Cancer stem–like cells are hypothesized to be the major tumor-initiating cell population of human cutaneous squamous cell carcinoma (cSCC), but the landscape of molecular alterations underpinning their signaling and cellular phenotypes as drug targets remains undefined. In this study, we developed an experimental pipeline to isolate a highly enriched CD133+CD31CD45CD61CD24 (CD133+) cell population from primary cSCC specimens by flow cytometry. The CD133+ cells show enhanced stem–like phenotypes, which were verified by spheroid and colony formation in vitro and tumor generation in vivo. Gene expression profiling of CD133+/− cells was compared and validated, and differentially expressed gene signatures and top pathways were identified. CD133+ cells expressed a repertoire of stemness and cancer-related genes, including NOTCH and NOTCH1-mediated NF-κB pathway signaling. Other cancer-related genes from WNT, growth factor receptors, PI3K/mTOR, STAT pathways, and chromatin modifiers were also identified. Pharmacologic and genetic targeting of NOTCH1, IKKα, RELA, and RELB modulated NF-κB transactivation, the CD133+ population, and cellular and stemness phenotypes. Immunofluorescent staining confirmed colocalization of CD133+ and IKKα expression in SCC tumor specimens. Our functional, genetic, and pharmacologic studies uncovered a novel linkage between NOTCH1, IKKα, and NF-κB pathway activation in maintaining the CD133+ stem SCC phenotypes. Studies investigating markers of activation and modulators of NOTCH, IKK/NF-κB, and other pathways regulating these cancer stem gene signatures could further accelerate the development of effective therapeutic strategies to treat cSCC recurrence and metastasis. Mol Cancer Ther; 17(9); 2034–48. ©2018 AACR.

Human cutaneous squamous cell carcinoma (cSCC) has been increasing over the past several decades, with more than 700,000 cases in the United States annually (1, 2). Despite surgery, radiation, and chemotherapy, SCC cells can escape treatment and reform tumors and metastasis, increasing morbidity and mortality. This gives prominence to the existence of a subpopulation of SCC cells capable of tumor initiation and therapeutic resistance, and the importance of characterizing and targeting the molecular alterations mediating their maintenance for cancer prevention and treatment.

The cancer stem cell (CSC) hypothesis holds that tumors are a hierarchical organ derived from cell subpopulation(s) capable of self-renewal, called cancer stem–like cells or tumor-initiating cells (TIC; ref. 3). TICs exhibit stem–like and tumor-initiating properties, including increased self-renewal/colony formation, tumor-forming capacity, as well as altered migration, differentiation, and therapeutic sensitivity. Evidence for the existence of TICs has been obtained in different types of human solid tumors through identification of subpopulations enriched for surface determinants or enzymatic markers, such as CD133, CD44, CXCR4, and ALDH1 (3). We previously demonstrated that cell membrane protein CD133 (also called prominin-1) is specifically expressed by cSCC cells enriched for a TIC phenotype, and not by CD45 and other nonepithelial subpopulations overlapping CD44 (4). CD133+ cSCC cells exhibited long-term proliferative ability, self-renewal, sphere formation, regeneration of differentiated SCC tumor cells, and enriched tumor-initiating capacity for xenografts in immunodeficient mice. CD133+ has also been identified as a biologically and clinically relevant marker for subpopulations of lung and head and neck SCC (HNSCC), where it is associated with tumor-initiating capacity, aggressive clinical features, and resistance to cytotoxic therapies (5, 6).

Recent studies of cSCC, HNSCC, and lung SCC tumors have identified significant genetic alterations involving components of several common and distinct pathways important in cell growth, death or survival, migration, and epithelial/mesenchymal differentiation (7–10). These alterations are implicated in inactivation or activation of several canonical pathways, including NOTCH, WNT, HEDGEHOG, NF-κB, growth factor receptors, RAS-mitogen–activated protein kinase, PI3K–Akt–mTOR, and TP53, but their expression and role in TIC versus other populations in established cSCC have not been dissected. Among these are numbers of genes linked to stem cell maintenance or differentiation (11–14). NOTCH signaling is important in epithelial differentiation and as a suppressor of tumor development in a subset of HNSCC, but is activated in others (8, 15). IKK and NF-κB signaling has been implicated in promoting tumor cell survival, inflammatory, and angiogenesis responses (9). However, how these signaling pathways and function of corresponding molecular components contribute to the regulation of phenotype of the CSC/TIC subpopulation in cSCC tumors are not well understood. Identification of deregulated components of pathways critical to maintenance of the CD133+ CSC phenotype could potentially help identify targets for precision medicine for prevention and therapy.

In this study, we integrated the molecular profiling, signaling pathway, mechanistic, and pharmacological studies of the CD133+ subpopulation in primary human cSCC tumors and cell line models. Through cell sorting with multiple positive and negative selection markers by flow cytometry, we successfully isolated live CD133+ cells that form spheroid colonies in vitro and tumors in vivo. This small distinct CD133+ population differentially expresses stem-like and cancer gene signatures linked to NOTCH1-mediated NF-κB modulation, NF-κB, and WNT pathways. Characterization of the landscape of gene signatures in these CD133+ stem cells revealed activation of multiple pathways, which were linked to NOTCH and NF-κB signaling networks and showed sensitivity to genetic and pharmacologic inhibitors of NOTCH and NF-κB. Our functional, genetic, and pharmacologic studies uncovered a linkage between NOTCH1, IKKα, and NF-κB pathway activation in maintaining the CD133+ population and its self-renewal ability in established primary cSCC and cell lines.

Human tumor tissue samples

Deidentified primary human skin cSCC tissue samples were obtained under an exemption from IRB approval by the Office of Human Subjects Research, National Institutes of Health, from Potomac Ambulatory Surgery Center, Bethesda, MD, and Braun Dermatology Associates, Washington, DC.

Preparation of single-cell suspension from cSCC specimens

Tumor tissues were gently chopped into small pieces and incubated with collagenase III and dispase, then further dissociated using GentleMACS Dissociator. The dissociated tissues were spun down, incubated with trypsin and EDTA and filtered, spun down, and the cSCC single-cell suspension was then ready for staining and FACS sorting for sphere, tumorigenicity assays, and expression profiling.

Flow cytometry analysis and FACS sorting

cSCC single-cell suspensions were labeled with multiple antibodies and sorted by BD FACSAria-II following standard protocol as detailed in Supplementary Methods. Analysis was performed with BD LSR II and FlowJo7.6.5.

Sphere formation assay

Fifteen cSCC samples were collected and used for sphere experiments after preparation of single-cell suspension. The cSCC single-cell suspension of unsorted or sorted CD133+ or CD133 cells was seeded on NIH 3T3 fibroblast cells feeder layer and cultured for 14 to 21 days. Spheres ≥50 μm were counted under the microscope.

In vivo mouse model

Animal studies were performed under a protocol approved by the Institutional Animal Care and Use Committee of the National Cancer Institute. Six- to 8-week-old nude mice were obtained from NCI Frederick. To establish a niche, 106 human endothelium and fibroblast cells with 120 μL Matrigel were injected into 0.5 × 0.5 cm GelFoam sponges implanted subcutaneously 2 weeks prior to tumor cell transplantation. Freshly obtained, dissociated, and sorted CD133+, CD133, and unsorted cells at doses of 106, 105, 104, 103, or 102 cells were mixed with 106 human endothelium and fibroblast cells in Matrigel, for injection into the established GelFoam niche. Each dose level included 15 mice, with 5 mice each that received freshly isolated primary sorted CD133+, CD133, or unsorted SCC tumor cells. To obtain 106 rare CD133+ and matched CD133 cells, we sorted dissociated suspensions from pools of 5, 5, 6, 6, and 6 (28 total) fresh tumors for inoculation of 5 pairs of mice. CD133+ and CD133 cells sorted from 5 tumors were sufficient to inoculate 5 paired recipients at 102, 103, 104, and 3 pairs at 105, and cells pooled from 3 and 4 additional tumors each (12 total) were required to sort 105 CD133+ and CD133 cells for 2 additional pairs. Five tumors were freshly dissociated and 106, 105, 104, 103, and 102 cells from each were inoculated into paired recipients as unsorted controls for each dose level.

Affymetrix genechip human gene 1.0 ST array

Total RNA was isolated after FACS sorting, and 15 samples were used with RNA integrity numbers equal to or greater than 7.0, as the control for RNA quality. Single-stranded cDNA was generated, fragmented, labeled, and hybridized with standard Affymetrix protocol. Data were scanned with an Affymetrix GeneChip Scanner 3000 and analyzed with Partek and GeneGo. The microarray data have been submitted to NCBI with GEO submission number (GSE84588).

Nanostring nCounter gene expression assay

Total RNAs (100 ng) from sorted CD133+ and CD133 cells from seven cSCC samples were hybridized, purified, and immobilized in the Prep Station. The images were measured and further analyzed with nSolver.

NF-κB family transcription factor DNA-binding assays

Nuclear fraction of SCC13 cells was placed on a plate coated with immobilized NF-κB consensus oligonucleotides. Activated NF-κB is recognized by antibodies, followed by the HRP-conjugated second antibody, and quantified by spectrophotometry.

Drugs and treatment

The following compounds were used for inhibition of gamma-secretase-NOTCH and inhibitor-kappaB kinase in experiments: DAPT (https://pubchem.ncbi.nlm.nih.gov/compound/dapt), catalog D5942, Sigma Aldrich); RO4929097 (https://pubchem.ncbi.nlm.nih.gov/compound/49867930#section=Top), catalog No. ADV465749148, Sigma Aldrich); wedelolactone (https://pubchem.ncbi.nlm.nih.gov/compound/Wedelolactone#section=Top, catalog #56639, Sigma Aldrich).

SCC13 cells were treated with two different NOTCH inhibitors, DAPT or R04929097, in different doses and measured by FACS analysis. Wedelolactone (10 μmol/L) was used to treat SCC13 cells for 24 hours. Cells then harvested and stained with CD133 antibody for FACS analysis. Spheres or colonies were counted after 5 days of drug treatment. Medium with drug was changed every 3 days.

Luciferase reporter gene assays

SCC cells were cotransfected with NF-κB Luciferase and β-Gal and then treated with wedelolactone and TNF-α at different time points. Relative luciferase activity was normalized to the cells without treatment.

Western blotting

Whole cell lysates (15 μg) were used for electrophoresis, then protein was transferred onto a Nitrocellulose membrane, and incubated with first and second antibodies following standard protocol.

Statistical analysis

Mean and standard deviation (SD) were calculated for different experiments, and the statistical significance was calculated using Student t test. P value <0.05 was considered a statistically significant difference. To analyze microarray data, a two-way analysis of variance (ANOVA) for the paired samples was used.

Additional information about human SCC specimens, antibodies, and instrumentation settings is presented in the Supplementary Information.

Clinical characteristics of human primary cutaneous squamous cell carcinomas (cSCC)

Eighty-two primary cSCC were collected fresh after Mohs micrographic surgery from patients. The diagnosis and the tumor's grade was confirmed by a dermatopathologist, and the clinical characteristics are summarized in Supplementary Table S1. The majority of subjects were men (78%), and tumors were predominantly located on sun-exposed skin from head (62.2%) and limbs (34.2%). Most tumors were well or moderately differentiated (85.4 + 12.2 = 97.6%) cSCC of less than 2 cm2 in size. The collected tumors were freshly delivered from the clinics and processed on the same day, to isolate stem cell subpopulations for in vitro and in vivo characterization of purity, stemness phenotype, molecular profiling, and functional validation studies (Supplementary Table S1; Supplementary Fig. S1). The clinical characteristics of tumor samples used in different experimental studies below are summarized Supplementary Table S2.

CD133+CD31CD45CD61CD24 cells isolated from cSCC exhibit enriched CSC/TIC features by in vitro and in vivo assays

New protocols to isolate, dissociate, and separate cSCC cells from other populations in tumor specimens were optimized, as summarized in Supplementary Methods and Supplementary Fig. S1. Each cSCC sample was microdissected to carefully remove excess surrounding stroma, and regions with >90% malignant epithelia were obtained. We developed protocols for cell sorting and improved purification of CD133+ and CD133 tumor cells after gating out differentiating keratinocytes (CD24), stromal endothelial cells (CD31), leukocytes (CD45), and fibroblasts (CD61). The percentage of cells expressing these CD markers in cSCC varied among individual tumors by FACS analysis (Fig. 1A). CD133+ cells represent a rare population that resides within a relatively stable range from 0.14% to 1.7% (mean 0.62%; Fig. 1A and B) and displayed a slight increase with tumor size that was not significant (R2 = 0.037, P > 0.05). Sorted CD133+CD31CD45CD61 CD24 (hereafter called CD133+) cells have a purity as high as 95% as detected by FACS analysis (Fig. 1C). More than 99% of FACS sorted CD133+ cells are pan-Keratin positive (Fig. 1D), indicating that the CD133+ subpopulation represents highly purified CD133+ keratinocytes from cSCC tissues. Thus, we isolated highly purified CD133+ CSC and compared with those relatively pure CD133 keratinocytes without stemness characteristics using this improved sorting and selection protocol with the five different cellular markers.

Figure 1.

Isolated CD133+CD61CD31CD24CD45 cell subpopulation exhibits CSC features in human cutaneous SCC (hsSCC). A, Distribution of cell surface markers in human SCC cells isolated from tumor specimens. B, FACS detection of CD133+ CSC population in hsSCC. The red square gate was used to quantify the CD133+ population after gating out the other four CD marker–positive cells. C, Purity detection after FACS sorting for CD133+ cells after sorting from CD133 cells, and gated CD61CD31CD24CD45 cell subpopulation from hsSCC. D, Cytokeratin staining for CD133+ cell subpopulation. CD133+ cells (99%) were pan-cytokeratin positive as shown in the right square gate on the right. E, Human SCC CD133+ cells are enriched for spheroid colony formation in vitro. Normal human keratinocytes grew as adherent monolayer on the culture dishes with NIH 3T3 feeder layer (left top, 40×), whereas the dissociated cSCC cells grew as tethered spheres on the NIH 3T3 feeder layer (right top, 40×). The lower two panels show enlarged spheroid images under microscope (100×). F, CD133+ cells exhibit statistically significantly greater spheroid formation numbers than CD133 (P < 0.05) or unsorted SCC cells (P < 0.05) when seeded the same number cells (1 × 103). G,In vivo tumor formation assay in nude mice for comparison of transplant of different numbers of primary cSCC cells. Each group includes 15 mice with 5 mice each that received the indicated number of primary SCC tumor cells from unsorted (green bars), sorted CD133 (yellow bars) and CD133+ (blue bars) cells, as described in Materials and Methods.

Figure 1.

Isolated CD133+CD61CD31CD24CD45 cell subpopulation exhibits CSC features in human cutaneous SCC (hsSCC). A, Distribution of cell surface markers in human SCC cells isolated from tumor specimens. B, FACS detection of CD133+ CSC population in hsSCC. The red square gate was used to quantify the CD133+ population after gating out the other four CD marker–positive cells. C, Purity detection after FACS sorting for CD133+ cells after sorting from CD133 cells, and gated CD61CD31CD24CD45 cell subpopulation from hsSCC. D, Cytokeratin staining for CD133+ cell subpopulation. CD133+ cells (99%) were pan-cytokeratin positive as shown in the right square gate on the right. E, Human SCC CD133+ cells are enriched for spheroid colony formation in vitro. Normal human keratinocytes grew as adherent monolayer on the culture dishes with NIH 3T3 feeder layer (left top, 40×), whereas the dissociated cSCC cells grew as tethered spheres on the NIH 3T3 feeder layer (right top, 40×). The lower two panels show enlarged spheroid images under microscope (100×). F, CD133+ cells exhibit statistically significantly greater spheroid formation numbers than CD133 (P < 0.05) or unsorted SCC cells (P < 0.05) when seeded the same number cells (1 × 103). G,In vivo tumor formation assay in nude mice for comparison of transplant of different numbers of primary cSCC cells. Each group includes 15 mice with 5 mice each that received the indicated number of primary SCC tumor cells from unsorted (green bars), sorted CD133 (yellow bars) and CD133+ (blue bars) cells, as described in Materials and Methods.

Close modal

To confirm whether the CD133+ population exhibits TIC features including self-renewal and tumorigenic ability established previously (3), we compared them with both CD133 and unsorted cells for sphere formation efficiency in vitro and tumor formation in vivo. Unlike normal human keratinocytes growing as monolayer colonies, dissociated cSCC cells grew as tethered spheres on NIH 3T3 feeder layers (Fig. 1E). Further, 1 × 103 sorted CD133+ cells formed 55 ± 8 spheres, while the same number of unsorted and CD133 cSCC cells grew only 23 ± 3 and 23 ± 4 spheres, respectively (Fig. 1F). The sphere formation in CD133+ was significantly increased (P < 0.0001) compared with unsorted cSCC or CD133 cells, whereas there is no difference between the unsorted and the CD133 tumor cell groups. The sphere numbers increased corresponding to increases in the seeded cell numbers, indicating that cSCC cell aggregation alone is unlikely to account for sphere formation.

To confirm the enrichment of TICs within the purified CD133+ population, we used the cSCC tumor xenograft assay in a mouse model established previously (3), modified to replenish the normal human fibroblasts and endothelial cells depleted by sorting. Varying concentrations of sorted CD133+, CD133, and unsorted cells were injected subcutaneously, to analyze tumor formation frequency by limiting dilution assay. A minimum of 105 unsorted cSCC cells was required for initiating tumor in this model (Fig. 1G, blue bar). By contrast, as few as 100 CD133+ cells generated tumors in vivo (Fig. 1G, green bars). At the same time, the CD133 subpopulation did not form any visible tumors with as high as 106 cells (Fig. 1G, yellow bar). Thus, the estimated tumor-initiating efficiency of the CD133+ subset is 1/245, which is dramatically enriched ∼340-fold compared with unsorted cSCC cells (1/83,333).

Gene expression profiling of CD133+ cells displays gene enrichment for stem cell markers and NOTCH and NF-κB signaling pathways

To investigate genes differentially expressed in the CD133+ cSCC population, we compared the whole genome expression profile between sorted CD133+ and CD133 cells after gating for the CD31CD45CD61CD24 population from 15 fresh cSCC tumor samples by using Affymetrix Whole Genome RNA microarray. The microarray data were normalized by the robust multichip average method and analyzed using the ANOVA paired test in Partek Genomics Suite, with P value ≤ 0.05 and expression level change > 1.5-fold. There were 2,089 genes significantly differentially expressed in CD133+ when compared with CD133 tumor cells. We further compared this gene list with 247 stem-cell–related genes from major publications of human stem cell microarray data (2, 11–14). We identified 80 overlapped genes between 2,089 genes identified in this study versus 247 stem-cell–related genes from major publications (Supplementary Fig. S2A). After subtraction of the 80 genes, the Venn diagram of pink circle shows total of 2,009 genes with significantly altered expression in CD133+ cells compared with CD133 cells. There are 167 stem genes from the literature after subtraction of 80 genes (green circle). Our data indicate that the sorted CD133+ cells are enriched for a stem cell gene expression signature. Moreover, a heat map of 80 signature genes shows that the genes are mainly distributed in two clusters (Fig. 2A). Cluster I includes 32 upregulated stem-related genes in CD133+ cells, including several NOTCH signaling pathway genes NOTCH1, JAG2, and MAML1; WNT pathway genes WNT3, FZD1, and CTNNB1; and Sonic Hedgehog pathway genes GLI2, SUFU, PTCH1, and GCNF (Germ Cell Nuclear Factor). Cluster II shows 48 downregulated genes in the CD133+ group, including NOTCH signaling-related genes JAG1, NOTCH4, and downstream components RBPJ, and NUMB; Wnt signaling negative regulator gene GSK3B and hedgehog signaling regulator HHIP. Unexpectedly, we also detected increased expression of hematopoietic cell determinant CD8A in several CD133+ samples. CD8a has been associated with CD133+ lymphomyeloid stem cells and cytotoxic T lymphocytes, which could potentially infiltrate cSCC and be detected among a subset of samples (16–19).

Figure 2.

Characteristics of the transcriptional profile of the CD133+ CSC subpopulation and its stem gene expression signature. A, Hierarchical clustering of altered expression of 80 upregulated or downregulated stem gene signature in sorted CD133+ cells compared with CD133 cells after gating for the CD61CD31CD24CD45 population from 15 tumor specimens using Affymetrix microarray (fold change >1.5, ANOVA analysis, P < 0.05). Section I: upregulated gene clustering and section II: downregulated gene clustering. Hierarchical cluster analysis of NOTCH (B) and NF-κB (C) pathways from 2,089 significantly different expressed genes in CD133+ CSCs compared with CD133 tumor cells. IPA analysis of the network of NF-κB and NOTCH pathway interaction (D), and NF-κB and Wnt pathway interaction (E). The solid line indicates direct interaction, and dotted line indicates indirect interaction. The pink color indicates upregulated genes, and the blue color indicates downregulated genes. The different shapes indicate different function category of the molecules as indicated.

Figure 2.

Characteristics of the transcriptional profile of the CD133+ CSC subpopulation and its stem gene expression signature. A, Hierarchical clustering of altered expression of 80 upregulated or downregulated stem gene signature in sorted CD133+ cells compared with CD133 cells after gating for the CD61CD31CD24CD45 population from 15 tumor specimens using Affymetrix microarray (fold change >1.5, ANOVA analysis, P < 0.05). Section I: upregulated gene clustering and section II: downregulated gene clustering. Hierarchical cluster analysis of NOTCH (B) and NF-κB (C) pathways from 2,089 significantly different expressed genes in CD133+ CSCs compared with CD133 tumor cells. IPA analysis of the network of NF-κB and NOTCH pathway interaction (D), and NF-κB and Wnt pathway interaction (E). The solid line indicates direct interaction, and dotted line indicates indirect interaction. The pink color indicates upregulated genes, and the blue color indicates downregulated genes. The different shapes indicate different function category of the molecules as indicated.

Close modal

GeneGo analysis of this 80-stem gene signature revealed the NOTCH pathway, NOTCH1-mediated NF-κB pathway, WNT pathway, Hedgehog, and NOTCH-mediated EMT signaling pathways are the top five most significantly ranked pathways in CD133+ cSCC cells (Supplementary Fig. S2B; Supplementary Table S3). To further investigate what kinds of signaling pathways contributed in the CD133+ population, we analyzed pathways ranking for all 2,089 significantly different expressed genes based on the ANOVA paired test results (Supplementary Fig. S2C). The broader analysis showed that molecular mechanisms of cancer, regulation of EMT pathways, PI3K/AKT, mTOR, NF-κB, ERK/MAPK signaling, and RAR activation are within the top 20 pathways in the CD133+ cell subset.

Because NOTCH, NOTCH1-mediated NF-κB, and WNT signaling appeared in top-ranked pathways in both stem cell supervised and unsupervised analyses, we generated heat maps for genes included in the top-ranked signaling/stem gene signatures from the 2,089 differentially expressed genes in cSCC (Fig. 2B and C; Supplementary Fig. S2D). The heat map of the NOTCH pathway showed two distinct clusters (Fig. 2B): cluster I displayed relative increase in gene expression, of ligand JAG2, receptors NOTCH1, NOTCH2, NOTCH3, and downstream transcription factor MAML1; cluster II and III revealed downregulated gene expression, including ligands JAG1, DLL1, receptors NOTCH4, glycosylation modification factor POFUT2, and downstream transcription factor component RBPJ. CD133+ cSCC did not exhibit significant differential expression of canonical NOTCH target genes HES and HEY1, consistent with a recent study that reported defective transcription of HES and HEY1 in subsets of human HNSCC with wild-type NOTCH genes (15).

We next examined the profile of NF-κB–related pathway genes in cSCC based on our hypothesis that Notch and other mediators could contribute to regulation of the NF-κB signaling network (20–23). The clustering data revealed increased gene expression of several key signal and transcription factor subunits for the NF-κB pathway in CD133+ cells (Fig. 2C). These include IL1R1, IRAK2/4, TRAF2/3, TRAF3IP2, CHUK (IKKalpha), IKBKB, REL, NFKB2, RELB, as well as AKT1/2, implicated in NF-κB activation by PI3K–Akt signaling. In addition, the heat map of the WNT pathway showed increase in WNT3, DVL2, and CTNNB1 and decrease in expression of the repressors such as AXIN2 and GSK3B, in CD133+ SCC cells (Supplementary Fig. S2D).

We hypothesized that the NOTCH, NF-κB, and WNT pathways could form interactive networks in cSCC CD133+ CSC populations and used Ingenuity Pathway Analysis (IPA) to reveal the potential regulatory relationships annotated among these three pathways. The results showed experimental evidence for altered expression and annotated links between NF-κB with the NOTCH (Fig. 2D) or WNT pathway (Fig. 2E). The network map suggests the hypothesis that noncanonical NOTCH1 signaling could contribute to the activation of REL and RELB in cSCC (Fig. 2D). The interaction of NF-κB with the WNT–β-catenin pathway is more indirectly through other nodes and pathways (Fig. 2E).

Independent validation of stem and pathway signature gene expression using Nanostring in a different set of cSCC tissue samples

To independently validate the microarray results, we verified gene expression levels with sorted CD133+ and CD133 cells from seven additional SCC tumors, using the Nanostring nCounter Gene Expression assay, for 149 selected genes from the stem- and top-ranked cancer-related pathway gene signatures. Notch signaling pathway components NOTCH1, NOTCH2, and MAML1 genes were confirmed to have higher expression levels, while JAG1 and EP300 showed decreased expression (Fig. 3A). The genes involved in the NF-κB pathway also showed upregulated expression, including IL1R1, IRAK2/4, CHUK, and RELB, as well as FAS, TLR6, PRKACA, MMP8, and ICAM4. In the WNT pathway, increased WNT3 and DVL2 expression and decreased GSK3B and DVL3 expression in CD133+ cells were independently verified using Nanostring.

Figure 3.

Validation of differentially expressed genes in the CD133+ CSC population using Nanostring expression assay. Totally 149 significantly altered genes involved in stem cell gene signature or signaling pathways were selected and validated using Nanostring nCounter gene expression assay. The consistently altered genes detected by both microarray and Nanostring are shown in NF-κB, NOTCH, WNT, and SHH pathways (A), EGFR/ERBB, FGF, MAPK, PI3K/AKT/MTOR, JAK/STAT, TP53/P73, and TGFβ pathways (B), and cell cycle, chromatin regulation, and other function categories (C). Blue bar, microarray; pink bar, Nanostring nCounter assay.

Figure 3.

Validation of differentially expressed genes in the CD133+ CSC population using Nanostring expression assay. Totally 149 significantly altered genes involved in stem cell gene signature or signaling pathways were selected and validated using Nanostring nCounter gene expression assay. The consistently altered genes detected by both microarray and Nanostring are shown in NF-κB, NOTCH, WNT, and SHH pathways (A), EGFR/ERBB, FGF, MAPK, PI3K/AKT/MTOR, JAK/STAT, TP53/P73, and TGFβ pathways (B), and cell cycle, chromatin regulation, and other function categories (C). Blue bar, microarray; pink bar, Nanostring nCounter assay.

Close modal

Furthermore, genes in several other signaling pathways displaying genomic or expression alterations and implicated in pathogenesis of SCC (8–10) were confirmed by Nanostring validation (Fig. 3B and C). In CD133+ cells, increased gene expression was observed in EGFR/ERBB, IGF1R, AKT, and mTOR and STAT family genes, as well as those implicated in FGF, MAPK, and TGF-β signaling pathways (Fig. 3B). Among other validated genes, cell cycle–related genes FBXW11 and PPP2R5C are overexpressed, while YAP1 and CTNNA1 related to cell growth and differentiation are decreased (Fig. 3C). Additionally, chromatin regulatory factors including DNMT1 and EZH2 genes were decreased, while HDAC6 is overexpressed. The differential expression of mRNA for several of these genes in CD133+ versus CD133 cells is not always congruent with the predicted effects of gene mutations (NOTCH1), copy-number alterations (YAP1), or overall expression (EZH2) observed in SCC subsets versus nonmalignant epithelia (7–9, 19, 20), underscoring the importance of functional validation.

Inhibiting NOTCH reduces the CD133+ cell population

As NOTCH signaling has been implicated in suppressing differentiation or promoting growth in different SCC (7, 15, 21), we first examined the functional effects of two different γ-secretase inhibitors (DAPT and RO4929097) that inhibit NOTCH, upon the CD133+ subset in the human cSCC cell line SCC13 (23). We confirmed that SCC13 exhibits a similar small percentage of CD133+ cells varying between 0.2% and 6%, depending on culture conditions, and displays clonogenic and sphere-forming potential, as observed by primary cSCC tumors. Three concentrations of DAPT (1, 5, 10 μmol/L) and RO4929097 (1, 3, 5 μmol/L) were selected based on ranges found to include IC50 for other tumor cell lines (24). FACS analysis shows that the CD133+ cell population is significantly inhibited by 5 to 10 μmol/L DAPT and ≥ 1 μmol/L RO4929097, supporting a functional role of NOTCH1 in the maintenance of CD133+ cells (Fig. 4A). We next tested if these two NOTCH inhibitors could affect CD133+ cell colony formation of SCC13, as well as UM-SCC-46, an HNSCC line verified to express wtNOTCH1 and pathway genes by whole exome and RNA sequencing (Hui Cheng, unpublished observations). Colony formation of both SCC13 and UM-SCC46 cells was significantly decreased by DAPT or RO4929097 (Fig. 4B; blue, red bars).

Figure 4.

Blocking NOTCH inhibits the CD133+ cell population and reduces NF-κB reporter activity and p65, p52, and RELB nuclear activation. A, Relative CD133+ cell population in human skin SCC13 cells was treated with two different NOTCH inhibitors, DAPT or R04929097, in different doses and measured by FACS analysis. B, The relative colony formation ratio in both SCC13 and UMSCC46 cells were examined after DAPT or R04929097 treatment. C, The specific NOTCH1 siRNA were transfected either alone or in combination into SCC13 or UMSCC 46 cells for 24 hours, and proteins after knockdown were shown by Western blot (top). The CD133+ cell population of SCC13 was examined by FACS. Nontargeting siRNA control was used as the control, and its CDl33+ population was set as 100% (bottom). D, NF-κB reporter activity of SCCl3 cells was measured with NOTCH inhibitors DAPT or R04929097 treatment for 24, 48, and 72 hours. NF-κB activity was stimulated with TNF-α (10 ng/mL) and normalized with β-gal. E, Silencing NOTCH 1, NOTCH2, or combined NOTCH1, and 2 siRNA partially decreases p65, p52, and RelB nuclear activation by binding assay. All results are shown as mean + SD; *, P < 0.05; **, P < 0.01; ***, P < 0.001. The experiments were done in triplicate.

Figure 4.

Blocking NOTCH inhibits the CD133+ cell population and reduces NF-κB reporter activity and p65, p52, and RELB nuclear activation. A, Relative CD133+ cell population in human skin SCC13 cells was treated with two different NOTCH inhibitors, DAPT or R04929097, in different doses and measured by FACS analysis. B, The relative colony formation ratio in both SCC13 and UMSCC46 cells were examined after DAPT or R04929097 treatment. C, The specific NOTCH1 siRNA were transfected either alone or in combination into SCC13 or UMSCC 46 cells for 24 hours, and proteins after knockdown were shown by Western blot (top). The CD133+ cell population of SCC13 was examined by FACS. Nontargeting siRNA control was used as the control, and its CDl33+ population was set as 100% (bottom). D, NF-κB reporter activity of SCCl3 cells was measured with NOTCH inhibitors DAPT or R04929097 treatment for 24, 48, and 72 hours. NF-κB activity was stimulated with TNF-α (10 ng/mL) and normalized with β-gal. E, Silencing NOTCH 1, NOTCH2, or combined NOTCH1, and 2 siRNA partially decreases p65, p52, and RelB nuclear activation by binding assay. All results are shown as mean + SD; *, P < 0.05; **, P < 0.01; ***, P < 0.001. The experiments were done in triplicate.

Close modal

To independently examine the specificity of these findings for NOTCH1 or 2 differentially expressed in tumor CD133+ cells in Fig. 3A, we examined the expression of NOTCH1 and 2 protein, and tested the effects of small interfering (si)RNAs silencing NOTCH1 and NOTCH2 on the CD133+ cell population in two cell lines. We confirmed that NOTCH1 protein of expected size is expressed and knocked down efficiently by siRNA in both SCC13 and UMSCC46 cells, while weak expression of NOTCH2 protein is detected only with prolonged exposure and weakly inhibited (Fig. 4C, top). Consistent with the differential expression and knockdown of NOTCH proteins, the relative CD133+ cell number decreased ∼40% of control with NOTCH1 knockdown alone, while no effect was attributable to the limited expression and knockdown of NOTCH2 alone or when combined with NOTCH1 siRNA (Fig. 4C, bottom). Together, the functional effects observed with pharmacologic and genetic approaches suggest NOTCH1 signaling contributes to the CD133+ population and colony-forming capability in a cSCC as well as an HNSCC cell line expressing wild-type NOTCH1.

Inhibiting NOTCH reduces NF-κB transcriptional and nuclear DNA-binding activity in cSCC cells

Because NOTCH1-mediated NF-κB activation is the second top-ranked GeneGo analysis signature for the CD133+ cell population, and aberrant activation of NF-κB has been observed in SCC (8, 25), we explored the possibility that NOTCH signaling enhances NF-κB pathway signaling, by assay for NF-κB reporter gene activity after NOTCH inhibition. As shown in Fig. 4D, inhibitors of γ-secretase and Notch activation, DAPT and RO4929097, decrease TNF-α–induced NF-κB reporter activity in human SCC13 cells in a dose-dependent manner. To further confirm the above results, siRNA was used to specifically knock down NOTCH1 and/or NOTCH2, and nuclear DNA binding of five NF-κB nuclear transcription factor subunits was assayed and quantified. The results showed that after transfection with NOTCH1 siRNA, the DNA-binding activity of canonical NF-κB pathway subunit p65 (RelA), and noncanonical pathway subunits p52, and RELB, were significantly decreased to 64%, 80%, and 77% respectively (P < 0.001) compared with the control siRNA group (Fig. 4E, left). NOTCH2 siRNA showed similar but weaker inhibitory effects on nuclear DNA-binding activity of these three transcription factor components, and combination of NOTCH2 with NOTCH1 siRNA did not further inhibit NF-κB nuclear binding activity (Fig. 4E, middle and right). These results provide evidence for NOTCH-MODULATED NF-κB pathway activity, which could potentially promote the CD133+ population in cSCC cells.

IKK–NF-κB pathway signaling promotes the CD133+ cell population in cSCC

As IKKα (CHUK) and other NF-κB pathway components were differentially overexpressed in CD133+ SCC tumor cells, we screened SCC13 cells for the effects of wedelolactone, an inhibitor shown to inhibit IKKα and β kinases, and IKKα-dependent NF-κB activation in SCC cells (25, 26). Wedelolactone inhibits TNF-α inducible NF-κB reporter activity assayed 24 and 48 hours after treatment in SCC13 cells (Fig. 5A). To determine the effects of NF-κB inhibition on SCC CD133+ cells, we analyzed the CD133+ cell population with or without wedelolactone treatment. The FACS results showed lower concentrations of 1 to 5 μmol/L wedelolactone reduce CD133+ cells, while 10 μmol/L wedelolactone significantly decreases the CD133+ population (Fig. 5B, P < 0.001). The effect of wedelolactone on the CD133+ population is not due to general cytotoxicity. The above results suggest the NF-κB pathway contributes to maintaining the cSCC CD133+ CSC population in vitro.

Figure 5.

Modulation of CD133+ CSCs by IKK inhibitor and IKKα expression, which colocalized with CD133 expression in cSCC tissues. A, IKK inhibitor wedelolactone inhibits NF-κB activity in SCC13 cells. Cells were transfected with NF-κB luciferase reporter and β-Gal plasmids and treated with wedelolactone and TNF-α alone or in combination for 24 and 48 hours. Red, no TNF-α; blue, TNF-α treatment. *, P < 0.05, compared with TNF-α-treated, non-wedelolactone group; #, P < 0.05, compared with non–TNF-α-treated group. B, Wedelolactone reduces CD133+ CSC population in SCC13 cells. Cells were treated with wedelolactone in different doses for 24 hours, stained with CD133-PE antibody, and the CD133+ population was analyzed with FACS. C, IKKα (CHUK) siRNA exhibited high knockdown efficiency in both cytoplasmic and nuclear compartment of SCC13 cells. D, IKKα (CHUK) knockdown significantly decreased CD133+ CSC population both in SCC13 and UMSCC46 cells. E, Overexpression IKKα (CHUK) active forms dramatically increases the CD133+ CSC population in SCC13 cells. F, Decreased CD133+ population caused by IKKα (CHUK) siRNA knockdown could be reversed by introducing WT IKKα or IKKα active forms (EE). Cells were cotransfected with different active forms of IKKα (WT, EE, AA, and KA) plus IKKα siRNA with control siRNA. CD133+ population was analyzed after 48 hours transfection. *, P < 0.05, compared with WT or EE control siRNA groups; #, P < 0.05, compared with vector control siRNA group. G, The frozen sections of cSCC tissues were coimmunostained with anti-IKKα and CD133 primary antibodies followed by Alexa-Fluor-488 (green) and Alexa-Fluor-555 (red) secondary antibodies. DAPI (blue) was used for visualizing nuclei, and H&E staining is showing morphology of tumor specimens. Magnification, 200×. All data represent mean and SD of one representative of three independent experiments, and each conducted in triplicate.

Figure 5.

Modulation of CD133+ CSCs by IKK inhibitor and IKKα expression, which colocalized with CD133 expression in cSCC tissues. A, IKK inhibitor wedelolactone inhibits NF-κB activity in SCC13 cells. Cells were transfected with NF-κB luciferase reporter and β-Gal plasmids and treated with wedelolactone and TNF-α alone or in combination for 24 and 48 hours. Red, no TNF-α; blue, TNF-α treatment. *, P < 0.05, compared with TNF-α-treated, non-wedelolactone group; #, P < 0.05, compared with non–TNF-α-treated group. B, Wedelolactone reduces CD133+ CSC population in SCC13 cells. Cells were treated with wedelolactone in different doses for 24 hours, stained with CD133-PE antibody, and the CD133+ population was analyzed with FACS. C, IKKα (CHUK) siRNA exhibited high knockdown efficiency in both cytoplasmic and nuclear compartment of SCC13 cells. D, IKKα (CHUK) knockdown significantly decreased CD133+ CSC population both in SCC13 and UMSCC46 cells. E, Overexpression IKKα (CHUK) active forms dramatically increases the CD133+ CSC population in SCC13 cells. F, Decreased CD133+ population caused by IKKα (CHUK) siRNA knockdown could be reversed by introducing WT IKKα or IKKα active forms (EE). Cells were cotransfected with different active forms of IKKα (WT, EE, AA, and KA) plus IKKα siRNA with control siRNA. CD133+ population was analyzed after 48 hours transfection. *, P < 0.05, compared with WT or EE control siRNA groups; #, P < 0.05, compared with vector control siRNA group. G, The frozen sections of cSCC tissues were coimmunostained with anti-IKKα and CD133 primary antibodies followed by Alexa-Fluor-488 (green) and Alexa-Fluor-555 (red) secondary antibodies. DAPI (blue) was used for visualizing nuclei, and H&E staining is showing morphology of tumor specimens. Magnification, 200×. All data represent mean and SD of one representative of three independent experiments, and each conducted in triplicate.

Close modal

Expression and activation of IKKα promote the CD133+ cell population in cSCC

As both our microarray and Nanostring verification data showed there is increased CHUK (IKKα) gene expression in CD133+ cells, and IKK inhibitor wedelolactone could decrease the CD133+ population in human SCC13 cells, we independently investigated the specific role and function of IKKα for the maintenance of CD133+ CSC in cSCC. To achieve this, we first tested the effect of siRNA knockdown of endogenous IKKα on the CD133+ cells. Knockdown by siRNA efficiently inhibited cytoplasmic and nuclear of IKKα (Fig. 5C) and reduced the CD133+ population by ∼50% compared with the control group in SCC13 cells (Fig. 5D). Similar effects were also observed in the HNSCC cell line UMSCC46. Next, we examined the effects of introducing IKKα wild-type (WT) or activated or inactivated genetic IKKα mutants upon the CD133+ population. These include IKKα with serine phospho-acceptor sites substituted with constitutively activating glutamate residues (EE); inactivating alanines (AA); or Lysine44 substituted with alanine, which produces a catalytically inactive form (KA; ref. 25). Cells expressing IKKα WT or EE protein significantly increased the CD133+ population 2.1- (P < 0.01) and 3.5-fold (P < 0.001) in SCC13 cells, whereas the inactive IKKα forms AA or KA exhibited no increase in the CD133+ subset (Fig. 5E). Furthermore, we confirmed that the increase in the CD133+ population by overexpression of wild-type IKKα (WT) or its active form (EE) can be significantly inhibited by IKKα siRNA knockdown (Fig. 5F).

To further examine the potential relationship between IKKα and CD133+ population in cSCC in vivo, we performed immunofluorescent (IF) staining of cSCC specimens. Figure 5G shows colocalization of IKKα and CD133+ immunostaining in cell subpopulations in two different primary cSCC tissues. The tumor specimens from 2 patients in Fig. 5G and Supplementary Fig. S3C and S3D show merged staining of CD133 and IKKα, as well as with DAPI, at the basal layer and leading edge of epithelial tumor cell nests shown in adjacent H&E sections. The costained population of cells is seen along the interface of malignant epithelia with intervening inflammatory (Supplementary Fig. S3C and 3D) or fibrovascular stroma (Fig. 5G). In addition, Supplementary Fig. S3A and 3B shows merged staining of CD133 and IKKα for cells in two additional SCC tumors, over a wider region at 100× magnification. These data are consistent with the differential coexpression of these detected in CD133+ cells from multiple tumors, and functional effects of IKKα genomic mutants on CD133+ cells in cell lines. Together, these observations support a role for increased IKKα expression in the CD133+ population in cSCC cells and tumors.

Knockdown of RELA and/or RELB inhibits the cSCC CD133+ population and its colony/spheroid formation phenotype

IKKα can mediate downstream canonical and alternative NF-κB pathway activation through RELA and RELB, which are found to be overexpressed in CD133+ cells. Next, we examined the knockdown effects of RELA and/or RELB on the CD133+ population in cSCC SCC13 and HNSCC UMSCC46 cells (Fig. 6A). The CD133+ cell populations were clearly reduced by 54% with RELA siRNA, 43% with RELB siRNA, and 43% with combined knockdown in SCC13 cells (Fig. 6B, blue bars). Similar results were observed in UMSCC46 cells (Fig. 6B, green bars). Next, we quantified the colony/spheroid formation ability in SCC13 and UMSCC46 cells after knockdown of RELA and/or RELB. Clonogenic colonies from CD133+ cells were reduced to 78%, 63%, and 55% by RELA or RELB siRNA knockdown alone, or in combination in SCC13 cells (Fig. 6C, pink bars). Spheroid formation in UMSCC46 cells also decreased to 76%, 69%, and 68%, respectively (Fig. 6C, purple bars; Fig. 6D). Together, the above data indicate both IKKα and NF-κB components RELA and RELB play an important role in maintaining the CD133+ population and its cancer stem–like features in human SCC.

Figure 6.

Knockdown NF-κB RelA and/or RelB inhibit CD133+ population and its colony/spheroid formation abilities, and the summary of NOTCH and NF-κB pathway activation. A, RELA and RELB siRNA exhibited high efficiency of siRNA knockdown. B, Knockdown of RELA and/or RELB inhibited CD133+ population. After 48 hours of transfection, the CD133+ population was analyzed by FACS. C, Knockdown of RELA or RELB inhibited colony and spheroid formation abilities. After transfection for 24 hours, cells were seeded for colony formation of SCC13 cells (left) and spheroid formation of UMSCC46 cells (right). Results were shown as mean ± SD, and data represent three independent experiments. *, P < 0.05; **, P < 0.01 and ***, P < 0.001. D, The images of UMSCC46 spheroids with different siRNA knockdown after 14 days of culture (40×). E, Schematic illustration of activated Notch enhances NF-κB pathway activation in cSCC CSC population. The CD133+ cSCC CSC gene signature is demonstrated with activated NOTCH signaling, which includes NOTCH1, NOTCH2, and JAG2 increased expressing. The activated NOTCH pathway then enhances CHUK(IKKα), one of the key upstream components of NF-κB signaling pathway. The active CHUK(IKKα) forms complexes with other IKKs and translocates to the nucleus, which further increased the NF-κB2 p52/RelB complexes activation and induces target gene expression. The increased NOTCH1 and NOTCH2 signal to the nucleus and regulate transcription factor MAML1 and RBPJ to control the downstream target gene expression.

Figure 6.

Knockdown NF-κB RelA and/or RelB inhibit CD133+ population and its colony/spheroid formation abilities, and the summary of NOTCH and NF-κB pathway activation. A, RELA and RELB siRNA exhibited high efficiency of siRNA knockdown. B, Knockdown of RELA and/or RELB inhibited CD133+ population. After 48 hours of transfection, the CD133+ population was analyzed by FACS. C, Knockdown of RELA or RELB inhibited colony and spheroid formation abilities. After transfection for 24 hours, cells were seeded for colony formation of SCC13 cells (left) and spheroid formation of UMSCC46 cells (right). Results were shown as mean ± SD, and data represent three independent experiments. *, P < 0.05; **, P < 0.01 and ***, P < 0.001. D, The images of UMSCC46 spheroids with different siRNA knockdown after 14 days of culture (40×). E, Schematic illustration of activated Notch enhances NF-κB pathway activation in cSCC CSC population. The CD133+ cSCC CSC gene signature is demonstrated with activated NOTCH signaling, which includes NOTCH1, NOTCH2, and JAG2 increased expressing. The activated NOTCH pathway then enhances CHUK(IKKα), one of the key upstream components of NF-κB signaling pathway. The active CHUK(IKKα) forms complexes with other IKKs and translocates to the nucleus, which further increased the NF-κB2 p52/RelB complexes activation and induces target gene expression. The increased NOTCH1 and NOTCH2 signal to the nucleus and regulate transcription factor MAML1 and RBPJ to control the downstream target gene expression.

Close modal

To our knowledge, this is the first study to demonstrate the CD133+ gene expression signature and its top-ranked signaling pathways in a purified cancer stem cell–like subset from primary cSCC tumors. There are four important findings from our research. First, our optimized isolation and separation further enriched for high purity CD133+ cytokeratin+ cells, overcoming a significant technical obstacle and enabling whole gene expression profiling detection and validation of stem and cancer genes in the small CD133+ population from clinical tumor samples. Second, our study provides a global overview of the gene signatures of CD133+ CSC in the human cSCC. Using microarray, Nanostring validation and bioinformatics analyses, these data provide a valuable resource for hypothesis generation and testing for functional and clinical role of different signaling components and pathways in cSCC CD133+ CSC. Our stem gene expression profiling data identified ∼2,000 candidate genes differentially expressed in CD133+ cSCC cells, including an 80-gene set of known stem cell signatures. Third, we demonstrated upregulated expression of genes of key signaling pathways and networks from the CSC population by analyzing the top-ranked pathways involved with GeneGO and IPA platform for the CD133+ transcript signature. The CD133+ transcript signatures represent pathways implicated in self-renewal and epithelial differentiation, such as NOTCH, WNT, SHH, and NF-κB, and cancer-related pathways PTEN, PI3K/AKT/mTOR, NGF, MAPK, EGFR/ERBB, FGF, JAK/STAT, TP53, and TGF-β. Genomics and experimental studies provide evidence supporting the biologic and clinical significance of several of these pathways to SCC biology (7, 8). Complementing our observations, recent whole exome sequencing studies of cSCC and HNSCC provide evidence for gene copy alterations or mutations in components or coregulators of the NOTCH (NOTCH 1,2), WNT (AJUBA), cancer-related MAPK and PI3K growth (HRAS, RASA1, BRAF, PIK3CA), NF-κB prosurvival and death pathways (FADD, BIRC2/3, CASP8, RIPK4). Lastly, our findings revealed a link between the NOTCH and NF-κB signal pathways and expression of several of their key components, NOTCH1, IKKα, RELA, and RELB, contributing to the maintenance of the CD133+ CSC population and its stem-like phenotypic features (Fig. 6E).

We uncovered a NOTCH cSCC CSC gene signature and showed that NOTCH plays a role in maintaining the CD133+ cell population. Blocking NOTCH by either of two different γ-secretase inhibitors, or NOTCH1 by siRNA, reduced the percentage of CD133+ SCC cells, and their capacity for clonogenic sphere formation, which are established surface and phenotypic markers of CSC/TIC (3). The NOTCH pathway plays a key role in cell–cell communication via interaction of membrane-bound JAG ligands and NOTCH receptors (27). The NOTCH receptor family is composed of 4 family members (27). In HNSCC and cSCC, inactivating and missense mutations in NOTCH1 and 2 have been observed in a subset of tumors, suggesting that the NOTCH may function as a tumor suppressor in those contexts (8, 9, 28, 29). Complexes of NOTCH receptor intracellular domains (NICD) with canonical NOTCH transcriptional cofactors can activate HES1/HEY1 repressors that trigger epithelial differentiation. Intriguingly, differential expression of HES1/HEY1 was not detected among CD133+ cSCC overexpressing NOTCH1 and other signal components, and there was decreased expression of key canonical transcription factor RBPJ. These findings are consistent with recent studies in human HNSCC, where tumors expressing WT NOTCH1 included subsets without evidence of HES1/HEY1 activation, and where NOTCH inhibitors reduced proliferation, through an unknown mechanism (15).

In this regard, we provide evidence that NOTCH can promote noncanonical cross-activation of nuclear factor-κB (NF-κB/REL), a family of signal activated transcription factors that are often aberrantly activated in cSCC, HNSCC, and cervical SCC (25, 30–33). The activation of canonical NF-κB1/RELA and alternate NF-κB2/RELB transcription factors is mediated by inhibitor-kappaB kinase (IKK)α/β/γ or IKKα complexes, respectively (25). Our profiling of CD133+ cSCC cells unveiled increased expression of multiple components of the NOTCH and NF-κB signaling pathways, including key components NOTCH1, IKKα (CHUK), RELA, and RELB (Fig. 6E). We showed that inhibition of NOTCH by γ-secretase or NOTCH1 by siRNA functionally partially inhibits NF-κB reporter and nuclear DNA binding of NF-κB p52, RELA and RELB along with the CD133+phenotype, and that knocking down overexpressed IKKα, RELA, and RELB similarly inhibited the CD133+ phenotype. The partial reduction of NF-κB activity by inhibition of NOTCH signaling is expected, because NOTCH are not the only molecules that directly modulate NF-κB activity in SCC (8, 25).

Interestingly, our study also establishes that IKKα is an important regulator for maintaining the CD133+ CSC population in cSCC cells. Notably, NOTCH1 and IKKα have previously been shown to physically and functionally interact in normal keratinocytes and cervical SCC cells (33, 34). NOTCH1 knockdown modulated TNFα-induced NF-κB activity and promoter binding in cervical SCC cells (33) similar to our findings for Notch γ-secretase pharmacologic inhibition of TNF-induced NF-κB activity in cSCC cells. In cervical SCC, NOTCH1 and IKKα also promoted chemotherapeutic resistance, a feature often attributed to CSC. Our data indicate that WT or activating phospho-acceptor site mutants for TNF-mediated IKKα activation increase CD133+ cells, whereas inactivated IKKα mutants did not, implicating the role of IKKα in the expansion of CD133+ CSC. Further supporting this, IKKα colocalized in CD133+ cells in malignant epithelia of cSCC tumors. Besides its role in NF-κB activation, IKKα may also function as a cofactor to regulate other key transcription factor genes that control cell proliferation and differentiation, such as E2F1/BMI1 (35). Intriguingly, IKKα has also been reported to bind the promoter and reciprocally repress Notch canonical target gene HES1 (36), potentially helping to explain the lack of HES1 induction among NOTCH-related genes we detected, and the paradoxical role of NOTCH1 and IKKα in promoting CSC expansion without triggering NOTCH-mediated terminal differentiation.

Although the involvement of the IKK–NF-κB signaling pathway in promoting SCC has been demonstrated (25, 30–33), its role in the maintenance of a CD133+ CSC population in cSCC has not previously been reported. Our gene expression profiling and functional analysis reveals the important role of increased NF-κB signaling pathway and key transcription factors RELA and RELB in promoting CD133+ CSCs in cSCC. Our finding that elevated NF-κB signaling is important to maintain CSCs is supported by microarray expression profiling in human embryonic stem cell (ESC), which demonstrates that RELA is essential to maintain ESC pluripotency (37).

The evidence for altered expression of components of the PI3K/AKT/mTOR, MAPK, FGF, ERBB, and STAT signaling or activating pathways is consistent with evidence for frequent genomic alterations affecting these pathways in HNSCC and cSCC (38). These pathways are also important for controlling the self-renewal and growth of CSCs (39–43). PI3K and MAPK signaling is implicated in SCC growth and CSC (44–48). The regulatory role of the STAT pathway on CSCs self-renewal was reported in HNSCC (49). Hence, our profiling data suggest that multiple signaling pathways may contribute to the stemness signature in cSCC. Consistent with this, we observed only partial inhibition of CSC phenotypes when blocking NOTCH and IKK–NF-KB signal. Thus, combination therapy inhibiting CSCs may be more effective, because pharmacologic inhibition of one pathway may only partially reduce the CD133+ cell population.

Several stemness makers have been studied in SCC and other cancers, including CD44, ALDH1, and CD133 (3, 4, 5, 50). Among these, CD44 and ALDH1 are expressed in relatively broader populations in tumor and cell lines, and additional markers or criteria are often used to enrich and characterize the CSC subpopulations. In our prior (3) and current study, the CD133+ population showed relatively lower expression of CD44, and <1.5-fold difference in ALDH1 compared with the CD133 population. As ALDH1, CD44, and CD133 populations do not always overlap, they may contain distinct stem cell subpopulations. Our prior and current study of skin SCC tumors established that the relatively rare CD133+ population is a biologically and therapeutically relevant subset enriched for TIC and sphere-forming stem phenotypes. While our studies did not preclude that there are CSC-bearing overlapping or distinct markers, this study explores and experimentally validates a role for related NOTCH and IKK–NF-κB signaling in the CD133+ population CSC phenotype. This study provides information about gene expression patterns and signaling pathways for this CSC population in human primary cSCC, which could help identify potential drug targets for CSC to complement current therapies for cSCC.

No potential conflicts of interest were disclosed.

Conception and design: X.X. Quan, Z. Chen, C. Van Waes

Development of methodology: X.X. Quan, N.V. Hawk, Z. Chen

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): X.X. Quan, P.S. Meltzer, A. Montemarano, M. Braun, Z. Chen

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): X.X. Quan, W. Chen, S.K. Lee, P.S. Meltzer, Z. Chen, C. Van Waes

Writing, review, and/or revision of the manuscript: X.X. Quan, Z. Chen, C. Van Waes

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): X.X. Quan, S.K. Lee, D.W. Petersen, Z. Chen

Study supervision: X.X. Quan, Z. Chen, C. Van Waes

Other (ran experiments, lab technician): J. Coupar

This study has been supported by the Intramural Research Program of the National Cancer Institute and the National Institute on Deafness and Other Communication Disorders (X. Quan, J. Coupar, S.K. Lee, Z. Chen, and C. Van Waes were supported by intramural projects ZIA-DC-000016, ZIA-DC-000073, and ZIA-DC-000074, which were awarded to C. Van Waes). We especially thank Dr. Jonathan C. Vogel, who is deceased, for his guidance in initiating this project, helpful discussions and comments on the purification of the CD133+ cell subpopulation, and the design and setup of microarray experiment. We thank Dr. Mark Udey in the Dermatology Branch, NCI, NIH, Bethesda, Maryland, for many comments and strong support for the project. We thank Dr. Thomas Hornyak, in the Dermatology Department, University of Maryland, School of Medicine, Baltimore, Maryland, for his comments and support. We thank Dr. William G. Telford, in the Experimental Transplantation and Immunology Branch, NCI, NIH, Bethesda, for his advice on cell flow cytometry experiments. This study utilized the high-performance computational capabilities of the Biowulf Linux cluster at the NIH, Bethesda, Maryland (https://hpc.nih.gov/systems/).

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.
Kauvar
AN
,
Arpey
CJ
,
Hruza
G
,
Olbricht
SM
,
Bennett
R
. 
Consensus for nonmelanoma skin cancer treatment, part II: squamous cell carcinoma, including a cost analysis of treatment methods
.
Dermatol Surg
2015
;
41
:
1214
40
.
2.
Phillips
RL
,
Ernst
RE
,
Brunk
B
,
Ivanova
N
,
Mahan
MA
,
Deanehan
JK
, et al
The genetic program of hematopoietic stem cells
.
Science
2000
;
288
:
1635
40
.
3.
D'Angelo
RC
,
Wicha
MS
. 
Stem cells in normal development and cancer
.
Prog Mol Biol Transl Sci
2010
;
95
:
113
58
.
4.
Patel
GK
,
Yee
CL
,
Terunuma
A
,
Telford
WG
,
Voong
N
,
Yuspa
SH
, et al
Identification and characterization of tumor-initiating cells in human primary cutaneous squamous cell carcinoma
.
J Invest Dermatol
2012
;
132
:
401
9
.
5.
Zhang
Q
,
Shi
S
,
Yen
Y
,
Brown
J
,
Ta
JQ
,
Le
AD
. 
A subpopulation of CD133(+) cancer stem-like cells characterized in human oral squamous cell carcinoma confer resistance to chemotherapy
.
Cancer Lett
2010
;
289
:
151
60
.
6.
Eramo
A
,
Lotti
F
,
Sette
G
,
Pilozzi
E
,
Biffoni
M
,
Di Virgilio
A
, et al
Identification and expansion of the tumorigenic lung cancer stem cell population
.
Cell Death Differ
2008
;
15
:
504
14
.
7.
Durinck
S
,
Ho
C
,
Wang
NJ
,
Liao
W
,
Jakkula
LR
,
Collisson
EA
, et al
Temporal dissection of tumorigenesis in primary cancers
.
Cancer Discov
2011
;
1
:
137
43
.
8.
Pickering
CR
,
Zhou
JH
,
Lee
JJ
,
Drummond
JA
,
Peng
SA
,
Saade
RE
, et al
Mutational landscape of aggressive cutaneous squamous cell carcinoma
.
Clin Cancer Res
2014
;
20
:
6582
92
.
9.
Cancer Genome Atlas Network
. 
Comprehensive genomic characterization of head and neck squamous cell carcinomas
.
Nature
2015
;
517
:
576
82
.
10.
Hoadley
KA
,
Yau
C
,
Wolf
DM
,
Cherniack
AD
,
Tamborero
D
,
Ng
S
, et al
Multiplatform analysis of 12 cancer types reveals molecular classification within and across tissues of origin
.
Cell
2014
;
158
:
929
44
.
11.
Ivanova
NB
,
Dimos
JT
,
Schaniel
C
,
Hackney
JA
,
Moore
KA
,
Lemischka
IR
. 
A stem cell molecular signature
.
Science
2002
;
298
:
601
4
.
12.
Hofmann
WK
,
de Vos
S
,
Komor
M
,
Hoelzer
D
,
Wachsman
W
,
Koeffler
HP
. 
Characterization of gene expression of CD34+ cells from normal and myelodysplastic bone marrow
.
Blood
2002
;
100
:
3553
60
.
13.
Sperger
JM
,
Chen
X
,
Draper
JS
,
Antosiewicz
JE
,
Chon
CH
,
Jones
SB
, et al
Gene expression patterns in human embryonic stem cells and human pluripotent germ cell tumors
.
Proc Natl Acad Sci U S A
2003
;
100
:
13350
5
.
14.
Ramalho-Santos
M
,
Yoon
S
,
Matsuzaki
Y
,
Mulligan
RC
,
Melton
DA
. 
"Stemness": transcriptional profiling of embryonic and adult stem cells
.
Science
2002
;
298
:
597
600
.
15.
Sun
W
,
Gaykalova
DA
,
Ochs
MF
,
Mambo
E
,
Arnaoutakis
D
,
Liu
Y
, et al
Activation of the NOTCH pathway in head and neck cancer
.
Cancer Res
2014
;
74
:
1091
104
.
16.
Radtke
S
,
Gorgens
A
,
Kordelas
L
,
Schmidt
M
,
Kimmig
KR
,
Koninger
A
, et al
CD133 allows elaborated discrimination and quantification of haematopoietic progenitor subsets in human haematopoietic stem cell transplants
.
Br J Haematol
2015
;
169
:
868
78
.
17.
Forget
MA
,
Huon
Y
,
Reuben
A
,
Grange
C
,
Liberman
M
,
Martin
J
, et al
Stimulation of Wnt/ss-catenin pathway in human CD8+ T lymphocytes from blood and lung tumors leads to a shared young/memory phenotype
.
PLoS One
2012
;
7
:
e41074
.
18.
Lai
C
,
August
S
,
Albibas
A
,
Behar
R
,
Cho
SY
,
Polak
ME
, et al
OX40+ regulatory T cells in cutaneous squamous cell carcinoma suppress effector T-cell responses and associate with metastatic potential
.
Clin Cancer Res
2016
;
22
:
4236
48
.
19.
Freeman
A
,
Bridge
JA
,
Maruthayanar
P
,
Overgaard
NH
,
Jung
JW
,
Simpson
F
, et al
Comparative immune phenotypic analysis of cutaneous squamous cell carcinoma and intraepidermal carcinoma in immune-competent individuals: proportional representation of CD8+ T-cells but not FoxP3+ Regulatory T-cells is associated with disease stage
.
PLoS One
2014
;
9
:
e110928
.
20.
Xie
Q
,
Wang
H
,
Heilman
ER
,
Walsh
MG
,
Haseeb
MA
,
Gupta
R
. 
Increased expression of enhancer of Zeste Homolog 2 (EZH2) differentiates squamous cell carcinoma from normal skin and actinic keratosis
.
Eur J Dermatol
2014
;
24
:
41
5
.
21.
Adhikary
G
,
Grun
D
,
Balasubramanian
S
,
Kerr
C
,
Huang
JM
,
Eckert
RL
. 
Survival of skin cancer stem cells requires the Ezh2 polycomb group protein
.
Carcinogenesis
2015
;
36
:
800
10
.
22.
Egloff
AM
,
Grandis
JR
. 
Molecular pathways: context-dependent approaches to Notch targeting as cancer therapy
.
Clin Cancer Res
2012
;
18
:
5188
95
.
23.
Rheinwald
JG
,
Beckett
MA
. 
Tumorigenic keratinocyte lines requiring anchorage and fibroblast support cultured from human squamous cell carcinomas
.
Cancer Res
. 
1981
;
41
:
1657
63
.
24.
Saito
N
,
Fu
J
,
Zheng
S
,
Yao
J
,
Wang
S
,
Liu
DD
, et al
A high Notch pathway activation predicts response to gamma secretase inhibitors in proneural subtype of glioma tumor-initiating cells
.
Stem cells
2014
;
32
:
301
12
.
25.
Nottingham
LK
,
Yan
CH
,
Yang
X
,
Si
H
,
Coupar
J
,
Bian
Y
, et al
Aberrant IKKalpha and IKKbeta cooperatively activate NF-kappaB and induce EGFR/AP1 signaling to promote survival and migration of head and neck cancer
.
Oncogene
2014
;
33
:
1135
47
.
26.
Kobori
M
,
Yang
Z
,
Gong
D
,
Heissmeyer
V
,
Zhu
H
,
Jung
YK
, et al
Wedelolactone suppresses LPS-induced caspase-11 expression by directly inhibiting the IKK complex
.
Cell Death Differ
2004
;
11
:
123
30
.
27.
Katoh
M
,
Katoh
M
. 
Notch signaling in gastrointestinal tract (review)
.
Int J Oncol
2007
;
30
:
247
51
.
28.
Wang
WM
,
Zhao
ZL
,
Ma
SR
,
Yu
GT
,
Liu
B
,
Zhang
L
, et al
Epidermal growth factor receptor inhibition reduces angiogenesis via hypoxia-inducible factor-1alpha and Notch1 in head neck squamous cell carcinoma
.
PLoS One
2015
;
10
:
e0119723
.
29.
Martin
A
,
Salvador
F
,
Moreno-Bueno
G
,
Floristan
A
,
Ruiz-Herguido
C
,
Cuevas
EP
, et al
Lysyl oxidase-like 2 represses Notch1 expression in the skin to promote squamous cell carcinoma progression
.
EMBO J
2015
;
34
:
1090
109
.
30.
Duffey
DC
,
Chen
Z
,
Dong
G
,
Ondrey
FG
,
Wolf
JS
,
Brown
K
, et al
Expression of a dominant-negative mutant inhibitor-kappaBalpha of nuclear factor-kappaB in human head and neck squamous cell carcinoma inhibits survival, proinflammatory cytokine expression, and tumor growth in vivo
.
Cancer Res
1999
;
59
:
3468
74
.
31.
Kobielak
A
,
Fuchs
E
. 
Links between alpha-catenin, NF-kappaB, and squamous cell carcinoma in skin
.
Proc Natl Acad Sci U S A
2006
;
103
:
2322
7
.
32.
Lun
M
,
Zhang
PL
,
Pellitteri
PK
,
Law
A
,
Kennedy
TL
,
Brown
RE
. 
Nuclear factor-kappaB pathway as a therapeutic target in head and neck squamous cell carcinoma: pharmaceutical and molecular validation in human cell lines using Velcade and siRNA/NF-kappaB
.
Ann Clin Lab Sci
2005
;
35
:
251
8
.
33.
Song
LL
,
Peng
Y
,
Yun
J
,
Rizzo
P
,
Chaturvedi
V
,
Weijzen
S
, et al
Notch-1 associates with IKKalpha and regulates IKK activity in cervical cancer cells
.
Oncogene
2008
;
27
:
5833
44
.
34.
Yugawa
T
,
Narisawa-Saito
M
,
Yoshimatsu
Y
,
Haga
K
,
Ohno
S
,
Egawa
N
, et al
DeltaNp63alpha repression of the Notch1 gene supports the proliferative capacity of normal human keratinocytes and cervical cancer cells
.
Cancer Res
2010
;
70
:
4034
44
.
35.
Ammirante
M
,
Kuraishy
AI
,
Shalapour
S
,
Strasner
A
,
Ramirez-Sanchez
C
,
Zhang
W
, et al
An IKKalpha-E2F1-BMI1 cascade activated by infiltrating B cells controls prostate regeneration and tumor recurrence
.
Genes Dev
2013
;
27
:
1435
40
.
36.
Aguilera
C
,
Hoya-Arias
R
,
Haegeman
G
,
Espinosa
L
,
Bigas
A
. 
Recruitment of IkappaBalpha to the hes1 promoter is associated with transcriptional repression
.
Proc Natl Acad Sci U S A
2004
;
101
:
16537
42
.
37.
Armstrong
L
,
Hughes
O
,
Yung
S
,
Hyslop
L
,
Stewart
R
,
Wappler
I
, et al
The role of PI3K/AKT, MAPK/ERK and NFkappabeta signalling in the maintenance of human embryonic stem cell pluripotency and viability highlighted by transcriptional profiling and functional analysis
.
Hum Mol Genet
2006
;
15
:
1894
913
.
38.
Li
YY
,
Hanna
GJ
,
Laga
AC
,
Haddad
RI
,
Lorch
JH
,
Hammerman
PS
. 
Genomic analysis of metastatic cutaneous squamous cell carcinoma
.
Clin Cancer Res
2015
;
21
:
1447
56
.
39.
Pattabiraman
DR
,
Weinberg
RA
. 
Tackling the cancer stem cells - what challenges do they pose? Nat Rev Drug Discov
2014
;
13
:
497
512
.
40.
Vermeulen
L
,
Snippert
HJ
. 
Stem cell dynamics in homeostasis and cancer of the intestine
.
Nat Rev Cancer
2014
;
14
:
468
80
.
41.
Schneider
MR
,
Yarden
Y
. 
The EGFR-HER2 module: a stem cell approach to understanding a prime target and driver of solid tumors
.
Oncogene
2016
;
35
:
2949
60
.
42.
Zhou
BB
,
Zhang
H
,
Damelin
M
,
Geles
KG
,
Grindley
JC
,
Dirks
PB
. 
Tumour-initiating cells: challenges and opportunities for anticancer drug discovery
.
Nat Rev Drug Discov
2009
;
8
:
806
23
.
43.
Pardal
R
,
Clarke
MF
,
Morrison
SJ
. 
Applying the principles of stem-cell biology to cancer
.
Nat Rev Cancer
2003
;
3
:
895
902
.
44.
Li
Q
,
Rycaj
K
,
Chen
X
,
Tang
DG
. 
Cancer stem cells and cell size: a causal link?
Semin Cancer Biol
2015
;
35
:
191
9
.
45.
Xia
P
,
Xu
XY
. 
PI3K/Akt/mTOR signaling pathway in cancer stem cells: from basic research to clinical application
.
Am J Cancer Res
2015
;
5
:
1602
9
.
46.
Mohan
S
,
Vander Broek
R
,
Shah
S
,
Eytan
DF
,
Pierce
ML
,
Carlson
SG
, et al
MEK inhibitor PD-0325901 overcomes resistance to PI3K/mTOR inhibitor PF-5212384 and potentiates antitumor effects in human head and neck squamous cell carcinoma
.
Clin Cancer Res
2015
;
21
:
3946
56
.
47.
Li
H
,
Gao
Q
,
Guo
L
,
Lu
SH
. 
The PTEN/PI3K/Akt pathway regulates stem-like cells in primary esophageal carcinoma cells
.
Cancer Biol Ther
2011
;
11
:
950
8
.
48.
Sun
S
,
Liu
S
,
Duan
SZ
,
Zhang
L
,
Zhou
H
,
Hu
Y
, et al
Targeting the c-Met/FZD8 signaling axis eliminates patient-derived cancer stem-like cells in head and neck squamous carcinomas
.
Cancer Res
2014
;
74
:
7546
59
.
49.
Bourguignon
LY
,
Earle
C
,
Wong
G
,
Spevak
CC
,
Krueger
K
. 
Stem cell marker (Nanog) and Stat-3 signaling promote MicroRNA-21 expression and chemoresistance in hyaluronan/CD44-activated head and neck squamous cell carcinoma cells
.
Oncogene
2012
;
31
:
149
60
.
50.
Pan
Q
,
Li
Q
,
Liu
S
,
Ning
N
,
Zhang
X
,
Xu
Y
, et al
Concise review: targeting cancer stem cells using immunologic approaches
.
Stem cells
2015
;
33
:
2085
92
.