, the cells were grown for 24 or 48 h in serum-free medium, and the migration of cells into the wounded area was photographed using a JULI TM Smart fluorescent cell analyzer (Digital Bio source, Seoul, Korea). The wound closure rate was determined using the initial and final wound widths, and the wound closure percentage was calculated by dividing the change in wound width by the initial wound width, as described previously (40).

Soft Agar Colony Formation Assay
Soft agar colony formation assay was performed to evaluate the anchorage-independent growth ability of the cells. Approximately 5×10 3 cells were suspended in the top soft agar layer (0.35% soft agar) and seeded into 6-well plates, which were pre-coated with 0.5% base agar. Colonies were allowed to grow at 37°C in a 5% CO 2 incubator for 2-3 weeks, and colony numbers were counted using an ECLIPSE Ti inverted microscope and the NIS-Elements AR (V. 4.0) computer software program (NIKON Instruments Korea, Seoul, Republic of Korea), as described previously (35).

Correlation of CD133/NRF2 With Colorectal Cancer Prognosis
We used gene expression data of colorectal adenocarcinoma patients (n=526) available at the Cancer Genome Atlas (TCGA) Pan-Cancer Atlas data set. Analyzed data were visualized using the cBioPortal (http://cbioportal.org) to investigate the gene expression levels of NRF2, NQO1, and PIK3CA depending CD133 mRNA levels. Levels of mRNA are presented as RSEM processed using the RNA-Seq by Estimation Maximization (RSEM) algorithm in log2 scale. In addition, we examined the overall survival rates depending on CD133 expression levels and genetic alterations in CD133 and NRF2. These Kaplan-Meier survival estimates were generated by log-rank nonparametric test in the cBioPortal. P-values are derived from student t-test, and qvalues are obtained from Benjamini-Hochberg procedure.

Statistical Analysis
We conducted multiple comparison tests for different treatment groups using histomorphometric analysis. The data were analyzed using one-way analysis of variance (ANOVA) followed by Tukey's multiple comparison test to determine which pairs of groups were significantly different. Statistical analyses were conducted using GraphPad Prism 5 (GraphPad Software, Inc., La Jolla, CA, USA). Differences were considered statistically significant at P < 0.05.

High CD133 Levels Are Associated With Facilitated Sphere Growth and CSC Marker Expression
To examine the relationship between CD133 and CSC-like properties, we first examined the expression levels of CD133 in different carcinoma cell lines. When the basal levels of CD133 were determined in colorectal carcinoma cell line HCT116 and Colo205, pancreatic carcinoma cell line PANC-1 and MIA PaCa-2, and breast carcinoma cell line MCF7, HCT116 showed the highest level of CD133 (Supplementary Figures 1A, B, 2B). FACS analysis with CD133-specific antibody showed that 79.8% of HCT116 total cell fraction expressed CD133 ( Figure 1A). In an attempt to investigate the role of CD133 in CSC-like property development, we used a sphere culture system, which was shown to be a CSC-enriched system (42). When HCT116 cells were cultured in ultra-low attachment plates, the transcript level of CD133 increased 3.16-fold in HCT116 spheres and CD133 protein levels were also elevated ( Figures 1B, C). Whereas, sphere culture of Colo205, PANC-1, and MIA-PaCa-2 did not show the elevations in CD133, which shows that HCT116 can be used as our experimental model (Supplementary Figures  2A, B). In accordance with CD133 elevation, levels of CSC markers, including KLF4 and ABCG2, were significantly increased in HCT116 colonospheres ( Figures 1D, E). We then assessed the association of CD133 with CSC-like properties by silencing CD133 in HCT116 ( Figure 1F). The CD133-silenced colonospheres expressed lower levels of KLF4 and ABCG2 than the nonspecific siRNA-transfected colonospheres ( Figure 1G). In addition, the number of spheres with a diameter greater than 70 mm was markedly diminished by CD133 silencing ( Figure 1H). These results showed the critical contribution of CD133 to the development of CSC-like properties in colon cancer cells.

NRF2 Elevation Is Mediated by CD133 in Sphere Cultured Colon Cancer Cells
In HCT116 colonospheres, transcription levels of NRF2 and its target genes, such as GCLC, AKR1C1, and NQO1 were all increased compared to the HCT116 monolayer ( Figure 2A). Increased protein levels of NRF2, GCLC, AKR1C1, and NQO1 were also confirmed by western blotting ( Figure 2B). Next, we tested the potential involvement of CD133 in NRF2 signaling activation by silencing CD133 in colonospheres. When CD133silenced HCT116 was cultured in sphere conditions, the elevations in NRF2 and its targets GCLC, AKR1C1, and NQO1 were attenuated when compared to the control siRNAtransfected spheres (Figures 2C, D). Whereas, NRF2 mRNA levels, which were elevated in spheres, were not significantly

Activation of PI3K/AKT Signaling Is Involved in CD133-Mediated Elevation of NRF2
There have been reports showing that CD133 induces PI3K/ AKT signaling activation in colon cancer cells as well as glioma stem cells (43,44). In addition, AKT-mediated GSK-3b phosphorylation stabilizes NRF2 (27). Based on these results, we monitored the activation levels of AKT/GSK-3b using western blotting to elucidate the molecular events involved in CD133mediated NRF2 elevation in colonospheres. The level of phosphorylated AKT (p-AKT) at Ser473 was higher in colonospheres than in monolayer HCT116 cells, implying the activation of PI3K ( Figure 3A). Subsequently, the level of phosphorylated GSK-3b (Ser9), an inactive form of proteasomal NRF2 degradation, was elevated in sphere-cultured HCT116 cells.
These results show that CD133-mediated PI3K/AKT activation and resultant GSK-3b inactivation could be a cause of NRF2 elevation in colonospheres. Indeed, when CD133 was silenced, levels of p-AKT and p-GSK-3b were reduced in colonospheres ( Figure 3B), and treatment with the PI3K inhibitor LY294002 transfection were grown in the sphere culture system, and transcript levels of CD133, GCLC, AKR1C1, NQO1, and NRF2 were measured. a P < 0.05 compared with the monolayer. b P < 0.05 compared with the siCtrl group. (D) Protein levels for NRF2, GCLC, AKR1C1, and NQO1 were monitored in sphere cultured HCT116 cells with either the non-specific siRNA (siCtrl) or CD133-specific siRNA (siCD133) transfection. Bar graph represents quantified protein levels from at least three experiments. Quantification results of western blotting were relative values to the loading control GAPDH. a P < 0.05 compared with the monolayer. b P < 0.05 compared with the siCtrl group. NT, non-transfection group. (10 mM) repressed these elevations ( Figure 3C). Additionally, PI3K inhibitor treatment diminished NRF2 levels and target gene expression levels in HCT116 spheres ( Figures 3C, D). These results showed that CD133 contributes to NRF2 upregulation via PI3K/AKT activation and subsequent inhibition of p-GSK-3b.

NRF2 Contributes to Facilitated Sphere Formation and CSC Marker Elevation
In order to assess the role of NRF2 in colonosphere formation, NRF2-silenced HCT116 was cultured in sphere culture condition ( Figure 4A). NRF2-silencing affected the sphere-forming capacity of HCT116 cells, and the sphere number was reduced ( Figure 4B). In line with this, the levels of CSC markers KLF4 and ABCG2 were significantly lower in the NRF2-silenced colonospheres than in the control siRNAtransfected colonospheres ( Figure 4C). These results confirmed the functional contribution of NRF2 to the sphereforming capacity.

CD133 high HCT116 Cells Display Enhanced CSC-Like Properties and NRF2 Activation
To confirm the relationship between CD133/NRF2 and CSC-like properties, we isolated CD133-positive and CD133-negative cell fractions from total HCT116 cells, and established CD133 low and CD133 high cell lines ( Figure 5A). These cell lines were cultured for up to 1 month, and high CD133 expression was maintained along with elevated ABCG2 and KLF4 levels ( Figure 5B). As phenotypic characteristics, the cell growth rate of CD133 high HCT116 cells was higher than that of CD133 low cells ( Figure 5C), and anchorage-independent colony formation was enhanced ( Figure 5D). In addition, cell motility in the wound healing assay ( Figure 5E) and sphere forming capacity ( Figure 5F) were higher in CD133 high HCT116 cells than in CD133 low cells. In line with these CSC-like properties, the cytotoxic response to anticancer treatment was determined using doxorubicin. We observed that 0.5 to 1 mM doxorubicin treatment for 24 h showed similar rates of growth inhibition, and  the cell growth inhibition by 0.5 mM doxorubicin treatment was alleviated in CD133 high HCT116 cells ( Figure 5G). When CD133 levels were silenced in CD133 high HCT116 cells, the levels of the CSC markers ABCG2 and KLF4, which were elevated in this cell line, were diminished ( Figure 5H). Additionally, the levels of NRF2 and NQO1 were higher in CD133 high HCT116 cells than in CD133 low cells and were repressed by CD133 silencing ( Figure 5H). These results provide direct evidence that the CD133-enriched population exhibited enhanced colony formation, cell migration, colony formation, and drug resistance, which is accompanied by NRF2 activation.

CD133 Alterations Are Associated With NRF2 Elevation and Poor Clinical Outcome in Patients With Colorectal Cancer
In an attempt to investigate the clinical implication of the linkage between CD133 and NRF2, we analyzed clinical data from the TCGA Pan-Cancer Atlas database using the cBioPortal interface. A total of 526 gene expression data from colorectal  Figure 6A). Additionally, patients whose tumors have NRF2 or CD133 gene alteration showed shorter overall survival estimates (median survival months=51.48) than patients with unaltered CD133 gene groups (median survival months=83.24) ( Figure 6B). These imply the potential correlation between CD133 gene alteration and clinical outcome of colorectal cancer patients. The CD133 high and CD133 low HCT116 cells were grown in the absence of fetal bovine serum (FBS) and relative cell growth was assessed using MTT assay after 24 h and 48 h of plating. a P < 0.05 compared to each 24 h group. b P < 0.05 compared to CD133 low HCT116 cells. (D) The CD133 high and CD133 low HCT116 cells were suspended in the top soft agar layer (0.35% soft agar) and anchorage-independent growth was monitored for 2 weeks. Colony number was counted using an ECLIPSE Ti inverted microscope and the NIS-Elements AR (V. 4.0) software. Bar graph represents quantified results from at least three experiments. a P < 0.05 compared to the CD133 low HCT116 cells. (E) Cell migration ability was assessed in the CD133 high and CD133 low HCT116 cells using a wound-healing assay for 24 h and 48 h. Bar graph represents quantified results from at least three experiments. a P < 0.05 compared to the CD133 low HCT116 cells. (F) CD133 high and CD133 low HCT116 cells were grown in sphere culture condition, and sphere formation capacity was assessed by measuring the number and average size of the spheres. Bar graph represents quantified results from at least three experiments. a P < 0.05 compared to the CD133 low HCT116 cells. (G) Inhibition of cell growth was monitored following doxorubicin (0.5 and 1 mM) incubation for 24 h using MTT assay. Values represent the mean ± SEM from four independent experiments. a P < 0.05 compared to the doxorubicin-treated CD133 low group. (H) CD133 high and CD133 low HCT116 cells were transfected with CD133 siRNA and protein levels of CD133, NRF2, NQO1, KLF4, and ABCG2 were determined. Bar graph represents quantified protein levels from at least three experiments. Quantification results of western blotting were relative values to the loading control GAPDH. All values represent the mean ± SEM from three independent experiments. a P < 0.05 compared with the CD133 low siCtrl group. b P < 0.05 compared with the CD133 high siCtrl group.
Among 526 colorectal cancer patients, 52 patients exhibited higher CD133 mRNA levels when compared to unaltered CD133 mRNA group (n=474). The Kaplan-Meier survival analysis showed that the median survival month of patients with high CD133 expression is 81.37 month, which is lower than that of patients with unaltered CD133 expression (83.24 month), although no statistical significance obtained ( Figure 6C). In addition, CD133 mRNA levels were associated with increased mRNA levels of NRF2 and NQO1 in these patients. Mean log2 mRNA expressions of NRF2 were 10.79, 11.21, and 11.28 in CD133-low, unaltered, and CD133-high group, respectively ( Figure 6D). NQO1 levels were increased depending on CD133 levels ( Figure 6D). PIK3CA mRNA levels were also relatively high in CD133-high group when compared to unaltered group ( Figure 6F). These results indicate that CD133 expression is associated with NRF2 signaling activation in colorectal cancers, and further suggested the correlation of

DISCUSSION
The major characteristics of CSCs, which include refractory response to conventional chemotherapy and radiotherapy, can be explained by the elevation of drug efflux transporters, enhanced DNA repair ability, and activation of the ROS defense system. The side population of cancer cells, which excludes the fluorescent dye Hoechst 33342, expresses a high level of ABCG2 (BCRP) and displays chemoresistant phenotypes (45). The CD133-high fraction from glioma cells was more resistant to radiotherapy than the CD133-negative fraction, and CD133-high cells activated the molecular event for the DNA damage checkpoint and enhanced DNA repair capacity following radiotherapy (46). The CD44 + /CD24cell fraction was found to be resistant to radiation, and maintenance of low ROS levels was associated with radioresistance (47). CSCs from human breast tumors exhibited less DNA damage and a higher rate of survival after irradiation compared to non-CSCs, and low levels of ROS in CSCs were attributed to increased expression levels of ROS detoxifying systems, such as glutamate-cysteine ligase and GSH synthetase (9). In leukemia with a high frequency of stem cells, ROS levels were low and ROS-scavenging GPX3 levels were high compared to leukemia with a low frequency of stem cells (48). Leukemic stem cells exhibited high levels of FoxoO3a expression, and deletion of foxO blocked the initiation of myeloid leukemia in a mouse model (49). In head and neck squamous cell carcinoma, a CSC marker CD44 variant was found to directly bind to the cystine/glutamate antiporter xCT, thereby increasing GSH synthesis, which is involved in therapy resistance (50). These reports consistently support the importance of the ROS detoxifying system in CSCs for maintaining low ROS levels and high survival under therapy stress.
In the current study, we demonstrated a positive linkage between CD133 and NRF2 in CSC-like properties using the colonosphere culture system and the CD133 high subpopulation. Sphere culture of the colorectal cell line HCT116 led to an increase in CD133 expression along with elevated CSC markers, such as KLF4 and ABCG2. CD133-silencing suppressed sphere-forming capacity and expression of KLF4 and ABCG2, which indicates the critical role of CD133 in CSC-like properties development. NRF2 signaling was also upregulated in colonospheres and partly responsible for CSC marker elevation and sphere-forming capacity. Notably, NRF2 activation in colonospheres was CD133-dependent: CD133silencing inhibited NRF2 elevation and attenuated the expression of GCLC, AKR1C1, and NQO1. The relationship between CD133 and NRF2 was confirmed in an isolated CD133 high subpopulation from HCT116 cells. Established CD133 high cell line showed higher doxorubicin resistance, colony formation, sphere formation, and migration capacity than CD133 low cell line, and CD133-silencing in CD133 high cells suppressed NRF2 activation and KLF4 elevation. These in vitro results were supported by clinical relationship between CD133 and NRF2, which were obtained from ATCG database. In colorectal carcinoma patients, overall survival rates were diminished by CD133 and CD133/NRF2 gene alterations, and high CD133 mRNA levels also showed a relationship with reduced overall survival rates. Additionally, high CD133 mRNA levels in colorectal carcinoma showed positive correlations with high transcript levels of NRF2 and NQO1. Taken together, these results suggest that CD133 mediates the activation of NRF2 signaling, which in turn contributes to the CSC-like properties of CD133 high colon cancers.
As a molecular event for CD133-mediated NRF2 activation, we suggest the involvement of PI3K/AKT/GSK-3b. Multiple reports have suggested the activation and contribution of PI3K/AKT signaling in CD133-positive CSCs. Microarray analysis revealed that expression of genes related to the PI3K/ AKT signaling pathway was elevated in sphere-cultured CD133 + /CD44prostate CSCs, and knockdown of phosphatase and tensin homolog (PTEN) stimulated sphere formation by inhibiting PI3K/AKT signaling (51). Phosphorylation of Tyr828 residue in the cytoplasmic domain of CD133 mediates binding with PI3K, and subsequently activates PI3K/AKT signaling for self-renewal and tumorigenicity of glioma stem cells (44). In line with these findings, an inhibitor of PI3K/AKT suppressed the proliferation and stemness of colon CSCs (52). In our colonosphere system, activation of the PI3K/AKT axis and consequent phosphorylation of GSK-3b at Ser9 were observed. Since AKT-mediated phosphorylation inhibits GSK-3b activity for b-TCRP-dependent degradation of NRF2, activated PI3K/ AKT signaling is often associated with NRF2 activation in multiple types of cancers. In breast cancers with oncogenic PI3K/AKT activation, NRF2-driven GSH biosynthesis is stimulated, which is required for oxidative stress resistance, tumor spheroid formation, and colony formation. In addition, elevation of NRF2 targets showed a positive correlation with mutation status in the PI3K/AKT pathway (53). In the absence of KEAP1, the deletion of PTEN could further elevate NRF2 levels, which accompanied GSK-3b inactivation via PI3K/AKT activation (8,54). In our study with colonospheres, CD133silencing reduced phosphorylation of AKT/GSK-3b, and the PI3K inhibitor LY294002 blocked AKT-mediated GSK-3b phosphorylation and attenuated NRF2 target genes expression. These results suggested that CD133 activates PI3K/AKT signaling, which in turn stabilizes NRF2 protein via GSK-3b inhibition in colonospheres. Of note, we observed that transcript levels of NRF2 were also higher in colonospheres than those in monolayer cultured cells (Figures 2A, C). As NRF2 transcription is regulated by its 5-flanking upstream ARE as a positive feedback loop (55), it can be plausible that PI3K/AKTmediated NRF2 stabilization elevates NRF2 transcription.
Several reports have demonstrated that NRF2 signaling plays a role in CSC maintenance and therapy resistance. In primary glioma stem cells from human glioblastoma tissues, NRF2 knockdown disrupts self-renewal and pluripotency (38). NRF2 signaling is elevated in spheroid cultured breast cancer cells, and high NRF2 levels are required for the maintenance of low ROS levels and taxol resistance (56). Similarly, NRF2 elevation in mammospheres resulted in high levels of drug efflux transporters and antioxidant genes, and NRF2-silencing blocked sphere growth and induced chemosensitization (37). Approximately 3.1% of cervical CSCs were isolated from tumor specimens, NRF2 was aberrantly upregulated, and NRF2 silencing could sensitize cervical CSCs to DNA damage-induced apoptosis (34). CD133 + /CD44 + colon CSCs express high levels of ABCB1 via NRF2 elevation, which is associated with doxorubicin resistance (57). In our previous study, CD44 high CD24 low breast CSCs exhibited high levels of NRF2 signaling, and NRF2-silencing led to retarded tumor growth, suppression of sphere formation and invasion capacity, and anticancer sensitization (36). The ovarian CSC fraction with high ALDH1 retained low levels of ROS, which was accompanied by NRF2 signaling activation (35). These results suggested a critical role of NRF2 signaling in CSC maintenance and tumor resistance/recurrence and support our current observation of the association between CD133 and NRF2.
Taken together, our results indicate that CD133, a molecular marker of colon CSCs, leads to PI3K/AKT-associated NRF2 activation (Figure 7). High NRF2 levels in spheroid cultured HCT116 cells and the CD133 high subpopulation contributed to the aggressive CSC phenotypes, including anticancer resistance, sphere formation, anchorage-independent colony formation, and migration potential. Therefore, the NRF2 axis might be a promising target for the inhibition of therapeutic resistance and enhancement of survival capacity under stress conditions in CD133 high CSCs.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding author.

AUTHOR CONTRIBUTIONS
JP carried out the experiment with support from SKK and SPH. SKK and SPH contributed to sample preparation, data analysis, and figure preparation. JP and MKK wrote the manuscript with the help from SKK and SPH. BHC contributed to data analysis and figure preparation. MKK conceived the original idea and supervised the project. All authors provided critical feedback and helped shape the research, analysis and manuscript preparation. All authors contributed to the article and approved the submitted version.

SUPPLEMENTARY MATERIAL
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fonc.2021.808300/ full#supplementary-material FIGURE 7 | A schematic diagram of the activation of NRF2 signaling and the development of CSC-like properties in CD133-enriched cancer cells. In colonospheres and the CD133-positive cell fraction, high CD133 levels induce PI3K/AKT activation and subsequently inactivate GSK-3b to stabilize NRF2 protein, which results in the acquisition of CSC-like properties, including anchorage-independent growth, sphere formation, facilitated migration, and resistance to anticancer drugs.