Stemness in Human Thyroid Cancers and Derived Cell Lines: The Role of Asymmetrically Dividing Cancer Stem Cells Resistant to Chemotherapy

Abstract

Context:

Cancer stem cells (CSCs) have the ability to self-renew through symmetric and asymmetric cell division. CSCs may arise from mutations within an embryonic stem cell/progenitor cell population or via epithelial-mesenchymal transition (EMT), and recent advances in the study of thyroid stem cells have led to a growing recognition of the likely central importance of CSCs in thyroid tumorigenesis.

Objective:

The objectives of this study were to establish the presence of a stem cell population in human thyroid tumors and to identify, isolate, and characterize CSCs in thyroid cancer cell lines.

Results:

1) Human thyroid cancers (n = 10) and thyroid cancer cell lines (n = 6) contained a stem cell population as evidenced by pluripotent stem cell gene expression. 2) Pulse-chase experiments with thyroid cancer cells identified a label-retaining cell population, a primary characteristic of CSCs, which at mitosis divided their DNA both symmetrically and asymmetrically and included a population of cells expressing the progenitor marker, stage-specific embryonic antigen 1 (SSEA-1). 3) Cells positive for SSEA-1 expressed additional stem cell markers including Oct4, Sox2, and Nanog were confirmed as CSCs by their tumor-initiating properties in vivo, their resistance to chemotherapy, and their multipotent capability. 4) SSEA-1-positive cells showed enhanced vimentin expression and decreased E-cadherin expression, indicating their likely derivation via EMT.

Conclusions:

Cellular diversity in thyroid cancer occurs through both symmetric and asymmetric cell division, and SSEA-1-positive cells are one form of CSCs that appear to have arisen via EMT and may be the source of malignant thyroid tumor formation. This would suggest that thyroid cancer CSCs were the result of thyroid cancer transformation rather than the source.

Thyroid cancers are the most common endocrine malignancy, comprising approximately 1% of all human malignancies, but they have been increasing in incidence more rapidly than any other in North America (1). Papillary (PTC), follicular (FTC), and anaplastic (ATC) thyroid carcinomas are derived from the follicular epithelium, whereas medullary thyroid cancer is of neuroectodermal origin. PTC comprise 80–85% and FTC comprise 10–15% of all thyroid neoplasms and are together termed differentiated thyroid cancers. The least common (1–2%) histotype is the ATC, which has a rapid lethal progression.

It has been shown that cancers, including thyroid cancer (2–6), have a cellular hierarchy and that only a small population of cells called cancer stem cells (CSCs) drive cancer growth. CSCs are cells within a tumor that possess the capacity for self-renewal and can generate heterogeneous lineages of cells that comprise a tumor (7). It should be noted that this definition does not indicate the source of these cells; these tumor-forming cells could hypothetically originate from stem, progenitor, or differentiated cells, but all have the ability to self-renew and generate the diverse cells that comprise the tumor. CSCs may, therefore, be responsible for sustaining and enlarging the tumor. Increasing evidence suggests that CSCs also mediate tumor metastases and are resistant to conventional anticancer therapeutics, thus contributing to relapse. Therefore, the identification and characterization of such a tumorigenic population may represent a crucial step in the development of effective therapies. However, several recent studies have demonstrated that CSCs and non-CSCs can exhibit plasticity with a transition from one state to another (8, 9), and this raises the possibility that approaches to just target CSCs will not be sufficient to cure the patient because the remaining non-CSCs may be reprogrammed and reinitiate tumorigenesis. One form of such transition is epithelial-mesenchymal transition (EMT), which may be a common source of cells expressing stemness and having multipotent potential (10).

There remains a paucity of information about thyroid CSCs and how to definitively identify such cells within thyroid tumors. Recent studies suggest that several markers, including CD133 (2, 4, 5), ABCG2 (11), and ALDH1 (3), are associated with thyroid cancer progression and resistance to current modes of chemotherapy but have not proven to be useful specific markers of CSCs. In search of an alternative and/or more general enrichment marker for thyroid cancer CSCs, we now demonstrate that the progenitor marker—stage-specific embryonic antigen 1 (SSEA-1)—can be detected in several human thyroid cancer lines, and that SSEA-1-positive (SSEA-1⁺) thyroid cancer cells divide both symmetrically and asymmetrically, an important characteristic of CSCs. In addition, these cells showed evidence of EMT initiation with enhanced vimentin expression, decreased E-cadherin expression, and increased Snail gene expression—all markers of EMT. These data provide definitive evidence that cellular diversity, especially in ATC, may occur through both symmetric and asymmetric cell division. Furthermore, it is likely that the asymmetrically dividing cells have arisen via EMT and may be the source of malignant thyroid tumor cell formation.

Materials and Methods

Patient thyroid cancer tissues

Thyroid cancer tissues were obtained from surgical pathology specimens provided by patients at the Mount Sinai Medical Center affected by PTC (n = 10) and benign thyroid adenomas (n = 10) with corresponding normal-appearing tissues available with each specimen. Samples were obtained with approval of the Institutional Review Board.

Thyroid cancer cell lines

The thyroid cancer cell lines employed are listed in Supplemental Table 1 (published on The Endocrine Society's Journals Online web site at http://jcem.endojournals.org). Tad2 cells were derived by transfection of normal human fetal thyroid follicular cells with the SV40 A gene encoding the large tumor (T) antigen and selected for cAMP responsiveness to TSH (12). Three FTC cell lines were examined: ML1 was kindly supplied by Dr M. Szkudlinski (Trophogen Inc), and FTC236 and FTC238 were purchased commercially (Sigma). T238 and SW1736, established from human ATCs, and TPC1, established from PTC, were kindly provided by Dr James Fagin (Memorial Sloan-Kettering Cancer Center). Cells were grown in RPMI 1640 (Tad2, T238), DMEM (ML1, SW1736, and TPC1), or DMEM/F12 (FTC236 and FTC238) medium supplemented with 10% fetal bovine serum, 100 IU penicillin/mL, and 100 μg/mL streptomycin and with 100 μm sodium pyruvate (ML1), or with 10 μg/mL insulin, 0.01 U/mL TSH (FTC236). All cells were maintained in 5% CO₂ at 37°C in a humidified incubator.

Cell identification

Bromodeoxyuridine (BrdU) administration and chase were performed as described (13). Flow cytometry analysis and cell sorting were performed as previously described (14). Immunofluorescence staining of cells was performed by standard techniques using cell permeabilization to detect intracellular antigens, and cells were mounted using hard set mounting media containing DAPI (Vector Laboratories). To detect asymmetrically dividing cells, cells were plated at low density and incubated for 24–48 hours before immunodetection. Pairs of cells whose nuclei were close and connected by α-tubulin staining were considered to be cells originating from the same cell. The numbers of symmetrically and asymmetrically dividing cells were counted as a proportion of all dividing cells in the microscope fields. Cell viability was determined using a cell count kit-8 (Dojindo Molecular Technologies).

Differentiation of thyroid CSCs into adipocytes

Cells were cultured in medium alone or medium supplemented with 10 μg/mL transferrin, 0.5 mg/mL insulin, 0.2 nm T₃, 1 μm dexamethasone, and 1 mU/mL TSH for up to 15 days. Cells were fixed in 4% paraformaldehyde in PBS for 15 minutes at room temperature, washed with PBS, stained with 0.5% oil red O (Sigma) for 30 minutes, and then washed with H₂O to remove the staining solution.

Chemotherapy enrichment assays

Single-cell suspensions of human thyroid cancer cells were plated at 5 × 10⁴ cells/mL in culture medium containing dimethyl sulfoxide (DMSO) vehicle control, or fluorouracil (5-FU) (Sigma) for 3 or 6 days, depending on the analysis time points. A range of drug concentrations was initially used to determine the toxicity of the drug.

In vivo tumorigenicity assays

Five-week-old nude male mice (The Jackson Laboratory) were housed in pathogen-free conditions with the approval of the Institutional Animal Care and Use Committee of Mount Sinai School of Medicine. Various numbers of SSEA-1-negative (SSEA-1⁻) cells, SSEA-1⁺ cells sorted from T238 cells, and T238 cells were resuspended in Matrigel-PBS and injected sc. Mice were killed 4 to 8 weeks after cell injection, and the “tumors” were harvested for further examination. All animal experiments were performed according to approved institutional protocols.

RNA isolation and RT-PCR

Total RNA was extracted from cultured thyroid cancer cells using the RNeasy system (QIAGEN Ltd). cDNA synthesis was performed using the SuperScript III system (Invitrogen Corp). Relative expression levels of each RT-PCR product were analyzed using the 2^−ΔΔCT method and normalized to the expression of the housekeeping gene GAPDH. Data presented (mean) are from three independent experiments in which all sample sets were analyzed in triplicate.

Western blotting

Western blotting was performed according to a standard protocol using corresponding antibodies and secondary horseradish peroxidase-conjugated antimouse or antirabbit antibodies (Vector Laboratories) as described (15).

Results

Pluripotent stem cell markers in normal, adenomatous, and malignant human thyroid cancer

Human PTCs, thyroid adenomas, and adjacent normal thyroid tissues from the same patient were evaluated for the expression of the thyroid-specific marker, thyroglobulin (Tg), and a number of stemness markers using quantitative RT-PCR (qRT-PCR) (Figure 1). In each pair of samples from each patient, we studied the expression of Oct4, ABCG2, CD44, Nanog, and SSEA-1 genes. We considered Tg expression as a measure of thyroid cell differentiation because this may be lost in dedifferentiated thyroid cancer cells and it was present to a variable degree in all samples examined (Figure 1A). In contrast, the expression of stemness marker genes indicates a state of relative undifferentiation because such gene expression is lost as a cell differentiates into its end phenotype (16, 17). However, stemness markers Oct4, ABCG2, CD44, Nanog, and SSEA-1 were also detected in all the tissues examined, confirming the presence of cells with stemness characteristics in normal thyroid, thyroid adenomas, and PTC (Figure 1B). The ratio of expression of Tg to each individual stemness marker is shown in Supplemental Figure 1. When a tissue appears to be more undifferentiated, this ratio can be expected to decrease. The data clearly demonstrate a greater degree of undifferentiation in PTC due to increased stem cell marker expression and reduced Tg expression when compared to thyroid adenomas and normal thyroid tissue.

Figure 1.

qRT-PCR analysis of the thyroid specific marker Tg (A) and stemness markers (Oct4, ABCG2, CD44, Nanog, and SSEA-1) (B) in human PTC, thyroid adenoma, and normal thyroid tissues. The data were normalized using GAPDH as an endogenous control. The results are expressed as mean ± SEM of three independent experiments with three parallels.

Characterization of thyroid cancer cell lines

RT-PCR analysis of thyroid gene markers was performed in a sample of thyroid cancer cell lines described in Supplemental Table 1 (Figure 2A). The expression of Tg and thyroid peroxidase was detected in the Tad2 and ML1 cells, as was the expression of the TSH receptor, although less than seen in normal thyroid tissue. There was loss of Tg and thyroid peroxidase expression in the ATC lines T238 and SW1736, the FTC lines F236 and F238, and the PTC line TPC1 cells, indicative of their dedifferentiated status. NIS gene expression was not detected in any of the cells, whereas Pax8 was detected in all of them.

Figure 2.

A, RT-PCR analysis for thyrocyte markers in human thyroid cancer cell lines. N, negative control; ES, murine embryonic stem cells; Thy, normal thyroid tissue; Tad2, the immortal thyroid cell; T238 and SW1736, human ATC cell lines; ML1, F236, and F238, human follicular thyroid cancer cell lines; TPC1, human PTC cells. B, qRT-PCR analysis for pluripotent stem cell markers in human thyroid cancer cell lines. Data were expressed as mean ± SEM and represent one of three separate experiments. C, Detection of BrdU in T238 cells either not grown in or grown at the end of the BrdU pulse by flow cytometry. Gating was based on an isotype control. All cells at the end of the BrdU pulse were BrdU positive. D, Immunofluorescent staining for BrdU in T238 cells at the end of the BrdU pulse. A representative image is shown in which all of the cells at various degrees of condensed chromatin were BrdU positive (green). (Scale bar, 100 μm.) E, Immunofluorescent staining for BrdU in T238 cells on the 14th day after the BrdU was washed out. A representative image is shown in which only a few cells were BrdU positive (green). (Scale bar, 100 μm.) F and G, Representative immunofluorescent staining for BrdU in T238 cells that partition their BrdU-labeled template DNA (green) either randomly to both daughter cells (F) or exclusively to one daughter cell (G).

The expression of pluripotent stem cell markers was also examined in these thyroid cancer cell lines using qRT-PCR (Figure 2B). The cells showed increased expression of stem cell markers in comparison to normal thyroid, indicating the presence of more undifferentiated cells. One exception was the immortal thyroid cell Tad2, which showed poor expression of stem cell markers evidencing a more differentiated state, as shown in Figure 2A.

Characterization of label-retaining cancer thyroid cells

We first used the ability of potential CSCs to retain DNA label over time as a selection property known to be associated with adult stem cells. Such label-retaining cells (LRCs) have been identified by repeatedly exposing cells either in vivo or in vitro to nucleotide analogs such as BrdU (13, 18, 19). We initially cultured T238 cells for 7 days in the presence of 1 μm BrdU to ensure that both sets of DNA strands were labeled in the vast majority of cells. We then stained cells at the end of the pulse with an anti-BrdU antibody and examined them by flow cytometry and immunofluorescence. As expected, almost 100% of the cells were initially BrdU⁺ (Figure 2, C and D), verifying that the BrdU pulse was sufficient. However, only sporadic cells were BrdU⁺ by the 14th day after withdrawing BrdU (Figure 2E), giving approximately 6% of LRCs.

Recent data suggest that CSCs can self-renew and repopulate the heterogeneous tumor bulk via asymmetric cell division (13, 19, 20). To determine whether our potential CSCs in the form of LRCs had the ability to asymmetrically divide their BrdU-labeled template DNA exclusively to one daughter cell or another, we used pulse-chase experiments with BrdU to label DNA and track the LRCs asymmetrically dividing (18). According to the model, asymmetric division of template DNA should be observed after the first cell division of the chase (13, 18). We examined the BrdU-retaining pattern of the template DNA in T238 cells during cell division; among the 251 dividing cells examined, 210 (84%) cells demonstrated a positive BrdU label segregated to both daughter cells (Figure 2F), whereas 41 (16.3%) cells demonstrated preferential segregation of the BrdU-labeled template DNA to one daughter cell and less BrdU label to the opposing cell (Figure 2G). These data confirmed the high likelihood that the thyroid cancer cell lines contained CSCs within their cell population.

SSEA-1 expression marks a major CSC subset in thyroid cancer

SSEA-1, a human embryonic stem cell early differentiation marker (16), has been previously identified as a marker for CSCs in human and mouse brain tumors (21–23) and in tumors induced by transforming fibroblasts (24). These SSEA-1-expressing cancer cells were shown to exhibit self-renewal and repopulation abilities through asymmetric division both in vivo and in vitro. To identify whether SSEA-1 was a potential marker of putative CSCs in human thyroid cancer, we examined SSEA-1 expression in the same panel of human thyroid cancer cell lines. RT-PCR analysis (Figure 3A) and flow cytometric analysis (Table 1) revealed that SSEA-1 was expressed in all the cell lines examined but to a variable degree. Similarly, immunofluorescent staining illustrated the surface expression of SSEA-1 on thyroid cancer cells (Figure 3B). These observations suggested that SSEA-1 may be used as a marker of putative CSCs for thyroid cancer cells.

Figure 3.

Detection of expression of SSEA-1 in human thyroid cancer cell lines. A, RT-PCR analysis for SSEA-1 in human thyroid cancer cell lines. B, Immunofluorescent staining for SSEA-1 in human thyroid cancer cells. A representative image of T238 cells is shown in which SSEA-1 (green) was expressed on the cell surface membrane and α-tubulin (red) was expressed in the cytoplasm. (Scale bar, 100 μm.) C–G, Characterization of SSEA-1⁺ cells isolated from the anaplastic T238 cell line. C, Abundance of SSEA-1⁺ cells in a sorted SSEA-1⁺ cell population over time in culture. This is a representative time course experiment. D and E, Immunodetection of SSEA-1 (green) in dividing cells exhibited both asymmetric (D) and symmetric (E) cell division. α-Tubulin was stained red. Nuclei were counterstained with DAPI (blue). (Scale bar, 100 μm.) F, qRT-PCR analysis indicates that SSEA-1⁺ cells express higher levels of additional stem cell markers (Oct4, Sox2, and Nanog) than do SSEA-1⁻ cells. G, qRT-PCR analysis of EMT-related genes (vimentin, E-cadherin, and Snail) indicates enhanced vimentin expression and decreased E-cadherin expression with increased Snail expression in SSEA-1⁺ cells when compared to SSEA-1⁻ cells. Data were expressed as mean ± SEM and represent one of three separate experiments.

Table 1.

FACS Analysis of SSEA-1-Expressing Cells in Human Thyroid Cancer Cell Lines

Cell Line	% SSEA-1+
ML1	2.9 ± 0.36
FTC236	6.8 ± 0.58
FTC238	7.4 ± 0.66
T238	9.2 ± 1.20
SW1736	7.8 ± 0.82
TPC1	8.2 ± 0.88

Characterization of SSEA-1⁺ thyroid cancer cell division

A characteristic feature of stem cells is their ability to both self-renew and generate phenotypically distinct daughter cells (25, 26). To confirm that SSEA-1-expressing cells in thyroid cancer cells are CSCs, we first sorted cells into SSEA-1⁺ and SSEA-1⁻ cells. For example, data for the ATC line T238 showed approximately 9% of positive cells, and these SSEA-1⁺ sorted cells, when grown in culture, rapidly generated a heterogeneous progeny with the percentage of SSEA-1⁺ cells in the population decreasing to that of the unsorted population in 6 days (Figure 3C). In agreement with this observation, immunofluorescent staining revealed a large increase in asymmetric cell divisions, with 51% of dividing cells undergoing asymmetric cell division giving rise to one SSEA-1⁺ daughter cell and one SSEA-1⁻ daughter cell (Figure 3D), whereas 49% of dividing cells yielded two SSEA-1⁺ daughter cells (Figure 3E), demonstrating that SSEA-1⁺ cells can repopulate the tumor cell population via asymmetric division. Such studies do not exclude bidirectional interconversions between stem and nonstem compartments explaining how fully differentiated cells may acquire self-renewing capacity (27, 28) and do not exclude stem cells arising de novo from nonstem cells (29, 30). To answer such questions we would have needed a population of 100% SSEA-1⁻ cells, but even after cell sorting, culture of the SSEA-1⁻ cells produced a population of positive cells (data not shown).

Stemness gene expression in SSEA-1⁺ cells

We also characterized the stemness profiles of SSEA-1⁺ and SSEA-1⁻ thyroid cancer cells by qRT-PCR. The expression of additional stem cell markers (Oct4, Sox2, and Nanog) was significantly increased in the SSEA-1⁺ cell population (Figure 3F). Furthermore, these cells showed evidence of EMT initiation with enhanced vimentin expression and decreased E-cadherin expression (259.46 and 0.57%, respectively) with increased Snail expression when compared to SSEA-1⁻ cells (Figure 3G). Hence, it is likely that the SSEA-1⁺, asymmetric dividing cells had arisen via EMT and were the likely source of the malignant thyroid cell population.

Multipotentiality of SSEA-1⁺ cells

To test the multipotent potential of SSEA-1⁺ thyroid cancer cells, both SSEA-1⁺ and SSEA-1⁻ sorted cells from T238 were cultured in adipocyte differentiation medium. The early adipocyte differentiation markers, including the fatty acid-binding protein 4, CCAAT-enhancer-binding protein α, and lipoprotein lipase, were markedly increased in SSEA-1⁺ cells compared to SSEA-1⁻ cells after culturing the two groups of cells in adipocyte differentiation medium for 15 days (Figure 4A). These results indicated the greater multipotent potential of SSEA-1⁺ cells in keeping with the characteristics of CSCs.

Figure 4.

A, qRT-PCR analysis of adipocyte differentiation markers in SSEA-1⁺ and SSEA-1⁻ cells isolated from T238 cells. Cells were analyzed after culture with adipocyte differentiation medium for 15 days. Data were expressed as mean ± SEM and represent one of three separate experiments. FABP4, fatty acid-binding protein 4; CEBPα, CCAAT-enhancer-binding protein α; LPL, lipoprotein lipase. B and C, SSEA-1⁺ (B) and SSEA-1⁻ (C) cells from the T238 anaplastic cell line reconstituted tumors in nude mice. Mice injected with different number of cells developed tumors within 4 weeks. Hematoxylin-eosin analysis of these xenografts showed that they were remarkably similar.

In vivo characterization of SSEA-1⁺ cells

To assess the tumorigenic potential of SSEA-1⁺ cells in vivo, SSEA-1⁺ cells or SSEA-1⁻ cells were injected sc into nude mice. Injecting large numbers of cells (100 000) produced tumors in all mice (Figure 4, B and C). Although three of four mice injected with 10 000 SSEA-1⁺ cells developed tumors, no tumors were observed when mice were injected with 10 000 SSEA-1⁻ cells. These observations suggested that SSEA-1⁺ cells initiated tumor growth more efficiently. The hematoxylin-eosin analysis of these xenografted tumors was similar to tumors formed from the parent T238 cells.

Chemotherapy resistance of SSEA-1⁺ cells in human thyroid cancer

CSCs have been reported to have an intrinsic resistance to conventional chemotherapy and radiotherapy (2, 5, 31). Current treatment strategies may therefore affect the bulk of the tumor cells but leave CSCs behind, serving as a starting point for disease recurrence. To investigate whether chemotherapy resistance is also a characteristic of SSEA-1⁺ cells in human thyroid cancer lines, T238 cells were treated with either DMSO vehicle or increasing concentrations of 5-FU. After 3 days of treatment, DMSO-treated cells had expanded by 4-fold, whereas 5-FU-treated cells only contained a fraction of viable cells (Figure 5A). By fluorescence-activated cell sorting (FACS) analysis, we observed that cells expressing SSEA-1 protein were enriched by treatment with 5-FU from approximately 10.0% to > 50% of the cell population in T238 cells when treated with 250 μm 5-FU. The ML1, F236, and F238 cells also showed increases in SSEA-1⁺ cells (Figure 5B). The expression of stem cell markers SSEA-1, Oct4, Nanog, Sox2, and CD133 was significantly enhanced in the resistant T238 cells when analyzed by qRT-PCR (Figure 5C). Furthermore, asymmetric cell divisions were increased in these 5-FU-resistant cells (Figure 5D), and expression of pAKT was decreased (Figure 5E). This is consistent with the report that asymmetric cancer cell division results from suppression of AKT/PKB kinase signaling (20) and suggested that factors, including exposure to chemotherapy, that modulate AKT signaling can shift the dynamic between symmetric and asymmetric division.

Figure 5.

A, Cell survival curves in response to increasing concentrations of 5-FU after 48-hour treatment of T238 cells. B, FACS analysis of SSEA-1-expressing cells in the thyroid cancer cell lines treated with 250 μm 5-FU for 3 and 6 days. Gating was based on isotype control. C, qRT-PCR analysis of mRNA expression of pluripotent stem cell markers in T238 cells and 5-FU-treated T238 cells. Data were expressed as mean ± SEM and represent one of three separate experiments. D, T238 cells treated with DMSO or 5-FU (250 μm) for 3 days and counted for asymmetric cells. E, Signal transduction markers assessed by Western blot analysis of T238 cells treated with DMSO or 5-FU (250 μm) for 3 days.

These resistant T238 cells were also able to be induced to differentiate into adipocyte-like cells (Supplemental Figure 2A) and showed tumorigenesis after injection into nude mice as seen earlier (Supplemental Figure 2B).

Discussion

Dedifferentiation of thyroid cancers is characterized by a reduction or loss of thyroid-specific gene expression. In our study, the human PTCs, the thyroid adenomas, and the human thyroid cancer cells all showed a relative reduction or loss of Tg gene expression but an increase in pluripotent stem cell markers when compared to normal thyroid. This confirmed the presence of undifferentiated cells such as CSCs, as previously reported (32–34). The identification of thyroid CSCs marks a step toward finding new and effective ways to treat human thyroid cancer. The concept of CSCs has constructive significance for clinical practice because it has been well shown that CSCs contribute to relapse, chemoresistance, and radioresistance in cancers (5, 35).

In the current study, we found SSEA-1 expression to be one marker able to identify and enrich a subpopulation of thyroid cancer cells with the properties ascribed to CSCs. The studies presented here demonstrate that human thyroid cancers contain stem cells, that thyroid cancer cell lines contain SSEA-1-expressing cells with properties of self-renewal, and that these cells show asymmetric cell divisions, can generate tumors, and are resistant to chemotherapy. It has been reported previously that SSEA-1⁺ cells undergo both symmetric and asymmetric cell division and give rise to both SSEA-1⁺ and SSEA-1⁻ progeny (24). These findings indicate that in thyroid cancers, the SSEA-1⁺ and SSEA-1⁻ populations are in dynamic equilibrium during cancer progression. Furthermore, these cells showed a high expression of embryonic stem cell-related genes (Oct4, Nanog, and Sox2), confirming their stemness potential, and showed higher tumor-initiating ability (24).

Another characteristic of CSCs has been the ability to retain DNA labels over time. Detection of LRCs has been considered an important method of stem cell identification and is widely used in stem cell research. Recently, LRCs were identified in a variety of cancers (13, 19) and may be the result of either slow-cycling or asymmetric-cell-division segregating the older template DNA strands into daughter stem cells and newly synthesized DNA into daughter cells destined for differentiation (13, 19). In this study, we found that only a small subfraction of thyroid cancer cells were LRCs and that their detection did not include all the asymmetrically dividing cells. In contrast, drug resistance is a common characteristic of CSCs, which forms the basis of tumor recurrence after chemotherapeutic treatments (36). The population of SSEA-1⁺ thyroid cancer cells was enriched after chemotherapy treatment with 5-FU, and the surviving cells showed good expression of stem cell markers, were able to be induced to differentiate into adipocyte-like cells, and formed thyroid tumors after being injected into nude mice. Whether SSEA-1 expression can serve as a prognostic factor needs to be evaluated.

Several studies have demonstrated the existence of thyroid CSCs marked by CD133 expression (2, 5, 26), by aldehyde dehydrogenase expression (3), or as a side population in flow cytometry (10). Together with our results, these data confirm the presence of undifferentiated cells that appear to be resistant to chemotherapy and exhibit stem cell-like behavior. Furthermore, it has recently been shown that EMT plays a critical role in certain epithelial cancer cells acquiring CSC properties (28, 37). EMT is a normal feature of embryonic development, tissue remodeling, and wound healing. Recent studies have demonstrated that EMT plays a critical role not only in tumor metastasis but also in tumor recurrence, which is tightly linked with the biology of CSCs (10). The process of EMT involves disassembly of cell-cell junctions, actin cytoskeleton reorganization, and increased cell motility and invasion, as characterized by down-regulation of E-cadherin and up-regulation of mesenchymal molecular markers such as vimentin, fibronectin, and N-cadherin. Snail, a member of the Snail family of zinc finger transcription factors, is a central mediator of EMT both during development and in tumor progression, inducing EMT by directly repressing E-cadherin (38). EMT has also been demonstrated to play a critical role in thyroid carcinogenesis, and EMT and stem-like properties have been induced by up-regulation of Snail in thyroid cancer (37, 39). Here, the SSEA-1⁺ cells we examined also showed evidence of EMT initiation with enhanced vimentin expression and decreased E-cadherin expression (259.46 and 0.57%, respectively) and with increased Snail expression when compared to SSEA-1⁻ cells, indicating EMT as their likely origin.

In summary, this study demonstrated that human thyroid cancer cells contain SSEA-1-expressing cells with self-renewal ability. These cells generated tumors and were resistant to chemotherapy. The data also provide evidence that cellular diversity in thyroid cancers may occur through both symmetric and asymmetric cell division of cells expressing SSEA-1. Furthermore, it is likely that the asymmetrically dividing cells have arisen via EMT from malignant cells and may be the source of metastatic malignant thyroid tumor formation. These findings provide a new perspective for the therapy of thyroid cancer and point to targeting SSEA-1-expressing cells in the development of a therapeutic strategy.