Normal and malignant epithelial cells with stem-like properties have an extended G2 cell cycle phase that is associated with apoptotic resistance

Tissue was collected from biopsies of normal human oral mucosa (n = 5) and HNSCC tumours (n = 11) with written informed patient consent and ethical approval granted by the NE London & The City Ethics Committee. Cultures of normal oral keratinocytes (NOK) were grown as previously described [21]. Tumour tissue was minced into pieces of approximately 1 mm³ and incubated for 20 min at 37°C in PBS containing 2.5 mg/ml collagenase I, 0.25% trypsin, and 0.01 mg/ml DNase (all from Sigma, Dorset, UK). Cells for FACS analysis were used either alive or after fixation in ice-cold 70% ethanol.

Cell lines derived from HNSCC (CA1, H357, 5PT, UK1, CaLH3), breast (MCF7) and prostate (DU145) carcinomas were grown as previously described [19, 21]. Cells were plated at clonal density (2000 cells/ml) and individual colonies arising from single cells examined after 5-10 days for differences in morphology and staining for CD44. For comparison of rates of proliferation, H357 cells plated at clonal density were labelled in the S phase of the cell cycle by addition of 5-bromo-2-deoxyuridine (BrdU, Sigma) to culture media to a final concentration of 50 μg/ml for 2 hours. Colonies developed from plated cells were then fixed in 70% ethanol, and stained with an antibody recognizing BrdU adducts (BD Biosciences, Oxford, UK, Cat# 347583) as previously described [19]. For assays of ability to grow as "tumour spheres" in suspension, CA1, CaLH3, H357 and MCF7 cells were grown at clonal density for 4-5 days, removed from the dish by trypsinisation, and sorted to obtain populations of CD44^high and CD44^low cells. For MCF7 and H357 cell lines, the sorted CD44^highand CD44^low cell subpopulations were then resuspended in conditioned medium with 0.4% low melting point agarose (Sigma) to a cell density of 2000 cells per ml. Cells were plated in triplicate 6 well plates pre-coated with a layer of 1% agarose and left to grow for 2-3 weeks. To quantify sphere formation, 10 random fields of view were taken for each well and the average number of spheres per field calculated. For CA1 and CaLH3 cell lines, the sorted CD44^high and CD44^low cell subpopulations were plated as single cells in non adherent 48-well plates and left to grow for 3 weeks to grow for sphere formation. Three plates were quantified for each cell line and each cell sub-population, and the data is presented as % of the wells that contained spheres.

CA1, H357, 5PT, UK1, CaLH3, MCF7, DU145 and NOK cells, were plated at clonal density and after 7-10 days apoptosis was induced by exposure to doses of UVB from 1 to 40 mj/cm², to Cisplatin (50-1000 μM), to neocarzinostatin (NCZ) (10-100 ng/ml), to tumour necrosis factor (TNF, 10 ng/ml), or to etoposide (10-750 μM) (all from Sigma) for 24 or 48 h. Unsorted cells of the CA1, H357 and 5PT cell lines growing in situ at clonal density were examined by phase contrast microscopy and were also stained for cleaved Caspase 3 and Annexin V as markers of apoptosis (see below). To assess effects of Chk1 and Chk2 inhibition, debromohymenialdisine (DBH, Axxora, Nottingham, UK) was added to the growth media of the unsorted cells of the Ca1, 5PT, CaLH2 and CaLH3 cell lines grown at clonal density to a final concentration of 8 micromol/ml and cells analysed 18 hours later. Chk1 and Chk2 were also individually down-regulated by transfection with either Chk1, Chk2 or non-targeting siRNAs (ON-TARGETplus SMART pool siRNA, Dharmacon, Inc. Chicago, IL) in unsorted cells of the CA1 and 5PT cell lines with cells stained and assayed 48 h after transfection.

Sorted cells of the CA1 cell line (CD44^high and CD44^lowsubpopulations) were plated at the same density, exposed to UV (10 mj) and NCZ (20 ng/ml), left 4 days to recover and then the number of floating (dead) cells was counted. In another experiment, sorted cells of the CA1 cell line (CD44^high and CD44^low subpopulations) were plated at similar density, exposed to apoptosis-inducing drugs for 48 hours, and then trypsinized and plated again at clonal density. The number of colonies formed was counted in 5 different microscopic fields.

Unsorted cultured cells of the CA1, H357, 5PT, CaLH2, CaLH3, MCF7, and DU145 cell lines were fixed in 4% formalin and immunostained as previously described [21]. Primary antibodies were anti-CD44 (1:100, BD Biosciences, Oxford, UK) and anti-cleaved-Caspase-3 (1:250, Promega, Southampton, UK). For Annexin V staining, unsorted cells of the Ca1, H357 and NOK cell lines were grown on culture dishes, washed with Annexin V binding buffer (BD Biosciences) with 1% FBS, incubated for 10 min at 37°C with 10 μl FITC-Annexin V (BD Biosciences) per mL in the same buffer, washed, fixed in 4% paraformaldehide, and mounted in VectaShield-DAPI (Vector laboratories Inc, Burlingame, CA).

For assessment of apoptotic sensitivities of CD44^low and CD44^highcell fractions, cells isolated from fresh HNSCC samples were quadruple stained with anti-epithelial-specific-antigen antibody (anti-ESA-APC), anti-CD44 antibody (anti-CD44-PE), AnnexinV-FITC, and DAPI. This allowed distinction of tumour cells from stromal cells, identification of CD44^high and CD44^low cells, and their further classification into early (AnnexinV⁺ DAPI^-) and late (AnnexinV⁺ DAPI⁺) apoptotic fractions. IgG2ak-PE was used as isotype control. For cell cycle analysis, the cells isolated from fresh HNSCC were fixed and triple stained with anti-ESA-APC antibody, anti-CD44-PE antibody, and PI. Typically, the 3-5% of cells with the highest CD44 expression was designated as CD44^high cell subset and the remainder of the population was designated as CD44^low cell subset. Cell populations from CA1, H357, 5PT, UK1, CaLH3, MCF7, DU145 and NOK cell lines were triple stained with anti-CD44-PE, AnnexinV-FITC, and DAPI for assessment of apoptosis, or double stained with anti-CD44-FITC and PI for cell cycle assays. To assess apoptosis by loss of mitochondrial membrane potential, a DilC₁(5) assay (MitoProbe™ Invitrogen, Scotland) was used together with CD44 in a double staining procedure according to manufacturers instructions for CA1, H357 and MCF7 cell lines. To confirm the presence of a G2 block, CaLH3 and H357 cells exposed to various apoptotic-inducing agents were also stained with DAPI, anti-CD44 antibody and anti-Cyclin B1 antibody (H-433, Sc-752, Santa Cruz, Wiltshire, UK).

To assess cell progression from G2 into G1, unsorted cells of the CaLH3 and H357 cell lines were pulsed with 10 μM 2-Deoxy-5-Iodouridine (IdU) for 20 mins before washing 3 times in PBS and adding fresh medium. Immediately and at 3, 5, 7, and 9 hours, cells were detaching from the flask with Accutase, washed and pelleted, and fixed in ice cold 70% ethanol. Cells were then co-stained with anti-CD44-PE and FITC-conjugated antibody recognizing IdU adducts (BD Pharmingen, Cat# 347583), and processed for FACS analysis. Unless otherwise stated, all FACS reagents were from BD Biosciences (Oxford, UK) and cells were examined using Becton Dickenson LSRII equipment and analysed with FACS Diva software.

Sorted CD44^high and CD44^low cells of the CA1 and H357 cell lines, or rapid adherent and late adhering cells of the CA1 and CaLH3 cell lines [15] were isolated after treatment with apoptosis-inducing agents and lysed using RIPA Buffer (50 mM Tris pH 7.3, NaCl, 0.1% SDS, 1%NP40) containing 1% protease inhibitor (Roche Diagnostics, Lewes, UK) and 1% phosphatase inhibitor (Sigma). Protein content was determined using Bio-Rad protein assay (Bio-Rad Laboratories, Hercules, CA, USA). Ten μg total cell protein was mixed with SDS-PAGE buffer and resolved using 4-12% NuPAGE b/Bis Tris Gels (Invitrogen, Renfrew, UK). Proteins were transferred to polyvinylfluoride (PVDF) membranes (Hybond-C Extra, Amersham Pharmacia Biotech, Little Chalfont, UK). Primary antibodies were diluted 1:500 for anti-Chk1 pS317 (Bethyl Lab, TX, USA), 1 μg/ml anti-Chk2 pT26 (Abcam, Cambridge, UK), 1:1000 anti-β-actin (Abcam), and 1:10,000 anti-GAPDH (Abcam).

Colonies of unsorted cells with holo- or para-clone morphologies were laser micro-dissected from the CA1, UK1, and 5PT cell lines, and CD44^high and CD44^low expressing cells were FACS sorted from primary cultures of human normal oral keratinocytes (NOK). Total RNA was extracted using CellsDirect One-Step qRT-PCR kit, and cDNA generated using SuperScript III first-strand synthesis supermix for qRT-PCR (both from Invitrogen). Q-PCR used a Brilliant SYBR Green kit (Stratagene, La Jolla, CA, USA) and MX3005P apparatus with between 50 and 100 ng of cDNA probed using the following primers: Wee1 - forward: gct ctg tta aac tcc ggg gta, reverse: gac act gtc ctg agg aat gaa g; Myt1 - forward: caa ccg agg ctg tcg aga aa, reverse: ctg tcc aca cca tag ccc t; GAPDH forward: gtg aac cat gag aag tat gac aac, reverse: cat gag tcc ttc cac gat acc.

Numerical data are shown as mean ± SD and are based on a minimum of 3 repeats for each independent biological sample (fresh tissue or cell line). Data was analysed using SPSS version 15 statistical program (SPSS Inc. Chicago, IL, USA), Students t-tests, and analysis of variance (general linear model) with post-hoc tests (Bonferroni multiple comparisons).

Each of the seven human carcinoma cell lines investigated in this study (CA1, H357, 5PT, CaLH2, CaLH3, MCF7, and DU145) developed colonies with morphologies that corresponded, as previously described, to holoclone, meroclone and paraclone colonies (Fig. 1A, 1D). Holoclone colonies showed strong expression of CD44 (Fig. 1B, 1E). When examined by FACS after removal from culture dishes, the cell populations showed various gradients of cell staining for CD44, with distinct sub-populations of CD44^high cells found in some cell lines (Fig. 1F). A similar pattern could be observed when analysing carcinoma cells isolated from fresh biopsies of HNSCC (Fig. 1G and 1H). Normal oral keratinocytes showed differential levels of staining for CD44, although they did not show discrete sub-populations of CD44^high cells as seen for some malignant cells (Fig. 1I). Sorting cells for high or low levels of expression of CD44 showed, as previously indicated for carcinoma-derived cells [21], that the CD44^high cell fractions are more clonogenic when plated in standard culture conditions and form a greater proportion of colonies with holoclone morphologies for both malignant and normal keratinocytes (Fig. 1J, 1K and1L, respectively). Plating the CD44^high and CD44^low cancer cell fractions as single cells in suspension indicated that the CD44^high subpopulation of cells forms a markedly greater proportion of tumour spheres (Fig. 1N and 1O), a property related to tumour initiating ability [22, 30].

Bild vergrößern

Following UVB irradiation, counts of cells within individual colonies of unsorted HNSCC cell lines indicated significantly greater cell loss from paraclone colonies than from holoclones (Fig. 2A and 2B) and higher number of cells positive for Annexin V (Fig. 2C) and cleaved-Caspase 3 in paraclones and meroclones than in holoclones. Labelling with BrdU confirmed previous observations [19, 31] of high levels of cell proliferation in both holoclone and paraclone colonies (Fig. 2D and 2E), indicating that increased survival of cells in holoclones was not simply due to proliferative quiescence. Quantitative assessment of apoptosis by triple-staining with anti-CD44-PE, Annexin V-FITC, and DAPI and FACS analysis showed detectable (baseline) levels of apoptosis in control (un-stimulated) cultures that significantly increased on exposure to UVB, Cisplatin, NCZ, TNF or Etoposide (Fig. 2H, 2I and Table 1). In each of the HNSCC, breast and prostate carcinoma cell lines, the CD44^high cell fraction showed significantly lower rates of apoptosis as compared to the CD44^low subset (Table 1). Stimulation for various time lengths and concentrations of UVB, Cisplatin, NCZ, or Etoposide consistently induced a time- and dose- dependent apoptosis in the general population, but the rates of apoptosis in CD44^high populations remained proportionately lower (Fig. 2F, 2G and Table 1). The proportion of CD44^high cells remaining after apoptotic challenge was typically increased (Table 1), indicating the selective survival of this fraction. More significant levels of cell death could be observed in the CD44^high fraction of cells when the CaLH3 cells were exposed to very high doses apoptosis-inducing drugs. However, the cell death in this fraction was still at a couple of folds lower levels than in the CD44^low fraction of cells (Fig. 2K), displaying the same pattern of resistance to apoptosis-inducing drugs even at these very high doses. DilC₁(5) release due to the loss of the mitochondrial membrane potential in cells undergoing apoptosis was used as an alternative method to assess cell death, and it also indicated a clear pattern of resistance for the CD44^high cell fractions to apoptosis-inducing stimuli (Fig. 2J). Significantly higher numbers of floating (dead) cells were found 4 days after apoptosis-inducing treatment in the pre-sorted CD44^low cells than in the pre-sorted CD44^high cells of the CA1 cell line, indicating the increased resistance of the CD44^high fraction of cells long term after treatment as well (Fig. 2L). Sorted cells of the CA1 cell line (CD44^high and CD44^low fractions) were exposed to drugs for 48 hours, then trypsinized and plated again at clonal density. CD44^high fraction of the cells exposed to NCZ gave rise to significantly higher number of colonies than the CD44^low fraction (Fig. 2M, 2N), although for the exposure to UV this was not significantly different.

Bild vergrößern

FACS analysis after staining the cell populations for CD44 and DNA content indicated significant cell cycle differences between CD44^high and CD44^low cells (Fig. 3A and Table 1). In un-stimulated cells, the proportion of cells in the G2 phase of the cell cycle was consistently and significantly higher for the CD44^high subset of cells than for the CD44^low cells (Table 1), at the expense of G1 phase of the cell cycle (Fig. 3). Treatment with NCZ or Cisplatin induced small increases in the fraction of CD44^low cells in G2 but significant increases in the proportion of CD44^high cells (Fig. 3C). Exposure to increased concentrations of NCZ showed that the rates of apoptosis correlated inversely with the extent of the G2 block in the CD44^high cells (Fig. 3D). For the H357 and CaLH3 cell lines the proportion of cells present in the G2 phase was also assessed by an alternative method of quantifying the expression of Cyclin B1, a molecule predominantly expressed in G2/M phase. These cell lines showed a significantly greater (2-3 fold greater, p < 0.01) number of cell expressing Cyclin B1 in the CD44^high cell fractions (Fig. 4A).

Bild vergrößern

Bild vergrößern

Cells freshly isolated from specimens of HNSCC were examined immediately and at 24 and 48 h after treatment with Cisplatin or NCZ. Negative staining for ESA was used to exclude non-epithelial cells (i.e., fibroblasts, inflammatory cells, etc.) from the analysis. All samples showed a range of CD44 staining levels and some specimens showed distinct sub-populations of 3-5% of cells with high CD44 staining (Fig. 1D and 1E). The rates of apoptosis were relatively higher for the un-treated carcinoma cells isolated from fresh tumour biopsies then for un-treated cultured carcinoma cell lines, but when the CD44^high and CD44^low cell fractions were separately analysed, the rates of apoptosis were found to be consistently lower for the CD44^high cell fractions than for the CD44^low fractions (Fig. 3B, Table 2). This difference was statistically significant (p < 0.05) for 6 of the 8 individual tumour specimens examined for apoptosis. FACS analysis after staining cell populations for CD44 and DNA content indicated significant cell cycle differences between CD44^high and CD44^low un-treated cells in all tumours (Fig. 3B, Table 2), with up to a 2-fold greater number of CD44^high cells in the G2 phase of the cell cycle and lower cell numbers in the G1 phase of the cell cycle. Treatment with NCZ or Cisplatin (Fig. 3B) consistently increased overall apoptotic rates, but the apoptotic rates for CD44^high cells invariably remained a couple of folds lower than for the rest of the population. After apoptotic challenge an increased fraction of CD44^low cells was also found in G2, but consistently greater increases were found for CD44^high cells. Following apoptotic stimulation by exposure to Cisplatin and NCZ, the size of the remaining CD44^high cell fractions typically increased, suggesting again selective survival of the CD44^high cells (Table 2).

The greater proportion of CD44^high cells observed to be present in G2 could result from a general increase in the time taken for the CD44^high cells to transit G2 or from a more permanent block of a sub-fraction of the CD44^high cell population. To assess the time taken to transit G2, cultures of the H357 and CaLH3 cell lines were exposed to IdU to label the cells in the S phase and the time taken for these cells to transit through G2 and return to G1 was then determined by sequential FACS analysis of cell cohorts (Fig. 4B, C, D). Five hours after IdU labelling, a higher proportion (15.6 ± 2.3% and 11.6 ± 6.2% for H357 and CaLH3 respectively) of CD44^low cells (Fig. 4C) was found to return to G1 as compared to CD44^high cells (4.8 ± 2.3% and 3.9 ± 2.0% respectively, p < 0.05) (Fig. 4D), indicating that although both CD44^high and CD44^low cells transit G2, CD44^high cells do so more slowly. Results from four data series for each cell line estimate the return of CD44^high cells to G1 to take 1.5 -2.0 hr longer than for CD44^low cells even in un-treated control cultures (Fig. 4E and 4F).

Comparison of all data sets for CD44^high and CD44^low cells in variously-treated individual cell lines indicated that lower cell death was associated with larger fractions of CD44^high cells present in the G2 phase of the cell cycle (Tabel 1 and Fig. 3D). G2-cell cycle arrest is often associated with phosphorylation of the Chk1 and Chk2 check-point proteins and both the CD44^high and CD44^low cells, and the rapid and late adhering cells of un-stimulated populations contained low levels of phosphorylated Chk1 and Chk2 (Fig. 5A). After treatment with NCZ, their levels increased but the CD44^high and the rapid adhering cell fractions showed a higher increase, suggesting greater checkpoint activation in CD44^high or rapid adhering cells in response to DNA damage. Treatment with DBH, a broad inhibitor of Chk1/2 checkpoint actions, resulted a significant decrease in the percentages of CD44^high cells in the G2 phase of the cell cycle, and this reduction was associated with significantly (p < 0.05) increased rates of death of CD44^high cells after exposure to UV (Fig. 5B). Targeting Chk1 with siRNA significantly (p < 0.05) decreased the fraction of CD44^high cells in G2 and increased their rate of cell death but little effect was seen after targeting Chk2 (Fig. 5C). QPCR showed that CD44^high cell fractions or cells laser captured from holoclones had higher levels of transcription of Wee1 and Myt1, the up-stream molecules involved in activation of the G2 checkpoint response (Fig. 5D).

Bild vergrößern

Murine epithelial stem cells appear highly susceptible to apoptosis [26, 27] but there is little information about differential apoptotic sensitivities of stem and amplifying fractions of normal human oral keratinocytes (NOK). To examine whether the greater apoptotic resistance of malignant CD44^high cells is related specifically to malignant change, or is a general a property of human epithelial stem cells, low passage cultures (P3-4) of NOK isolated from normal human oral mucosa were examined (n = 5). Analysis of the 3-5% of NOK cells with the highest CD44 expression showed that these cells, as for the CD44^high malignant cells, were more resistant than the remainder of the population to apoptotic induction and spent greater time in the G2 phase of the cell cycle (Fig. 6). Baseline levels of apoptosis increased markedly in NOK cultures after exposure to UVB, Cisplatin, NCZ, and Etoposide (Fig. 6 and Table 2) but the proportion of apoptotic cells in CD44^high cell fractions consistently remained lower and the time spent in G2 became further extended.

Bild vergrößern