Cancer Stem Cell Subpopulations Are Present Within Metastatic Head and Neck Cutaneous Squamous Cell Carcinoma

Cancer stem cells (CSCs) have been identified in many cancer types including primary head and neck cutaneous squamous cell carcinoma (HNcSCC). This study aimed to identify and characterize CSCs in metastatic HNcSCC (mHNcSCC). Immunohistochemical staining performed on mHNcSCC samples from 15 patients demonstrated expression of the induced pluripotent stem cell (iPSC) markers OCT4, SOX2, NANOG, KLF4, and c-MYC in all 15 samples. In situ hybridization and RT-qPCR performed on four of these mHNcSCC tissue samples confirmed transcript expression of all five iPSC markers. Immunofluorescence staining performed on three of these mHNcSCC samples demonstrated expression of c-MYC on cells within the tumor nests (TNs) and the peri-tumoral stroma (PTS) that also expressed KLF4. OCT4 was expressed on the SOX2+/NANOG+/KLF4+ cells within the TNs, and the SOX2+/NANOG+/KLF4+ cells within the PTS. RT-qPCR demonstrated transcript expression of all five iPSC markers in all three mHNcSCC-derived primary cell lines, except for SOX2 in one cell line. Western blotting showed the presence of SOX2, KLF4, and c-MYC but not OCT4 and NANOG in the three mHNcSCC-derived primary cell lines. All three cell lines formed tumorspheres, at the first passage. We demonstrated an OCT4+/NANOG+/SOX2+/KLF4+/c-MYC+ CSC subpopulation and an OCT4+/NANOG-/SOX2+/KLF4+/c-MYC+ subpopulation within the TNs, and an OCT4+/NANOG+/SOX2+/KLF4+/c-MYC+ subpopulation within the PTS of mHNcSCC.


INTRODUCTION
The incidence of non-melanoma skin cancer (NMSC) is rising rapidly globally, particularly amongst Anglo-Celtic populations (1). Cutaneous squamous cell carcinoma (cSCC) is the second most common form of NMSC, affecting 118/100,000 people in New Zealand (1). Caucasian descent, pale complexion and advancing age increase the risk of developing cSCC (2). Tumor thickness, horizontal diameter, perineural and/or lymphovascular invasion, poor histological differentiation and certain anatomic sites increase the metastatic risk (1,(3)(4)(5).
Sixty percent of cSCC occur in the head and neck with a 2% incidence of metastasis, mainly to the parotid and/or neck nodes (4), making metastatic cSCC the most common parotid malignancy in New Zealand and Australia (6). Treatment for metastatic head and neck cSCC (mHNcSCC) is surgery and post-operative adjuvant radiotherapy with a 48% 5-year overall survival (2). This poor outcome has been attributed to the presence of cancer stem cells (CSCs) (7).
The CSC concept proposes that the development and progression of cancer are driven by CSCs (8), a small population of highly tumorigenic cells imbued with embryonic stem cell (ESC) properties (9), such as self-renewal (10), and pluripotency (9). CSCs divide asymmetrically to produce identical CSCs, as well as differentiated cancer cells that possess little or no tumorigenicity and form the bulk of the tumor (11,12).
Yamanaka et al. (13,14) have demonstrated induction of pluripotent stem cells from adult mouse and human fibroblasts by the introduction of the ESC markers OCT4, SOX2, c-MYC, and KLF4. Thomson et al. (15) have demonstrated that this can also be achieved with NANOG and LIN28 in place of c-MYC and KLF4. These studies underscore the sufficiency of these four transcription factors (13) to generate induced pluripotent stem cells (iPSCs). CSCs in several cancer types (16)(17)(18) including primary head and neck cSCC (HNcSCC) (19) have been shown to express these master regulators of pluripotency.
SOX2, a member of the SRY-related HMG-box (SOX) gene family of transcription factors that maintains developmental potential (20), is a marker of CSCs in cSCC (21). SOX2 expression is critical for tumor initiation and growth, and regulates selfrenewal and long-term growth of CSCs (21)(22)(23). OCT4, a POU domain transcription factor (24), forms the "core pluripotency network" with SOX2 and NANOG to regulate ESC pluripotency (25) and stemness of CSCs in HNcSCC (26). Its expression correlates with poor overall survival of patients with oral cavity SCC (OCSCC) (27). NANOG, a homeodomain containing transcription factor (28), also correlates with worsened survival of OCSCC (27). NANOG induces CSC characteristics in OCSCC (27) and colorectal cancer (29). KLF4, a zinc-finger transcription factor and part of the Krüppel-like family, plays a central role in cell cycle regulation, maintenance of pluripotency and somatic cell reprogramming (30). c-MYC, an oncoprotein within the MYC family, is involved in cell growth, differentiation, apoptosis, angiogenesis, and stem cell pluripotency and proliferation (31,32). MYC dysregulation occurs in various cancer types (33)(34)(35) and is associated with more aggressive tumors (36).
We have recently demonstrated the presence of CSCs in primary HNcSCC that express the iPSC markers OCT4, SOX2, NANOG, KLF4, and c-MYC (19). This study aimed to identify and characterize CSCs within mHNcSCC using these markers.

MATERIALS AND METHODS mHNcSCC Tissue Samples
Formalin-fixed paraffin-embedded (FFPE) sections of mHNcSCC tissue samples from 15 male patients, aged 53-92 (mean, 80) years ( Table 1) were sourced from the Gillies McIndoe Research Institute Tissue Bank for this study, which was approved by the Central Regional Health and Disability Ethics Committee (Ref. 12/CEN/74AM05). The mHNcSCC tissue samples used were from a separate cohort of patients than those used in our previous study that investigated CSCs in primary HNcSCC (19). Written informed consent was obtained from all patients.
Three mHNcSCC tissue samples from the original cohort of 15 patients underwent immunofluorescence (IF) staining to determine co-expression of proteins. VectaFluor Excel antimouse 488 (ready-to-use; cat#VEDK2488, Vector Laboratories, Burlingame, CA, USA) and Alexa Fluor anti-rabbit 594 (1:500; cat#A21207, Life Technologies, Carlsbad, CA, USA) were used to determine marker expression. All IF-stained slides were mounted in Vecta Shield Hardset mounting medium with 4 ′ ,6diamidino-2-phenylindole (Vector Laboratories). All antibodies were diluted in BOND TM primary diluent (Leica). All IHC and IF staining was performed on the Leica BOND RX TM auto-stainer (Leica, Nussloch, Germany).
Human tissues used for positive controls were seminoma for OCT4 and NANOG, skin for SOX2, breast carcinoma for KLF4, and colon for c-MYC. To determine the specificity of the amplification cascade used in IHC staining, negative controls were performed on sections of mHNcSCC tissue samples using a matched isotype control for both mouse (ready-to-use; cat#DK2488, Dako, Glostrup, Denmark) and rabbit (ready-touse; cat#DK1594, Dako) primary antibodies, and a combination of both was used for determining specificity of the amplification cascade used in IF staining.
Human tissues used for positive controls were seminoma for OCT4 and NANOG, skin for SOX2, breast carcinoma for KLF4, and colon for c-MYC. Negative controls were demonstrated on sections of mHNcSCC tissue samples using a probe for DapB (EF191515) (cat#312038, Advanced Cell Diagnostics).

Image Analysis and Quantification of IHC and ISH Staining
IHC-stained slides were visualized and imaged using an Olympus BX53 light microscope fitted with an Olympus SC100 digital camera (Olympus, Tokyo, Japan), and processed with the cellSens 2.0 Software (Olympus). IF-stained slides were viewed and imaged with an Olympus FV1200 biological confocal laserscanning microscope and subjected to 2D deconvolutional processing with cellSens Dimension 1.11 software (Olympus).
Cell counting was performed on IHC-stained and ISH-stained slides of mHNcSCC tissue samples using Cell Counter on ImageJ software (National Institutes of Health, Bethesda, MD, USA). Cell counting of IHC-stained slides was performed on three fields of view at 400x magnification, with each field including both the tumor nests (TNs) and the peri-tumoral stroma (PTS) at ∼50% of each image. A cell was considered positive for staining if it resembled the positive control for that marker, and was deemed negative for staining if it did not. A cell was deemed positively stained for OCT4, SOX2, NANOG, KLF4, and c-MYC if staining was present in either the nucleus or cytoplasm, and cells were distinguished from one another by the presence of their nuclei and counted. All positively stained cells in the TNs and the PTS for each field were counted and the proportions of positively stained cells out of the total number of cells within the field of view were then calculated and averaged across the three fields of view that had a minimum of 100 cells per field, for each of the 15 cases. Cell counting on ISH-stained slides was performed in the same manner, except the images were taken at 1000x magnification with each view having a minimum of 10 cells, for each of the six cases.

Reverse-Transcription Quantitative Polymerase Chain Reaction
Total RNA was isolated from four snap-frozen mHNcSCC tissue samples (20 mg/sample) and three mHNcSCC-derived primary cell lines (5 × 10 5 viable cells/sample) from the original cohort of 15 patients. Tissues were homogenized using the Omni Tissue Homogenizer (Omni TH, Omni International, Kennesaw, GA, USA) before preparation using the RNeasy Mini Kit (cat#74104, Qiagen). Frozen cell pellets were prepared using the RNeasy Micro Kit (cat#74004, Qiagen). A DNase digest step was included for both methods (cat#79254, Qiagen). Quantitation of the RNA was determined using a NanoDrop 2000 Spectrophotometer (Thermo Fisher Scientific). Transcript expression was determined using the Rotor-Gene Q (Qiagen) and the Rotor-Gene Multiplex RT-qPCR Kit (cat#204974, Qiagen). The TaqMan primer probes used were OCT4 (Hs03005111_g1), SOX2 (Hs01053049_s1), NANOG (Hs02387400_g1), KLF4 (Hs00358836_m1), and c-MYC (Hs00153408_m1; cat#4331182). The level of gene expression was normalized to that of the housekeepers GAPDH (Hs99999905_m1) and PUM1 (Hs00206469_m1; cat#4331182), all from Thermo Fisher Scientific. Universal human reference RNA (UHR; cat#636690, Clontech Laboratories, Mountain View, CA, USA) -total RNA from a range of normal adult human tissues, was used as a calibrator for the 2 Ct analysis. NTERA-2 cells were included as a positive control, and a no template control (NTC) included to control for contamination. End-point amplification products were checked for the presence of bands of the correct size by gel electrophoresis on 2% agarose gels (cat#G402002, Thermo Fisher Scientific). Graphs were generated using GraphPad Prism (v8.0.2, San Diego, CA, USA) and results presented as expression fold change relative to UHR.

In vitro Tumorsphere Formation Assays
Tumorsphere suspension cultures were developed using three mHNcSCC-derived primary cell lines. Briefly, 2.5 × 10 5 live cells from adherent cultures at passages 7-9 were seeded in 24 mL StemXVivo serum-free tumorsphere media (cat#CCM012, R&D Systems, Minneapolis, MN, USA), in T75 Nunclon TM Sphera TM EasYFlasks (cat#174952, Thermo Fisher Scientific) supplemented as per the manufacturer's protocol. Cells were maintained in a 5% incubator at 37 • C for up to 10 days, and fed every 3-4 days with the addition of 12 mL of media. Cultures were observed, photographed and measured daily from day 3 post seeding under an Olympus CKX53 Microscope (Tokyo, Japan). Three fields of view were captured for each cell line. From each field of view five representative spheres were identified and two measurements per sphere were performed to establish an average size within each. Culture results are reported as the average size of all spheres measured and the percentage of spheres measured that exceed the 50 µm threshold, which is the minimum required size for a sphere to be considered positive (45).

Statistical Analysis
Because cell counting was performed for each mHNcSCC tissue sample that underwent IHC staining and ISH staining was tested for all five iPSC markers, and because three technical replicates were done for each marker, a repeated measures analysis was performed to test for differences between markers using Generalized Estimating Equations (GEE) (46). The marginal means and their standard errors were estimated from the GEE model and compared between markers using sequential Bonferroni correction for multiple comparisons. A p-value of < 0.05 was taken as significant.  (Figure 1C) was expressed in the cytoplasm of cells within the TNs, but not the cells within the PTS. SOX2 (Figure 1D) was expressed on the nucleus and cytoplasm of cells within the TNs, and in the nucleus of cells within the PTS. KLF4 was present in the cytoplasm of cells throughout the TNs (Figure 1E), and to a lesser extent, cells within the PTS (Figure 1E). Cytoplasmic and nuclear expression of c-MYC ( Figure 1F) was observed in cells within the TNs and occasional cells within the PTS.  (Figure 2). There was no statistically significant difference between c-MYC and KLF4 (Figure 2).

IF Staining Showed Three CSC Subpopulations Within mHNcSCC Tissue Samples
IF staining demonstrated abundant expression of c-MYC (Figures 3A-C, green) on the cells within the PTS (arrowheads) and the TNs (arrows). The c-MYC+ subpopulations within the TNs and the PTS also demonstrated cytoplasmic expression of KLF4 (Figure 3A, red). Most of the c-MYC+ cells within the TNs and few c-MYC+ within the PTS also expressed NANOG (Figure 3B, red) with few c-MYC+ cells within the PTS not expressing NANOG (Figure 3B). These c-MYC+ subpopulations within the TNs and the PTS exhibited nuclear and cytoplasmic expression of SOX2 (Figure 3C, red). The KLF4+ (Figure 3D, red) cells within the TNs and the PTS also expressed OCT4 (Figure 3D, green). The majority of these OCT4+ (Figures 3E,F, green) cells within the TNs and the PTS also expressed NANOG (Figure 3E, red), and all of them expressed SOX2 (Figure 3F, red). Figure insets have been provided to show enlarged views of the corresponding images.
Images of individual stains of the merged images presented in Figure 3 are provided in Supplementary Figure 2. Specificity of secondary antibodies was confirmed on the negative control (Supplementary Figure 2M), which demonstrated minimal staining.

RT-qPCR Demonstrated mRNA Transcripts of iPSC Markers in mHNcSCC Tissue Samples and mHNcSCC-Derived Primary Cell Lines
RT-qPCR analysis demonstrated transcript expression of OCT4, NANOG, SOX2, KLF4, and c-MYC in all four mHNcSCC tissue samples ( Figure 6A) and in all three of the mHNcSCC-derived primary cell lines (Figure 6B).
In the mHNcSCC tissue samples c-MYC showed a biologically significant increase in expression, relative to UHR. SOX2 had a biologically significant (increase in expression, relative to UHR. SOX2 had a biologically significant decrease in expression, relative to UHR. The markers with the highest to the lowest mRNA expression were: c-MYC > OCT4, KLF4 ≥ NANOG ≥ SOX2. In the mHNcSCC-derived primary cell lines, OCT4, NANOG, and SOX2 all showed a biologically relevant decrease in expression, relative to UHR. The markers with the highest to the lowest mRNA expression were: Specific amplification of the products was demonstrated by electrophoresis of qPCR products on 2% agarose gels (Supplementary Figure 4). The expected size amplicons were observed, and no products were observed in the NTC reactions (Supplementary Figure 4).

Western Blotting Demonstrated
Expression of SOX2, KLF4, and c-MYC but Not OCT4 and NANOG by mHNcSCC-Derived Primary Cell Lines WB of mHNcSCC-derived primary cell lines derived from mHNcSCC tissue samples from three of the original cohort of 15 patients showed very low expression levels of SOX2 over two bands: one at the expected 35 kDa size and another at 40-42 kDa (Figure 7A), which is consistent with one of its known post-translational modifications (47). KLF4 (Figure 7B) and c-MYC (seen as two bands) ( Figure 7C) were expressed in all three mHNcSCC-derived primary cell lines; the lower band for c-MYC may be MYC-nick (48). NANOG ( Figure 7D) and OCT4 ( Figure 7E) were below detectable levels. Equal amounts of proteins were loaded into each lane, as confirmed by α-tubulin staining ( Figure 7F).

mHNcSCC-Derived Primary Cell Lines Demonstrated Tumorsphere Formation in vitro
All three mHNcSCC-derived primary cell lines demonstrated in vitro tumorsphere formation (Figures 8A-C) at the first passage with a mean diameter at cessation of culture averaging 62.6, 72.9, and 88.7 µm respectively (Supplementary Figure 5), meeting our threshold (spheres that were > 50 µm in size).
The percentage of spheres measured that exceeded the 50 µm threshold in these cell lines were 93.3, 100, and 100%, respectively (45). Tumorsphere culture was halted at the first signs of the FIGURE 5 | Statistical analysis of cell counting for in situ hybridization stained slides demonstrating that c-MYC had the highest mean proportion of positive cells, followed by KLF4 although there was no statistically significant difference between them. OCT4 and NANOG had lower mean proportions of positive cells compared with c-MYC and KLF4 although there was no statistically significance difference between them. The mean proportion of cells that stained positively for SOX2 was significantly lower (p < 0.0005) than the other four iPSC markers. Error bars: 95% confidence interval. ***p < 0.05. on-set of dark centers indicating cell death. This occurred on days 6, 7, and 5 respectively. This provides some preliminary evidence of stem cell functionality within the mHNcSCC-derived primary cell lines.

DISCUSSION
NMSCs are the commonest cancers diagnosed worldwide (49) and make up one third of all malignancies (50), with 20% being SCC (51).
Tumor initiation (23), development (52), recurrence (53), and chemoradiation resistance have been attributed to the presence of CSCs (54), which have been recently identified within primary HNcSCC (19) using the five iPSC markers used in this study. Targeting CSCs has been suggested as a novel treatment approach for cancer (55).
We have previously identified an OCT4+/NANOG+ /SOX2+/KLF4+/c-MYC+ CSC subpopulation within the TNs, and an OCT4+/NANOG+/SOX2+/KLF+/c-MYC+ and an OCT4+/NANOG-/SOX2+/KLF4+/c-MYC+ CSC subpopulation within the PTS of primary HNcSCC (19). The CSC subpopulations within primary HNcSCC differ from the CSC subpopulations identified in mHNcSCC demonstrated in this study. We have identified an OCT4+/NANOG-/SOX2+/KLF4+/c-MYC+ population within the TNs in mHNcSCC, which is not present in primary HNcSCC. The OCT4+/NANOG-/SOX2+/KLF4+/c-MYC+ CSC  subpopulation identified within the PTS in primary HNcSCC, is not demonstrated in mHNcSCC in this study. Using the same iPSC markers, we have also shown the presence of two CSC subpopulations within head and neck metastatic malignant melanoma (HNmMM): an OCT4+/SOX2+/KLF4+/c-MYC+ CSC subpopulation within the TNs, and another in the PTS with NANOG present only in two of the 20 cases studied (44). Three out of four primary cell lines derived from four different HNmMM tissue samples formed tumorspheres in vitro. As with the current study of mHNcSCC, the formation of tumorspheres in HNmMM-derived primary cell lines in vitro, provides preliminary functional evidence of the presence of CSCs in HNmMM tissues (44).
The presence of the CSC subpopulation within the PTS of mHNcSCC was supported by both protein and mRNA expression of OCT4, NANOG, SOX2, KLF4, and c-MYC on the cells within the PTS by IHC staining and ISH, respectively. The absence of NANOG in one of the two CSC subpopulations within the TNs aligns with previous studies demonstrating dispensability of NANOG in the presence of the other four iPSC markers for induction of pluripotency (13), the phenotype exhibited by CSCs (8). As with primary HNcSCC (19), IHC staining in this study demonstrated localization of NANOG to CSCs in the TNs, and ISH detected mRNA transcripts for NANOG in cells within the TNs and, to a lesser extent, those within the PTS. The presence of NANOG mRNA in cells within the TNs and the absence of its protein in one of the two CSC subpopulations within the TNs, is consistent with the dispensability of NANOG for pluripotent capability (14). The absence of NANOG in the aforementioned CSC subpopulation is interesting, as NANOG has been associated with the development of cervical SCC when it is present on stromal cells (56). NANOG has been used as a CSC marker in many cancer types (16,(37)(38)(39)(40)57). Targeting NANOG in head and neck squamous cell carcinoma-derived spheroid cells significantly inhibits tumor aggressiveness and increases chemosensitivity to cisplatin (57); perhaps targeting NANOG within mHNcSCC would have a similar effect on these parameters. The presence of CSCs within mHNcSCC tissue samples is supported by RT-qPCR which demonstrated the presence of all five iPSC markers in three mHNcSCC-derived primary cell lines, while WB showed expression of these markers except for NANOG and OCT4. The capability of the three mHNcSCC-derived primary cell lines to form initial tumorspheres in vitro provides preliminary evidence of the presence of CSCs within mHNcSCC. Further functional tests to confirm these findings, such as additional passages and in vivo xenotransplantation of the said tumorspheres into mice, should be undertaken.
As with our previous study of primary HNcSCC (19), SOX2 expression was observed within mHNcSCC tissue samples and mHNcSCC-derived primary cell lines, albeit at very low levels. This may suggest the dispensability of SOX2 following metastasis, as observed in metastatic malignant melanoma (58).
If SOX2 is dispensable following metastasis in mHNcSCC as in metastatic malignant melanoma (58), a correlation between low SOX2 expression levels and a more aggressive cancer may be expected. However, Li et al. (59) demonstrate that SOX2 is a downstream target of the Hippo effector TAZ in HNcSCC, which is capable of reprogramming differentiated cancer cells into CSCs, and that SOX2 and TAZ upregulation are both associated with poorer overall survival in HNcSCC. Given the presence of SOX2 in mHNcSCC (60) and its association with local recurrence and metastasis, we speculate the TAZ-SOX2 axis as an important determinant of CSC stemness in this tumor. The low expression of SOX2 within mHNcSCC-derived primary cell lines and the lower proportions of SOX2 positive cells in mHNcSCC tissue samples on IHC staining and ISH staining compared to the other four iPSC markers may reflect the aggressiveness of this cancer, as decreased SOX2 expression has been associated with carcinogenesis and the development of OCSCC (60).
RT-qPCR and WB analyses showed some differences of the expression profile in the iPSC markers between the mHNcSCC tissue samples and mHNcSCC-derived primary cell lines. It has been observed that the expression of up to 10% of genes is altered within five passages during tumorsphere formation in vitro (61). These alterations in gene expression driven by the cell culture environment may explain why SOX2 was expressed at a very low level, and why OCT4 and NANOG were not detected by WB in mHNcSCC-derived primary cell lines, compared with the RT-qPCR results for the mHNcSCC tissue samples. It may also account for the differences in the RT-qPCR results between the mHNcSCC tissue samples and the mHNcSCC-derived primary cell lines.
Given its role and overexpression in cSCC (62), it was unsurprising that KLF4 was expressed in all mHNcSCC tissue samples and mHNcSCC-derived primary cell lines. Our finding of c-MYC expression in mHNcSCC aligns with reports of c-MYC overexpression in other types of carcinoma (29,37,39,40).
The finding of OCT4 expression in mHNcSCC is consistent with our previous study demonstrating OCT4 expression in primary HNcSCC (19) and other solid tumors (63). As a key regulator of pluripotency in the early mammalian embryo and given it is consistently expressed by ESCs and hematopoietic and neural stem cells (26), the presence of OCT4 in mHNcSCC supports the presence of primitive CSC subpopulations in this tumor.
We have demonstrated the presence of three putative CSC subpopulations within mHNcSCC. Further, investigating the expression of the RAS-an important regulator of stem cells (69,70)-by these CSC subpopulations may provide an avenue of therapeutic targeting of them in the treatment of this aggressive cancer (55). These findings are novel, and further investigation with a larger sample size with functional studies are needed to confirm the CSC subpopulations identified in this study.

CONCLUSIONS
This study demonstrates expression of the iPSC markers OCT4, NANOG, SOX2, KLF4, and c-MYC in mHNcSCC, with the presence of three CSC subpopulations. mHNcSCC-derived primary cell lines express transcripts of OCT4, NANOG, SOX2, KLF4 and c-MYC. SOX2, KLF4, and c-MYC, but not OCT4 and NANOG proteins. These cell lines form tumorspheres in vitro, at a single passage. The identification of CSCs within mHNcSCC may open novel avenues for therapeutic targeting of this aggressive cancer.

DATA AVAILABILITY STATEMENT
The datasets generated for this study are available on request to the corresponding author.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by Central Regional Health and Disability Ethics Committee (Ref. 12/CEN/74AM05). The patients provided their written informed consent to participate in this study.