CD15 and CD15s expression is associated with G1 phase of the cell cycle in glioma cell lines

Overexpression of the tetrasaccharide carbohydrate epitopes, CD15 and CD15s are associated with non-central nervous system malignancies. While CD15 and CD15s expression is rare in gliomas, recent reports suggest that CD15 may serve as a marker for brain tumour ‘stem-like’ cells. The aim of this study was to determine if this apparent discrepancy may, in part, be explained by temporal expression of CD15 and CD15s at different phases of the cell cycle. We used flow cytometry, immunocytochemistry and a fluorescence cell cycle indicator (FUCCI) system to examine expression in glioblastoma (GBM) cells (UP-007 and SNB-19) and non-neoplastic astrocytes (SC-1800) synchronised via serum starvation, Hydroxyurea and Nocodazole, respectively. CD15 and CD15s expression was significantly increased in glioma cells synchronised to G1 phase compared with non-synchronised cells (p<0.001). This was supported by qualitative results obtained with the (FUCCI) system. Few studies have considered the possibility of cell-cycle dependent CD15 and CD15s expression which may explain the inconsistencies reported in the literature in terms of expression in ‘glioma stem-like cells’ where cells are more likely in S phase where CD15 and CD15s expression would be low.

Others have reported that CD15 expression is not a marker for glioma 'stem-like' cells [27]. The aim of this study was to gain a better understanding of CD15 and CD15s expression in GBM. Herein we demonstrate that the expression of CD15 and CD15s is cell cycle dependent in the glioma cell lines tested.

Expression of CD15 and CD15s in non-synchronised brain tumour cells
Expression of CD15 and CD15s were characterised in non-synchronised cell cultures as per the flow chart

Expression of CD15s at different cell cycle stages
CD15s expression also shows cell-cycle dependency. In non-synchronised cells, 2% of SC-1800 cells, 8% of SNB-19 cells and 16% of UP-007 cells were CD15s positive (Supplemental Figure 2b). CD15s was expressed at a lower level in SNB-19 (9.5%) and UP-007 (13.4%) compared to isotype control (p<0.05 and p<0.01). As with CD15, CD15s expression in the astrocyte cell line was low and did not significantly change with synchronisation.
In GBM cells, CD15s expression was significantly higher in cells arrested at G1 phase: 39% and 48% in UP-007 (p<0.001) and SNB-19 (p<0.001) respectively compared to non-synchronised counterparts (Figure 4a and b). In addition, in each of the GBM cell lines there was significantly higher CD15s expression in G1 compared to S (p<0.001) and G2/M (p<0.001). The increase in CD15s expression in glioma cell lines synchronised to G0/G1 was dramatically higher than CD15 expression. In non-synchronised co-localisation studies, CD15s expression (green) was also co-localised with red nuclei indicative of G1 (Figure 4c). These results suggest that expression of CD15s correlates with G1 phase.

Discussion
CD15 and CD15s are fucosylated polysaccharide epitopes and tumour-associated cell adhesion molecules.
Overexpression of both epitopes has been correlated with malignancy of many non-CNS cancers [15,14]. Early studies in CNS tumours reported the absence or low CD15 expression in GBM cells [12,22]. However more recently, expression of CD15 in glioblastomas and specifically with tumour-initiating cells [23,24,28,25,26].
In contrast, Kenney-Herbert et al [27] reported that there were no phenotypic or genetic differences between CD15-and CD15+ GBM cells and CD15 expression was not enough to distinguish a discrete population of GBM cells. Mao et al [24] showed that in 20 GBM cases in which CD15 expression was investigated by immunohistochemistry, 12 cases were considered CD15 positive. In another study where CD15 expression was being investigated as a potential marker for rare extracranial metastases, 5/13 of GBM cases were considered CD15 positive [29]. Few studies have however, considered the possibility of cell-cycle dependent CD15 expression which may explain the inconsistencies reported in the literature. A recent study, however; showed elegantly, data to support the idea of an intrinsic glioma cancer stem cell plasticity which might help to explain the 'inconsistencies' [30].
CD15s expression has been shown to be variable in different non-synchronised GBM cell lines with positivity roughly from 4.8%-32.8% [31]. This agrees with our findings that show the variability in expression in nonsynchronised GBM cell lines (UP-007 and SNB-19). In non-neoplastic astrocytes, CD15 and CD15s expression was low and this did not change when cell lines were synchronised to G1, S, or G2/M consistent with reports showing low CD15 expression in astrocytes [22].
In the GBM cell lines tested, the expression of CD15 and CD15s was significantly higher when cells are synchronised in G1 phase. These findings suggest that expression of both CD15 and CD15s in GBM cells correlate with G1 phase. Our overall findings suggest that in GBM cell lines, CD15 and CD15s expression is correlated with a specific stage of the cell cycle. This may help to explain the conflicting reports in the literature concerning CD15 expression in glioma cells and 'glioma stem-like cells' as differences could be due to majority of the cells being in the S phase where CD15 and CD15s expression would be low. Future studies are needed to address if this is the case. An additional interesting question is whether CD15 and CD15s play a role during cell cycle progression, particularly while cells are in G1. It has been suggested since the late 1980's that cells in G1 were more susceptible to initiate differentiation in response to growth factors [32]. The concept of connecting the cell cycle phase and cell fate as well as the time spent in each phase has gained supportive experimental evidence over the years [reviewed in 33,34]. Recently Singh [35] proposed a hypothetical model of the relationship between cell cycle, pluripotent cells and heterogeneity and that G1 could serve as a 'differentiation induction point' [35] and as proposed by Hardwick et al [33] the length of time 'glioma stem-like cells' are in G1 be targeted to enhance differentiation and slow tumour progression? How does the glioma 'cancer stem cell niche' contribute to cell cycle length? In an interesting proteomic study using a breast cancer cell line arrested in G1, bioinformatic tools and databases, Tenga and Lazar [36] reported that three major clusters of interacting networks emerged and included oxidative phosphorylation, DNA repair and signalling. These reports highlight the complexity associated with labelling a subset of cells. Instead of relying on 'glioma stem-like cell' markers the idea of using molecular regulatory components that act within a network may prove promising in terms of understanding and identification of therapeutic targets for GBM and 'stem-like cells'. Two reports that examined CD15 as a potential glioma 'stemlike' cell also conducted CD15 IHC studies [23,24]. The data presented herein on expression of CD15 being correlated with specific cell cycle phases are based on in vitro studies and future work should include in vivo experiments to determine if this phenotype is the same. Future work should also include a comprehensive temporal and spatial investigation of CD15 and CD15s in GBM and other brain tumour biopsy material. Interestingly, Qazi et al [25] demonstrated that the CD15 positive brain tumour stem-like cell population obtained from primary GBM can be enriched following chemo radiotherapy in vitro and this model may represent the phenotype seen in recurrent GBM. Exploring CD15 and CD15s expression, regulation and function in light of recent findings of CD15 as a potential glioma 'stem-like' marker with this work demonstrating the expression of these molecules are dependent on cell cycle phase, will provide additional valuable insight and hopefully lead to potential new therapeutic targets.

Cell culture
This study included low passage adult human astrocytes (SC-1800) (passage 3-6) derived from the cerebral cortex (Caltag Medsystems, UK) and low and high passage glioblastoma multiforme, Grade IV (GBM) cells, UP-007 (passage 8-13) and SNB-19 (passage 38-42), respectively. UP-007 cells were established 'in house' from biopsies received from surgical resections while SNB-19 was purchased from the DSMZ German Brain Tumour Bank, Germany. All cells were examined for mycoplasma contamination on a routine quality control basis, using MycoAlert™ kit (Lonza, Germany) and genetically authenticated using a STR-PCR fragments kit (Agilent Technologies, USA) as per our previous published technique [37]. SC-1800 cells were grown in astrocyte basal medium (ABM) supplemented with astrocyte single Quots™ (AGM-2) (Lonza, Germany) and 3% human serum

Synchronization of cell cultures at cell cycle specific stages
Expression of CD15 and CD15s were characterised in non-synchronised cell cultures in primary and secondary brain tumour cells as per the flow chart described in Figure 1. Cell cultures were synchronised at G1 phase by serum starvation, at S phase via Hydroxyurea (1mM) and at G2/M phase via Nocodazole (2µg/ml). G1 phase: cells were arrested at G1 phase by serum deprivation. 1x10 6 cells were seeded in T25 tissue culture flasks containing serum supplemented growth medium until 50% confluency was reached. Cells were then washed with pre-warmed sterile Hank`s buffered salt solution (Fisher, UK) followed by addition of growth medium supplemented with 1% human serum followed by overnight incubation. Cells were then grown in serum-free medium for 48-72 hours then replaced every 12 hours to avoid cell cytotoxicity due to the pH change.
S phase: cells were arrested at S phase by first arresting cells at G1 phase then replacing the medium with growth medium supplemented with serum and Hydroxyurea (Sigma, UK) at a final concentration of 1mM and incubated overnight.
G2/M phase: cells were grown in serum-free medium for 24 hours followed by replacement with fresh growth medium supplemented with 2μg/mL Nocodazole (Sigma, UK). Growth factors in the medium induce cells to progress to G2/M phase while Nocodazole causes cell arrest at G2/M phase since Nocodazole depolarizes the tubulin in microtubules (Figure 1).

Detection of cell cycle stage
To determine the distribution of cell cycle stages, non-synchronized and synchronized cell cultures were harvested by gentle scraping. Cellular pellets were washed with Phosphate Buffered Saline (PBS) and fixed with ice-cold 70% Ethanol for 48 hours at 4 o C. Fixed cells were washed with PBS+2% goat serum, resuspended in 250μL of Propidium Iodide/RNase solution (FxCycle™) (Life technologies, UK) and incubated for an hour at room temperature. Cells were then washed with PBS+2% goat serum and cell cycle analysis was conducted using a BD FACS Calibur (BD Biosciences, UK).

Flow cytometry
Non-synchronized and synchronized cells were fixed using 70% Ethanol for 48 hours at 4 o C then washed with PBS+2% goat serum (Sigma, UK) and resuspended in 1mL of ice-cold PBS. Tubes contained approximately 1x10 5 cells. Two tubes served as staining controls (blank and isotype control) and three tubes as positive tests.
Positive samples were incubated in primary antibodies for 1 hour at 4 o C followed by 30 minutes incubation in secondary antibodies. Cells were then incubated in 250μL of Propidium iodide/RNase solution for 1 hour at room temperature prior to flow cytometry analysis. Samples were analyzed using a BD FACS Calibur.

Immunocytochemistry using Premo™ (FUCCI) Cell cycle Sensor (BacMam 2.0)
A fluorescence ubiquitination cell cycle indicator (FUCCI) was used according to the manufacturer`s instructions (Life Technologies, UK) to assess cell cycle progression. Cell lines were transfected with the BacMam 2.0 gene delivery system which combines two main cell cycle regulators: Cdt1-tagged with red fluorescent protein (RFP) and geminin-tagged with green fluorescent protein (GFP). Cells in G0/G1 phase expressed Cdt1-tagged with RFP were visualised as cells with red nuclei in G0/G1 phase. Cells in S phase co-expressed Cdt1-RFP and geminin-GFP and were visualised with yellow nuclei. Geminin-GFP is predominately expressed in cells during G2/M phase allowing cells with green nuclei to be observed. Briefly, 1x10 3 BacMam2.0™ particles were diluted in 200μL serum free Opti-MEM™ (Gibco, UK) followed by addition of 1x10 3 cells and incubated for 10 minutes.
Treated cells were seeded on 10mm sterile coverslips in 48-well plates followed by a 48-hour incubation. Cells were fixed with 4% paraformaldehyde (Sigma, UK) and non-specific antigens were blocked with 10% goat serum (Sigma, UK). CD15 and CD15s primary antibodies were applied for one hour followed by incubation in secondary conjugates for 30 minutes. Cells were counterstained with 10mM Hoechst blue (Sigma, UK).

Confocal microscopy
ICC images were obtained using X40 and X100 (oil immersion) objectives via a Zeiss LSM 510 Meta Axioskop2 confocal microscope using lasers with excitation wavelengths of 405nm (blue), 488nm (green), 568nm (red) and 674 (purple) with diode, argon and HeNe1 lasers respectively. Identical settings were used to image negative controls in which primary antibody was replaced with non-specific Isotype. Semi-quantitative analysis of antigen intensity was measured using Zeiss Zen image software.

Statistical analysis
All experiments were performed in triplicate and data was expressed as +/-SE. Statistical analyses were performed using one-way ANOVA followed by Tukey's multiple comparison post-hoc tests using Graph Pad Prism 6 software.
Author contributions: SAJ, ZM, GJP and HLF designed study and analysed data. SAJ and PC performed experiments. All authors contributed to writing of manuscript. This work was part of SAJ's PhD thesis.