There is increasing evidence that cancers contain their own stem-like cells called cancer stem cells (CSCs). A small subset of cells, termed side population (SP), has been identified using flow cytometric analysis. The SP cells have the ability to exclude the DNA binding dye, Hoechst33342, and are highly enriched for stem cells in many kinds of normal tissues. Because CSCs are thought to be drug resistant, SP cells in cancers might contain CSCs. We initially examined the presence of SP cells in several human thyroid cancer cell lines. A small percentage of SP cells were found in ARO (0.25%), FRO (0.1%), NPA (0.06%), and WRO (0.02%) cells but not TPC1 cells. After sorting, the SP cells generated both SP and non-SP cells in culture. The clonogenic ability of SP cells was significantly higher than that of non-SP cells. Moreover, the SP prevalence was dependent on cell density in culture, suggesting that SP cells preferentially survived at lower cell density. Microarray experiment revealed differential gene expression profile between SP and non-SP cells, and several genes related to stemness were up-regulated. However, non-SP population also contained cells that were tumorigenic in nude mice, and non-SP cells generated a small number of SP cells. These results suggest that cancer stem-like cells are partly, but not exclusively, enriched in SP population. Clarifying the key tumorigenic population might contribute to the establishment of a novel therapy for thyroid cancer.
THERE IS INCREASING evidence that cancers contain a small subset of their own stem-like cells called cancer stem cells (CSCs) (1, 2). CSCs can self-renew to generate additional CSCs and also differentiate to generate phenotypically diverse cancer cells with limited proliferative potential. The existence of CSCs might explain why there are so many recurrences, even after tumor disappeared completely by radio- or chemotherapy. Because current cancer therapeutics have been developed based on killing differentiated cancer cells, some cancer cells, possibly being CSCs, probably survive the treatment.
Identifying CSCs was initially accomplished in the context of acute myeloid leukemia. Several groups demonstrated that the cells capable of initiating leukemia in immunodeficient mice were found in CD34+CD38− fraction (3–5). Multiple myeloma might also have a CSC population. The majority of the myeloma cells express CD138; however, a minority of CD138− cells preferentially engrafted in immunodeficient mice and also generated CD138+ cells (6).
Recently such CSCs have been also isolated in several types of solid tumors. Al-Hajj et al. (7) demonstrated that breast cancer cells with CD44+CD24−/lowlineage− surface phenotype were tumorigenic in immunodeficient mice. Importantly, as few as 100 tumorigenic cells were able to form tumors in the mice, whereas 105 of CD44+CD24+ cells were not able to form tumors. These tumorigenic cells were serially passaged in the mice, and each time the generated new tumors contained additional CD44+CD24−/lowlineage− tumorigenic cells as well as phenotypically mixed nontumorigenic cancer cells. Further evidence supporting an existence of CSCs in solid tumors came from the studies of brain tumors (8). Singh et al. (9, 10) reported that only CD133+ brain tumor fraction contained cells that were capable of tumor initiation in immunodeficient mouse brains. CD133 is a marker expressed in normal neural stem cells, suggesting that the normal neural stem cell is the target of transforming events that lead to brain tumors. These cells have been identified by the expression of their unique surface markers. However, most tissue-restricted stem cells lack unique and specific expression markers, including those in thyroid.
A small subset of cells, termed side population (SP), has been identified using flow cytometric analysis (11–13). The SP cells have the ability to exclude the DNA binding dye, Hoechst33342, and are highly enriched for stem cells in many kinds of normal tissues. It has been shown that the SP phenotype is maintained by breast cancer-resistant protein-1 (BCRP1 or ABCG2), one of ATP-binding cassette transporters that are associated with multidrug resistance in many cancers by pumping out the drugs. Because CSCs are thought to be drug resistant, SP cells in cancers might contain CSCs. Indeed, the SP analysis has been recently used in an attempt to isolate CSCs from several types of cancers, and CSCs seem to be enriched in the SP population (14–17). However, there are some reports demonstrating that SP cells lack stem cell characteristics (18–20), although primitive stem cells are enriched in the SP in most cases. Little is known about stem cell or CSC in thyroid.
Anaplastic thyroid carcinomas provide less than 8 months of mean survival after diagnosis (21). These tumors are highly resistant to standard therapeutic procedures such as surgical treatment, radiation therapy, and chemotherapy. The identification of CSCs in such thyroid cancers has important therapeutic implications because the elimination of this key population might lead to improved outcome.
In this study, we identified SP cells in several thyroid cancer cell lines and characterized the population in comparison to main non-SP population.
Materials and Methods
Cell lines and reagents
Human anaplastic thyroid carcinoma cell lines, ARO and FRO, papillary thyroid carcinoma cell lines, NPA and TPC-1, and follicular carcinoma cell line, WRO, were initially provided by Dr. James Fagin (University of Cincinnati, Cincinnati, OH). All cell lines were grown in RPMI 1640 supplemented with 5% fetal bovine serum and penicillin/streptomycin at 37 C in a humidified atmosphere with 5% CO2. Hoechst33342 and verapamil were purchased from Sigma (St. Louis, MO).
Flow cytometry for SP cells
The cells were detached by trypsinization and washed with ice-cold PBS/2% FBS. Cells (1 × 106) were labeled in the growth medium with 5.0 μg/ml Hoechst33342 dye either alone or in combination with 50 μg/ml verapamil at 37 C for 90 min. After washing with PBS/2% FBS, the cells were then incubated with 2 μg/ml propidium iodide to exclude dead cells. SP analysis and sorting were done using a FACSVantage SE (BD Biosciences, San Jose, CA). The Hoechst dye was excited with UV laser and its fluorescence was measured with both 675/20 filter (Hoechst Red) and 424/44 filter (Hoechst Blue).
Cell proliferation assay
The cells were grown in 96-well plate and the relative cell number was determined using Cell Counting Kit-8 (Dojindo, Kumamoto, Japan) according to the manufacturer’s protocol. Briefly, 10 μl of WST-8 solution was added to each well and the plate was incubated for 1 h at 37 C. For quantification, OD was read at 450 nm using a microplate reader ImmunoMini NJ-2300 (System Instruments, Tokyo, Japan).
Nude mouse xenograft model
Animal care and experimental procedures described in this study were done in accordance with the Guidelines for Animal Experiments of Nagasaki University with the approval of the Institutional Animal Care and Use Committee (no. 0202060039). The cells were resuspended in growth medium (∼50 μl) and mixed with an equal volume of Matrigel (BD Biosciences). The mixed cell suspension was injected sc into 8-wk-old female BALB/c nu/nu mice (Charles River Japan, Tokyo, Japan).
GeneChip analysis
Total RNA was extracted from approximately 30,000 SP or non-SP cells using an RNeasy micro kit (QIAGEN, Valencia, CA) according to the manufacturer’s instruction. The RNAs were subjected to GeneChip expression array full service with two-cycle target labeling (Biomatrix Research Inc., Chiba, Japan). Briefly, first- and second-strand cDNA was synthesized from total RNA using T7-oligo(dT) primers (first cycle). After first in vitro transcription, second cycle first- and second-strand cDNA was generated. Biotinylated cRNA was then synthesized by second in vitro transcription. Fifteen micrograms of the labeled cRNA were hybridized to a human genome U133 Plus 2.0 array (Affymetrix, Santa Clara, CA). Array image was scanned and analyzed with GeneChip operating software (Affymetrix). After normalizaion, data mining was done using GeneSpring software (Agilent Technologies, Palo Alto, CA). Probes scored as absent call in both SP and non-SP were excluded from data analysis.
Real-time RT-PCR
Approximately 150 ng of total RNA were reverse transcribed using a SuperScript III first-strand synthesis supermix (Invitrogen, Carlsbad, CA) with random hexamers. The following PCR amplifications were performed using SYBR Premix Ex Taq Perfect real-time kit (TaKaRa Bio, Ohtsu, Japan) in a thermal cycler Dice real-time system (TaKaRa Bio). The cycle threshold value, which was determined using second derivative, was used to calculate the normalized expression of the indicated genes using Q-Gene software (22), using β-actin as a reference gene. The following primer pairs were used: β-actin, 5′-ctgaaccccaaggccaaccgcg-3′ and 5′-ggcgtacagggatagcacagcc-3′; ABCG2, 5′-ctggataaagtggcagactccaag-3′ and 5′-agccagttgtaggctcatccaag-3′; MYC, 5′-tgctctcctcgacggagtcc-3′ and 5′-tccacagaaacaacatcgatttcttc-3′; JUN, 5′-aaaaggaagctggagagaatcgc-3′ and 5′-tgttccctgagcatgttggc-3′; FZD5, 5′-gggactgtctgctcttctcg-3′ and 5′-ccgtccaaagataaactgcttc-3′; HES1, 5′-tctggaaatgacagtgaagcacct-3′ and 5′-gttcatgcactcgctgaagcc-3′; JAG1, 5′-ggtcttgcaaactcccaggtg-3′ and 5′-actgccagggctcattacagatg-3′.
Statistical analysis
Differences between groups were examined for statistical significance using one-way ANOVA followed by Fisher’s protected least significant difference or unpaired t test as appropriate. P value not exceeding 0.05 was considered statistically significant.
Results
Identification of SP cells in thyroid cancer cell lines
Several groups have already reported the presence or absence of SP cells in a variety of established cancer cell lines (14–17). We first sought to determine the presence of SP cells in five widely used thyroid cancer cell lines: two anaplastic cancer cell lines ARO and FRO; two papillary cancer NPA and TPC1; and one follicular cancer WRO. The SP was identified by its characteristic fluorescent profile in dual-wavelength analysis and presented as a distinct tail from the main population on the dot plot graph (11). As shown in Fig. 1, SP cells were identified in ARO (0.25%), FRO (0.1%), NPA (0.06%), and WRO (0.02%) but not TPC1 cells. In each case, the SP population was markedly diminished by the treatment with verapamil, which is the inhibitor of the pumps responsible for exclusion of the Hoechst dye, indicating that this population was truly SP. To further investigate the function of CSCs in anaplastic cancer, we concentrated on ARO cells, which possessed the highest percentage of SP cells.
Identification of SP cells in various thyroid cancer cell lines. Indicated cell lines were plated 1 d before analysis to achieve approximately 30% confluence on the day of analysis. The cells were detached and labeled with Hoechst33342 in the presence or absence of verapamil, and at least 1 × 105 cells were analyzed by flow cytometry. The SP cells, which disappeared in the presence of verapamil, are outlined. The percentage of SP cells is also shown. Similar results were obtained in at least two independent experiments.
SP cells can generate both SP and non-SP cells
To examine whether SP cells can generate both SP and non-SP cells, SP cells were sorted (Fig. 2) and cultured in vitro. After 10 d expansion, the cells were restained with Hoechst33342 and reanalyzed by flow cytometry. The SP cells repopulated both SP and non-SP cells (Fig. 2). The ratio of the SP cells to non-SP cells was still much higher than before sorting. The number of cells in the particular area pointed out by the asterisk was also increased, suggesting that those cells are SP cells from G2/M population. At the same time, sorted non-SP cells from G1 main population yielded cells mostly in main population, yet a small number of cells in SP area were also observed (Fig. 2).
SP cells can generate both SP and non-SP cells. ARO cells were labeled with Hoechst33342 and analyzed by flow cytometry. Outlined SP cells and non-SP cells were sorted and cultured separately (left panel). After 10 d culture, SP cells and non-SP cells were stained with Hochst33342 and analyzed by flow cytometry again (middle and right panels). The asterisk indicates possible SP cells from G2/M cell cycle.
Clonogenic ability of SP cells is higher than that of non-SP cells
We next examined clonogenic ability of SP and non-SP cells. The sorting gates were basically same as in Fig. 2. As shown in Fig. 3A, SP cells were more clonogenic than non-SP cells. There was no difference between non-SP cells and total ARO cells (control). We also investigated the short-term proliferative activity after sorting. Twenty-four hours after sorting, the viability of SP and non-SP cells was identical (data not shown, approximately >90%). The growth of SP cells during additional 72 h was slightly slower than that of non-SP cells (Fig. 3B).
SP cells have higher clonogenic ability. ARO cells were labeled with Hoechst33342 and analyzed by flow cytometry. A, One hundred SP, non-SP, and total ARO cells were sorted separately and directly put into each well of a six-well plate using a CloneCyt option (BD Biosciences). The plates were cultured for 2 wk, and the number of colonies in each well was counted. Each bar indicates the mean and sd of six wells. The data are representative of three independent experiments. *, P < 0.05 vs. non-SP or total ARO. B, Ten thousand SP and non-SP cells were separately sorted and seeded in each well of a 96-well plate. After 24 h incubation, 1000 cells were replated into a different well and cultured for an additional 72 h. Relative cell number was measured as described in Materials and Methods. Each bar indicates the mean and sd of at least three wells. The data are representative of three independent experiments. *, P < 0.05 vs. SP.
The prevalence of SP cells depends on culture density
It is generally accepted that plating efficiency of cells drops when the cells are plated at low density. Based on the finding that SP cells have high cloning ability, we investigated the relationship between the SP rate and culture density. As shown in Fig. 4, the prevalence of SP cells was apparently increased when cultured at low density in all cell lines except TPC1. TPC1 seemed to be a really SP-deficient line.
The prevalence of SP cells depends on culture density. The indicated cell lines were cultured for 2 wk as follows: high density, when reached approximately 90% confluence, one fifth of the cells were passaged to a new dish; low density, when reached approximately 10% confluence, one fifth of the cells were passaged to a new dish. All cells were passaged 1 d before analysis. The cells were detached and labeled with Hoechst33342 in the presence or absence of verapamil, and at least 1 × 105 cells were analyzed by flow cytometry. The percentage of SP cells is shown.
Tumorigenesis of SP cells in nude mice
We next explored in vivo tumorigenic ability of SP cells. Viable SP or non-SP cells at G1 phase were sorted separately (the sorting gates were same as in Fig. 2) and injected with matrigel into nude mice. Tumor formation was evaluated 4 wk after injection. Unexpectedly, tumors were formed in most cases regardless of cell type (Table 1). Even though two tumors with large size (+++) were from SP cells and two tumor-negative cases were from non-SP cells, there was no statistical significance between any groups.
Tumorigenesis of SP/non-SP cells in nude mice
| SP/non-SP | Injected cell number | Tumor size |
|---|---|---|
| Mouse 1 | ||
| SP | 1,000 | + (2 mm) |
| 10,000 | + (2 mm) | |
| Non-SP | 1,000 | + (3 mm) |
| 10,000 | ++ (6 mm) | |
| Mouse 2 | ||
| SP | 1,000 | + (2 mm) |
| 10,000 | +++ (12 mm) | |
| Non-SP | 1,000 | TN |
| 10,000 | ++ (4 mm) | |
| Mouse 3 | ||
| SP | 1,000 | +++ (8 mm) |
| 10,000 | + (2 mm) | |
| Non-SP | 1,000 | + (3 mm) |
| 10,000 | + (3 mm) | |
| Mouse 4 | ||
| SP | 1,000 | + (2 mm) |
| 10,000 | + (3 mm) | |
| Non-SP | 1,000 | + (3 mm) |
| 10,000 | TN |
| SP/non-SP | Injected cell number | Tumor size |
|---|---|---|
| Mouse 1 | ||
| SP | 1,000 | + (2 mm) |
| 10,000 | + (2 mm) | |
| Non-SP | 1,000 | + (3 mm) |
| 10,000 | ++ (6 mm) | |
| Mouse 2 | ||
| SP | 1,000 | + (2 mm) |
| 10,000 | +++ (12 mm) | |
| Non-SP | 1,000 | TN |
| 10,000 | ++ (4 mm) | |
| Mouse 3 | ||
| SP | 1,000 | +++ (8 mm) |
| 10,000 | + (2 mm) | |
| Non-SP | 1,000 | + (3 mm) |
| 10,000 | + (3 mm) | |
| Mouse 4 | ||
| SP | 1,000 | + (2 mm) |
| 10,000 | + (3 mm) | |
| Non-SP | 1,000 | + (3 mm) |
| 10,000 | TN |
TN, Tumor-negative.
Tumorigenesis of SP/non-SP cells in nude mice
| SP/non-SP | Injected cell number | Tumor size |
|---|---|---|
| Mouse 1 | ||
| SP | 1,000 | + (2 mm) |
| 10,000 | + (2 mm) | |
| Non-SP | 1,000 | + (3 mm) |
| 10,000 | ++ (6 mm) | |
| Mouse 2 | ||
| SP | 1,000 | + (2 mm) |
| 10,000 | +++ (12 mm) | |
| Non-SP | 1,000 | TN |
| 10,000 | ++ (4 mm) | |
| Mouse 3 | ||
| SP | 1,000 | +++ (8 mm) |
| 10,000 | + (2 mm) | |
| Non-SP | 1,000 | + (3 mm) |
| 10,000 | + (3 mm) | |
| Mouse 4 | ||
| SP | 1,000 | + (2 mm) |
| 10,000 | + (3 mm) | |
| Non-SP | 1,000 | + (3 mm) |
| 10,000 | TN |
| SP/non-SP | Injected cell number | Tumor size |
|---|---|---|
| Mouse 1 | ||
| SP | 1,000 | + (2 mm) |
| 10,000 | + (2 mm) | |
| Non-SP | 1,000 | + (3 mm) |
| 10,000 | ++ (6 mm) | |
| Mouse 2 | ||
| SP | 1,000 | + (2 mm) |
| 10,000 | +++ (12 mm) | |
| Non-SP | 1,000 | TN |
| 10,000 | ++ (4 mm) | |
| Mouse 3 | ||
| SP | 1,000 | +++ (8 mm) |
| 10,000 | + (2 mm) | |
| Non-SP | 1,000 | + (3 mm) |
| 10,000 | + (3 mm) | |
| Mouse 4 | ||
| SP | 1,000 | + (2 mm) |
| 10,000 | + (3 mm) | |
| Non-SP | 1,000 | + (3 mm) |
| 10,000 | TN |
TN, Tumor-negative.
SP cells are enriched by sequential sorts
Repeated sorting of SP cells gradually increased the proportion of SP cells. After four or more consecutive sorts, the SP rate reached 20–40% (Fig. 5). Interestingly, the SP rate did not return to initial level (around 0.2%) but stayed at 5–20% up to 2 months after the last sorting (data not shown). Serial sorting of SP cells seemed to accumulate the cells expressing ABCG2 constitutively (irreversible expression), suggesting the development of mechanism similar to drug resistance.
SP cells are enriched by sequential sorts. SP cells were serially sorted and cultured four times at intervals of approximately 10 d. The cells were then labeled and analyzed by flow cytometry.
Gene expression profile of SP cells
To study the differential gene expression profiles between SP and non-SP cells, we used oligonucleotide-based DNA microarrays, GeneChip Human Genome U133 Plus 2.0 array (Affymetrix).
Of 54,675 probe sets, 25,583 were scored as marginal or present call (not absent) in either SP or non-SP cells and applied to further analysis. Of the 25,583, 2417 probe sets were up-regulated (fold change > 2.0) and 1009 were down-regulated (fold change < 0.5).
To validate the microarray data, we performed quantitative real-time RT-PCR for six interesting genes (ABCG2, MYC, JUN, FZD5, HES1, and JAG1) which were up-regulated in the microarray data. ABCG2 has been shown to be responsible for SP phenotype. Wnt and Notch signaling pathways play important roles in normal stem cells. MYC, JUN, and FZD5 are related to Wnt signaling, and HES1 and JAG1 are related to Notch. Quantitative real-time RT-PCR revealed that all the six genes were up-regulated in SP cells (Fig. 6). The fold-change of FZD5 in microarray data were 33.63, which was far more than RT-PCR data. A possible explanation for this is that the flag of FZD5 expression was scored as absent in non-SP cells and present in SP cells, and thereby the calculation of normalization was not accurate. In terms of the other five genes, the flags in both SP and non-SP cells were all present.
Validation of microarray data using quantitative real-time RT-PCR. ARO cells were labeled with Hoechst33342 and analyzed by flow cytometry. SP and non-SP cells were separately sorted and total RNAs were extracted. Quantitative real-time RT-PCR for indicated genes was performed as described in Materials and Methods. Each bar indicates the mean and sd of the data collected in triplicate. Fold increase in microarray data are shown below the graph.
Chiba et al. (17) recently reported a gene expression profile of SP in two hepatocellular carcinoma cell lines, Huh7 and PLC/PRF/5. Sixty-two genes were commonly up-regulated in both cell lines. Those up-regulated genes were compared with our microarray data. Sixteen genes overlapped between the two studies are listed in Table 2.
Overlapping genes up-regulated in SP cells
| Gene symbol | Gene description | Probe set | Fold change |
|---|---|---|---|
| FOSL | FOS-like antigen 2 | 228188_at | 3.89 |
| TMF1 | TATA element modulatory factor 1 | 213024_at | 2.08 |
| TCERG1 | Transcription elongation regulator 1 (CA150) | 229706_at | 2.28 |
| PF6 | Hypothetical protein FLJ34497 | 1557593_at | 5.45 |
| HELZ | Helicase with zinc finger domain | 240486_at | 2.14 |
| XRCC5 | X-ray repair complementing defective repair in Chinese hamster cells 5 | 233007_at | 2.01 |
| FGF18 | Fibroblast growth factor 18 | 231382_at | 2.92 |
| CYP2E1 | Cytochrome P450, family 2, subfamily E, polypeptide 1 | 209975_at | 7.57 |
| NCOA7 | Nuclear receptor coactivator 7 | 225344_at | 3.68 |
| F2RL1 | Coagulation factor II (thrombin) receptor-like 1 | 213506_at | 2.42 |
| RTN4 | Reticulon 4 | 1556049_at | 2.28 |
| TM4SF1 | Transmembrane 4 superfamily member 1 | 238168_at | 5.07 |
| AKR1C1 | Aldo-keto reductase family 1, member C1 | 1555854_at | 5.82 |
| PRSS3 | Protease, serine, 2 (trypsin 2) | 205869_at | 2.73 |
| EST | 1564257_at | 3.28 | |
| EST | 1561989_at | 2.69 |
| Gene symbol | Gene description | Probe set | Fold change |
|---|---|---|---|
| FOSL | FOS-like antigen 2 | 228188_at | 3.89 |
| TMF1 | TATA element modulatory factor 1 | 213024_at | 2.08 |
| TCERG1 | Transcription elongation regulator 1 (CA150) | 229706_at | 2.28 |
| PF6 | Hypothetical protein FLJ34497 | 1557593_at | 5.45 |
| HELZ | Helicase with zinc finger domain | 240486_at | 2.14 |
| XRCC5 | X-ray repair complementing defective repair in Chinese hamster cells 5 | 233007_at | 2.01 |
| FGF18 | Fibroblast growth factor 18 | 231382_at | 2.92 |
| CYP2E1 | Cytochrome P450, family 2, subfamily E, polypeptide 1 | 209975_at | 7.57 |
| NCOA7 | Nuclear receptor coactivator 7 | 225344_at | 3.68 |
| F2RL1 | Coagulation factor II (thrombin) receptor-like 1 | 213506_at | 2.42 |
| RTN4 | Reticulon 4 | 1556049_at | 2.28 |
| TM4SF1 | Transmembrane 4 superfamily member 1 | 238168_at | 5.07 |
| AKR1C1 | Aldo-keto reductase family 1, member C1 | 1555854_at | 5.82 |
| PRSS3 | Protease, serine, 2 (trypsin 2) | 205869_at | 2.73 |
| EST | 1564257_at | 3.28 | |
| EST | 1561989_at | 2.69 |
Overlapping genes up-regulated in SP cells
| Gene symbol | Gene description | Probe set | Fold change |
|---|---|---|---|
| FOSL | FOS-like antigen 2 | 228188_at | 3.89 |
| TMF1 | TATA element modulatory factor 1 | 213024_at | 2.08 |
| TCERG1 | Transcription elongation regulator 1 (CA150) | 229706_at | 2.28 |
| PF6 | Hypothetical protein FLJ34497 | 1557593_at | 5.45 |
| HELZ | Helicase with zinc finger domain | 240486_at | 2.14 |
| XRCC5 | X-ray repair complementing defective repair in Chinese hamster cells 5 | 233007_at | 2.01 |
| FGF18 | Fibroblast growth factor 18 | 231382_at | 2.92 |
| CYP2E1 | Cytochrome P450, family 2, subfamily E, polypeptide 1 | 209975_at | 7.57 |
| NCOA7 | Nuclear receptor coactivator 7 | 225344_at | 3.68 |
| F2RL1 | Coagulation factor II (thrombin) receptor-like 1 | 213506_at | 2.42 |
| RTN4 | Reticulon 4 | 1556049_at | 2.28 |
| TM4SF1 | Transmembrane 4 superfamily member 1 | 238168_at | 5.07 |
| AKR1C1 | Aldo-keto reductase family 1, member C1 | 1555854_at | 5.82 |
| PRSS3 | Protease, serine, 2 (trypsin 2) | 205869_at | 2.73 |
| EST | 1564257_at | 3.28 | |
| EST | 1561989_at | 2.69 |
| Gene symbol | Gene description | Probe set | Fold change |
|---|---|---|---|
| FOSL | FOS-like antigen 2 | 228188_at | 3.89 |
| TMF1 | TATA element modulatory factor 1 | 213024_at | 2.08 |
| TCERG1 | Transcription elongation regulator 1 (CA150) | 229706_at | 2.28 |
| PF6 | Hypothetical protein FLJ34497 | 1557593_at | 5.45 |
| HELZ | Helicase with zinc finger domain | 240486_at | 2.14 |
| XRCC5 | X-ray repair complementing defective repair in Chinese hamster cells 5 | 233007_at | 2.01 |
| FGF18 | Fibroblast growth factor 18 | 231382_at | 2.92 |
| CYP2E1 | Cytochrome P450, family 2, subfamily E, polypeptide 1 | 209975_at | 7.57 |
| NCOA7 | Nuclear receptor coactivator 7 | 225344_at | 3.68 |
| F2RL1 | Coagulation factor II (thrombin) receptor-like 1 | 213506_at | 2.42 |
| RTN4 | Reticulon 4 | 1556049_at | 2.28 |
| TM4SF1 | Transmembrane 4 superfamily member 1 | 238168_at | 5.07 |
| AKR1C1 | Aldo-keto reductase family 1, member C1 | 1555854_at | 5.82 |
| PRSS3 | Protease, serine, 2 (trypsin 2) | 205869_at | 2.73 |
| EST | 1564257_at | 3.28 | |
| EST | 1561989_at | 2.69 |
Discussion
In the present study, we characterized the SP cells in anaplastic thyroid cancer cells. First, we examined five widely used thyroid cancer cell lines. Among them, ARO, FRO, NPA, and WRO possessed SP subpopulation but TPC1 did not. The rate of SP cells in ARO and FRO was higher than that in NPA and WRO. This was somewhat correlated with their tumorigenicities in nude mice. ARO and FRO form tumor very rapidly in the mice, whereas NPA and WRO grow relatively slower and TPC1 is hardly engrafted (in our hands, injection of up to 2 × 107 TPC1 cells collected from 60% confluent did not result in tumor formation, data not shown). This implies that the SP prevalence might be associated with tumor aggressiveness. To our knowledge, however, there are no reports showing significant correlation. Patrawala et al. (16) compared five breast cancer cell lines with a variety of tumorigenic abilities. Among them, MCF7 was the only cell line containing SP population, despite its low tumorigenicity. SP was undetectable in the other four lines including highly tumorigenic MDA-MB231 cells. Further studies are needed to clarify the relationship between the SP prevalence and tumor aggressiveness.
Although the concept of CSCs is not new, it is only recently that advances in stem cell biology have given the impetus to CSC theory (1, 2). Like somatic stem cells, CSCs have the properties of self-renewal, producing heterologous descendent cells and slow cell cycling. In this work, we showed that SP cells were capable of self-renewal and also generated non-SP cells by asymmetric division. This is completely consistent with previous studies showing that SP cells can generate non-SP cells in vitro or in vivo (14–17). In addition, the cell growth of SP cells was slightly slower than that of non-SP cells during 96 h after sorting. Note that the SP cells probably divided asymmetrically and produced progenitor-like cells, which might be rapidly cycling, implying that the determination of SP cell growth was not exact but rather overestimated.
The clonogenic ability of SP cells was higher than that of non-SP cells. This is consistent with the study by Patrawala et al. (16) using MCF7 cells. The difference is not likely to be a consequence of longer retention of Hoechst dye in non-SP cells because the viability of SP and non-SP cells after sorting was identical and the growth of non-SP cells was even faster. However, we cannot completely rule out the possibility that the difference is due to some effect of the Hoechst dye, which is potentially cytotoxic (23).
This study, for the first time, demonstrates that the SP prevalence depends on cell density in culture. Presumably, the SP cells preferentially survived at very low plating density, which is a severe condition for cells. Kondo et al. (14) reported that MCF7 and Hela cells contained 2.0 and 1.2% SP cells, respectively. On the other hand, Patrawala et al. (16) identified 0.2% SP cells in MCF7 and did not detect SP in Hela cells. The latter group presumed that the difference was due to variation of flow cytometer and staining protocol. Our findings provide a new possible explanation. It is necessary to take into account culture condition when seeking SP population in certain cell lines. If SP cells are not identified in some type of cells that have been cultured at very low density, it is most likely that such cells do not possess SP population, like TPC1 cells.
Comprehensive analysis of gene expression using microarray chip revealed distinct differences between SP and non-SP cells. It is noteworthy that we sorted SP and non-SP cells from G1 population to avoid the effect of cell cycle-specific changes. A few thousand genes were up- or down-regulated in SP cells, indicating that ARO cells are not homogeneous. We next confirmed several interesting gene expressions by quantitative real-time RT-PCR. Wnt signaling is thought to be closely associated with stem cell self-renewal (24). MYC and JUN are important downstream components of Wnt pathway. In addition, Notch signaling maintains survival of stem cells and also inhibits differentiation in certain cell types (25–27). JAG1 is a ligand and HES1 is a transcription factor regulated by Notch signaling. In SP cells, these genes were up-regulated, suggesting that SP cells retain some stem-like properties. We also compared our microarray data with those by Chiba et al. (17). They isolated SP cells from Huh7 and PLC/PRF/5 cells. In their study, as few as 1 × 103 SP cells were sufficient for tumor formation in nonobese diabetic/severe combined immunodeficiency mice, whereas injection of 1 × 106 non-SP cells did not yield tumors. Sixteen genes were commonly up-regulated in the two studies (Chiba et al. and our study). These genes probably play very important roles in maintenance of CSC properties. To our knowledge, however, none of them have been demonstrated to be involved in thyroid cancer. This can be explained by the fact that SP cells account for only 0.2%, even though they exhibited clearly distinct gene expression patterns from the bulk of cancer cells (main population), which have been used in previous studies for many years.
Previous studies using C6 glioma cells (14), LAPC9 prostate tumor cells (16), MCF7 breast tumor cells (16), U373 glioma cells (16), Huh7 hepatoma cells (17), PLC/PRF/5 hepatoma cells (17), and MOVCAR7 ovarian tumor cells (28) have demonstrated that SP cells are more tumorigenic in vivo using immunodeficient mice. In our xenograft experiment, all injections (eight of eight) with SP cells formed tumors. Regarding non-SP cells, however, six tumors were unexpectedly formed from eight injections. SP cells seemed to be more tumorigenic; however, there was no significant difference. Besides, non-SP cells generated a small number of SP cells (Fig. 2, right panel). Moreover, even though SP cells were more clonogenic than non-SP cells, the clonogenic ability of non-SP cells was not zero (Fig. 3A). These results suggest that tumorigenic cells are probably enriched in SP population, but non-SP main population also contains a small number of tumorigenic cells. Although we found possible SP cells from G2/M population (Fig. 2, middle panel), cancer stem cell is thought to be quiescent and should be in G0/G1 phase. Patrawala et al. (16) proposed the hypothesis in which the ABCG2− population contains primitive CSCs with higher self-renewal and proliferative potentials but slow cycling. These cells then generate ABCG2+ progenitor cells that are more actively proliferating but possess reduced self-renewal and long-term proliferative capacities. The ABCG2+ progenitor cells eventually give rise to ABCG2− fully differentiated tumor cells with limited proliferative potential. Our results are supportive of this hypothesis.
In conclusion, we have identified, for the first time, SP cells in several human thyroid cancer cell lines. Cancer stem-like cells are enriched in the SP population but not exclusively, indicating that CSCs are not identical with SP cells. Our results also suggest that CSCs that have distinct properties might be present even in established thyroid cancer cell lines, which have been cultured in the presence of high concentration of serum for many years. However, it should be noted that there is a possibility that highly clonogenic SP cells might be derived from cells with genomic instability cultured for many years. Further studies are needed to identify thyroid cancer stem cells precisely. The elucidation of key tumorigenic population might contribute to not only clarification of mechanism of thyroid carcinogenesis but also the establishment of a novel therapy for anaplastic thyroid cancer.
This work was supported in part by Grant-in-Aid for Scientific Research (18790637, 18591030, and 18590335) from the Ministry of Education, Culture, Sports, Science, and Technology of Japan.
Disclosure Statement: The authors have nothing to disclose.
Abbreviations:
