Identification of CD24 as a marker of Patched1 deleted medulloblastoma-initiating neural progenitor cells

High morbidity and mortality are common traits of malignant tumours and identification of the cells responsible is a focus of on-going research. Many studies are now reporting the use of antibodies specific to Clusters of Differentiation (CD) cell surface antigens to identify tumour-initiating cell (TIC) populations in neural tumours. Medulloblastoma is one of the most common malignant brain tumours in children and despite a considerable amount of research investigating this tumour, the identity of the TICs, and the means by which such cells can be targeted remain largely unknown. Current prognostication and stratification of medulloblastoma using clinical factors, histology and genetic profiling have classified this tumour into four main subgroups: WNT, Sonic hedgehog (SHH), Group 3 and Group 4. Of these subgroups, SHH remains one of the most studied tumour groups due to the ability to model medulloblastoma formation through targeted deletion of the Shh pathway inhibitor Patched1 (Ptch1). Here we sought to utilise CD antibody expression to identify and isolate TIC populations in Ptch1 deleted medulloblastoma, and determine if these antibodies can help classify the identity of human medulloblastoma subgroups. Using a fluorescence-activated cell sorted (FACS) CD antibody panel, we identified CD24 as a marker of TICs in Ptch1 deleted medulloblastoma. CD24 expression was not correlated with markers of astrocytes or oligodendrocytes, but co-labelled with markers of neural progenitor cells. In conjunction with CD15, proliferating CD24+/CD15+ granule cell precursors (GCPs) were identified as a TIC population in Ptch1 deleted medulloblastoma. On human medulloblastoma, CD24 was found to be highly expressed on Group 3, Group 4 and SHH subgroups compared with the WNT subgroup, which was predominantly positive for CD15, suggesting CD24 is an important marker of non-WNT medulloblastoma initiating cells and a potential therapeutic target in human medulloblastoma. This study reports the use of CD24 and CD15 to isolate a GCP-like TIC population in Ptch1 deleted medulloblastoma, and suggests CD24 expression as a marker to help stratify human WNT tumours from other medulloblastoma subgroups.


Introduction
Medulloblastoma is the most common malignant brain tumour in children. Despite recent advances in the treatment of this disease the 5-year survival rate remains at approximately 70%, and a significant number of patients suffer from long-term side effects including cognitive impairments and growth retardation. One major developmental pathway associated with medulloblastoma formation is the Sonic hedgehog (Shh)/Patched 1 (Ptch1) pathway. Ptch1 functions as an antagonist of the Shh pathway through suppression of the transmembrane protein Smoothened (Smo). Proper interaction between Shh and Ptch1 is critical to maintain normal Smo activity, which mediates the expression of the Gli transcription factors, and ultimately proper embryonic development [1]. Loss of Ptch1 has been attributed with tumour formation in many organs, including the skin [2] and liver [3], and in the brain, excessive Shh pathway activity has been well documented to be causative for medulloblastoma [4].
Recently, medulloblastoma have been classified into four subgroups: WNT, SHH, Group 3 and Group 4 that differ in their ontogeny, demographics and clinical outcomes [5,6]. The SHH subgroup shows the greatest incidence in infants (younger than three years of age), patients older than 16 years of age, and is largely attributable to mutations in PTCH1, SUFU and SMO genes [7][8][9][10]. While progress has been made in uncovering the cells of origin of medulloblastoma, the identification and targeting of the tumour initiating cells (TICs) remains a work in progress. The cancer stem cell hypothesis postulates that the TIC is a relatively rare cell that is responsible for tumour initiation, propagation and therapy resistance [11,12]. Recently, it was reported through the use of murine models of medulloblastoma that a cerebellar stem cell (SC) is a TIC population in Ptch1 deleted medulloblastoma [13]. Other medulloblastoma studies have also identified granule cell precursors (GCPs) as a cell of origin of medulloblastoma [4,[14][15][16][17]. Owing to the heterogeneous nature of medulloblastoma, a means to selectively identify the tumorigenic cell population prior to oncogenesis represents an important goal towards improving outcomes for this disease. Fluorescent-Activated Cell Sorting (FACS) has been used to identify and purify putative neural stem cells [18][19][20][21], but the ability to identify TICs with stem-like properties remains a difficult process largely due to the inherent limitation of TIC markers to disseminate these cells from normal neural stem cells. Nevertheless research has had a lot of success with utilising FACS to identify neural stem-like cells in the murine and human brain [22][23][24][25][26][27][28].
Current evidence for TICs in medulloblastoma is largely reliant on antibodies against cell surface receptors termed clusters of differentiation (CD) that allow for the stratification and selection of live cell populations. Given that similarities exist between normal stem cells and cancer stem cells, research has sought to utilise CD antibodies to identify cell populations that have both stem cell traits and tumour forming capabilities. In the neural field, CD133 and CD15 have been reported to label stem cells and GCPs, respectively, which can initiate tumorigenesis in Ptch1 deleted models of medulloblastoma [26,29,30]. CD133 has been reported to label glioma-derived stem cells [29,31], while CD15 has been shown to label a more lineage-specific neural progenitor population [21,25,32]. Recently, CD133 has been identified as a marker of tumour invasiveness in medulloblastoma [28]. Nevertheless, CD133 and CD15 do not appear to be sufficient to disseminate all TICs in medulloblastoma from the resident stem and progenitor cells.
Previous studies have undertaken various screens of CD antibodies through immunohistological methods [25]. These studies have provided seminal research identifying novel CD markers but are hampered by the inability to analyse large numbers of cells, and thus identify rarer TICs. To this end, we designed a flow cytometric screen of approximately 50 commercially available CD antibodies that have been characterised in the haematopoietic system. We chose to work with cells derived from the GFAP cre mediated Ptch1 conditional mutant mouse (Ptch1 lox/lox ;GFAP cre ), which develop medulloblastoma with 100% incidence by four weeks of age [13,[33][34][35]. Using this model, we aim to investigate the expression profiles of the CD antibodies and determine whether their expression profiles can be used to predict tumorigenicity, independently, or in combination with other markers. Following successful identification in the Ptch1 conditional model, we aim to determine if similar expression profiles can be observed in human medulloblastoma subgroups, with the intent of identifying novel CD antibody markers to further profile human medulloblastoma.

Animal models and human tissue specimens
This study was carried out in strict accordance with the Australian Code of Practice for the Care and Use of Animals for Scientific Purposes. All experimental and breeding procedures done throughout this study was approved by the University of Queensland animal ethics committee (protocol numbers IMB/588/08/BREED; IMB/092/10/BREED; IMB/444/08/BREED; IMB/526/08; IMB/254/10(NF); IMB/526/08; IBC/489/IMB/2007). All animals were bred and maintained in the Institute for Molecular Bioscience animal facility at the University of Queensland, and all animals used throughout the study were annually reported to the ethics committee. All surgery was performed under Ketamine/Ilium-Xylazil anaesthesia and all efforts were made to minimize suffering, including the administration of external warming, analgesics and constant animal monitoring until full recovery. Genetically modified mice and tumour-transplanted mice were monitored daily for their health and all animals were euthanized at the first sign of distress or when a tumour was visible. Death as an endpoint was never investigated in this study no tumour-transplanted animals died as a result of tumour-induced suffering or disability. The number of mice utilised in all primary tissue studies (CD antibody assays/tumour assays) ranged between 3 and 15 for each experiment. For histological analyses approximately 3-5 animals were used for each experiment. The number of animals used for any given experiment are shown as an n-value (biological repeat), throughout the study.
The Ptch1 lox/lox ;GFAP cre mouse line used in this study was described previously [33]. In Ptch1 lox/lox ;GFAP cre mice tumour visibility was evident between postnatal days 17-22 (collectively termed P17+ in the manuscript). SCID mice receiving in vivo transplantation of tumour cells were culled upon immediately visualising a subcutaneous or intracranial tumour. P17 + aged Ptch1 lox/lox ;GFAP cre mice and adult SCID mice were culled via CO 2 asphyxiation while P7 mice were culled via cranial amputation, as ethically required by the Australian Code of Practice for the Care and Use of Animals for Scientific Purposes. Mice were genotyped using 2 PCR protocols: One PCR for the transgenic LoxP sites using the primers 5'-CCACCAGT GATTTCTGCTCA-3' and 5'-AGTACGAGGCATGCAAGACC-3' and the other PCR for the GFAP-cre transgene using the primers 5'-ACTCCTTCATAAAGCCCTCG-3' and 5'-ATCACTCGTTGCATCGACCG-3. All human tumour specimens were serially collected in accordance with the ethics review board of the NN Burdenko Neurosurgical Institute (Moscow, Russia) between 1995 and 2007 as described [5,6,36,37].

FACS antibodies
The CD mouse antibodies used in the antibody screen were the following CD antibodies conjugated with PE from BD Biosciences: CD1d (553846) cycle analysis, all cells were stained with DAPI (4',6-Diamidino-2-Phenylindole, Dihydrochloride; 1/10,000 concentration; Thermo Fisher Scientific, Massachusetts, USA) for 2 minutes in 1mL PBS followed by three 5-minute PBS washes. CD antibody gated cells were measured using a violet (405 nm) laser to capture any DAPI nuclear signal, and cell cycle analysis later performed using ModFit LT 5 software (Verity, Maine, USA). Live imaging of labelled cells was performed using an Amnis Flow cytometer. All flow cytometry and sorting of cells were undertaken at the ACRF Brain Tumour Centre at Queensland Brain Institute, the University of Queensland.

Transplantation experiments
Orthotopic injections were carried out on severe combined immunodeficient (SCID) mice (6-12 weeks old). Injections of tumour cells into the murine brain were performed in two different regions, the striatum as described by [40], and the cerebellum as described by [25]. Briefly, cells were dissociated and placed into 2μL or 5μL of DPBS, for intra-striatal and intra-cerebellar injections, respectively. SCID mice were anaesthetised with Ketamine (Lyppard, Victoria, Australia, KETAI1), Ilium Xylazil-20 (Lyppard XYLAZIL20) and Acepromazine (Lyppard ACEMAV2) diluted in a 3:3:1 ratio in 1mL phosphate buffered saline and administered at an effective dose of 100μL/100g body weight via intra-peritoneal injection. Once anaesthetised, mice were positioned correctly in a stereotaxic frame with a mouse adaptor (KOPF instruments, California, USA) and an incision was made along the midline of the scalp to expose the underlying skull and a small hole was made at appropriate coordinates from bregma: For intra-striatal injections: anterior-posterior = 0; medio-lateral = +2.5 mm; for intra-cerebellar injections, a small hole was made 1mm lateral to the midline above the cerebellum. Using a 5μL Hamilton Syringe (Hamilton) with a 24-gauge bevelled needle, 2-5μL of cells (in DPBS) were injected (dorso-ventral = -3.5 mm) and cells released over a 5 minute period. The skin was closed using Vetbond adhesive (Coherent Scientific, South Australia, AU). Pain relief was achieved by administration of Torbugesic (20μL/100g; Lypard) preoperatively and Baytril (40μL/100g; Lypard) was given as a post-surgical antibiotic.
All efforts were made to reduce post-operative suffering including external warming and active monitoring until fully active with no signs of distress. Mice were monitored daily until a tumour was visibly identified or the animals showed any signs of distress, at which time they were euthanized. If no visible tumours were identified after approximately one year, all transplanted animals were euthanized and autopsied for tumour formation.

Immunohistochemistry and tissue micro arrays
For histological analyses brains were perfused in 4% PFA via heart perfusion, embedded in opitimal cutting temperature ((OCT): Sakura, Alphen aan den Rijn, The Netherlands, 25608-930) and sectioned using a Leica microtome in 10μm sections. Hemotoxylin and Eosin stainings were performed per standard methods. For Immunofluorescent labelling of cryosections primary antibodies were blocked in PBS containing 10% horse serum, 1% Bovine Serum Albumin (Sigma) and 0.1% TritonX and incubated overnight at 4 degrees Centigrade. Secondary antibody incubations were at room temperature for 1 hour. In situ hybridisations were done as per [41] using an in situ probe generated against murine CD24. Immunohistochemistry was performed immediately post in situ hybridisation as per Vectastain Elite avidinbiotin complex method instructions (Vector Laboratories, California, USA) and detection was carried out with 3,3'-diaminobenzidine reagent (Vector Laboratories). For co-immunofluorescent labelling of FACS sorted cells, post-sorted cells were pipetted onto a charged coverslide (Dako, California, USA, K802021) and cytospun using a benchtop centrifuge with a cytospin rotor. Cells were then immunolabelled as per standard immunofluorescent techniques. Cells and sections were visualised and counted using a Zeiss fluorescence microscope.

RNA isolation and real time analysis
RNA was isolated from cells using a QIAgen RNAeasy mini kit (Qiagen, Hilden, Germany, 74104) with DNAse digestion and cDNA synthesised using SuperscriptIII RT (Invitrogen 18080-044). For quantitative mRNA expression detection cDNA was assayed using Taqman Universal PCR master mix (Applied Biosystems, California, USA, 4304437). Assay on Demand primers were used for detection: Gli1 (Applied Biosystems, Mm00494645_mL), Nmyc (Applied Biosystems, Mm00476449_mL). GAPDH was used as a housekeeping control (Life Technologies 4352339E). The qPCRs were performed on a 7000 Sequence Detection System (Applied Biosystems) and analysis performed using ABI prism 7000 SDS software. To confirm Ptch1 deletion in tumour cells, and thereby prove that the tumours were derived from in vivo transplanted Ptch1 lox/lox ;GFAP cre cells, we utilised Exon2/6 primer PCR. Using 200ng cDNA synthesised from extrapolated RNA, PCR was performed using Exon 2 (5'-CACCGTA AAGGAGCGTTACCTA-3') and Exon 6 (5'-TGGTTGTGGGTCTCCTCATATT-3') specific primers.

Statistical analysis
All analyses were done using biological repeats and numbers are represented throughout the manuscript as 'n'. For murine medulloblastoma analysis comparisons of antibodies, cell counts, or cell cycle differences between two populations were analysed using the student's t-test with welches correction. Grouped antibody analysis was carried out using one-way ANOVA. Survival statistics were calculated using Mantel-Cox tests. Murine isolated cell counts were done using Adobe Photoshop (Adobe, California, USA). Flow cytometry analyses were undertaken using FlowJo (FlowJo, Oregon, USA). For human medulloblastoma analysis, expression of CD24 and CD15 (FUT4) was assessed using the R2 software (http://r2.amc.nl) in eight independent gene expression cohorts [5,6,44,45,[47][48][49]. Associations between gene expression and subgroup affiliation were evaluated using one-way ANOVA. Evaluation of human CD24 immunostaining was performed in a semi-quantitative manner. Analysis of IHC experiments was performed by a positive versus negative evaluation of stained cores as assessed by investigators blinded to clinical and molecular variables. Subgroup specific expression was determined using Chi Square statistics. For all analyses p-values < 0.05 were considered to be statistically significant. Graphs and statistical analyses were done using Prism (GraphPad software, California, USA). Artwork was done using Adobe Illustrator (Adobe).

CD antibody screening of Ptch1 lox/lox ;GFAP cre cerebella identified unique expression profiles compared to wild type cerebellar cells
To identify TIC markers in medulloblastoma we utilised the Ptch1 conditional mouse model [33]. GFAP-cre [35] mediated Ptch1 deletion (Ptch1 lox/lox ;GFAP Cre ) results in an aggressive medulloblastoma of the cerebellum that begins during postnatal neurogenesis and leads to a 100% mortality rate by approximately three weeks of age [13]. For the purpose of ethically analysing medulloblastoma in this mouse model, animals were investigated prior to severe symptoms that result in mortality (P17+). While age-equivalent Ptch1 lox/lox wild type cerebella were investigated as controls, P7 Ptch1 lox/lox granule cell precursors (GCPs) and P17+ Ptch1 lox/lox ; GFAP cre isolated neurospheres were also investigated for comparisons with stem and progenitor cells [13,25].
FACS screening and one-way ANOVA analysis on the four models tested identified a number of CD antibodies with differential expression profiles, including CD1d (p<0.0001), CD15 (p = 0.097), CD24 (p<0.0001), CD38 (p<0.0001), CD81 (p = 0.0002), and CD117 (p = 0.0031) ( Table 1). In addition to their identification in this screen, previous studies have reported these antibodies to be of interest in tumour identification. While not identified in the primary screen, CD133 has previously been a focus of CD expression research on neural tumours [25,28,29,50] and so its tumour identifying potential on Ptch1 lox/lox ;GFAP cre medulloblastoma cells was also investigated. CD117 has previously been shown to not label Ptch1 deleted murine medulloblastoma [25] but we wanted to investigate its potential as a marker in Ptch1 lox/lox ;GFAP cre medulloblastoma. Initial tumour transplantation studies of CD117 sorted cells resulted in tumours from both positive and negative fractions at greater than 50 days post cell transplantation: CD117+ tumours, n = 1/5 at 83 days post injection; CD117-tumours, n = 3/5 at 59-83 days post injection (S1 Fig). CD1d, CD38, and CD81 showed interesting expression profiles in the primary screen but have not been identified as markers of medulloblastoma and so they, along with CD117's low expression in Ptch1 deleted cells, were not further investigated in this study. In contrast, CD15 and CD133 have been reported to be expressed on Ptch1 deleted medulloblastoma and have previously been identified as tumour initiation markers in both medulloblastoma and gliomas [25,28,29,50]. In addition, CD24 has been reported to be expressed on human medulloblastoma [51], but its expression on murine models of medulloblastoma has yet to be fully investigated. To this end we selected CD24, CD15 and CD133 to investigate as TIC markers in Ptch1 lox/lox ;GFAP cre medulloblastoma.
To better identify what cell(s) CD24 expression associates with, we co-stained murine tissue with markers of neural cells. Co-staining with the Purkinje neuron marker Calbindin identified co-labelling with CD24 throughout the Purkinje layer of WT cerebella, but not with CD24 + cells in the IGL of Co-immunohistochemistry/in situ hybridisation for the S-phase marker PCNA identified discrete "islands" of CD24+/PCNA+ cells in Ptch1 deleted medulloblastoma, while adult wild type cerebella were predominantly PCNA-/CD24- (S2I and S2J Fig). These results indicate that CD24 is not a specific marker for mature astrocytes, or oligodendrocytes, but does show some co-labelling with oligodendrocyte progenitor cells (OPCs) and Purkinje neurons within the Purkinje layer.

CD24 labels a tumour-initiating cell population in Ptch1 deleted medulloblastoma
To determine whether CD24 labels TICs in medulloblastoma, we investigated the tumour propagating potential of CD24+ and CD24-Ptch1 lox/lox ;GFAP cre cell populations. CD24 expression has previously been shown to label proliferating, progenitor-like cells of the cerebellum [20,59,60] so we hypothesised that CD24 would label TIC's in Ptch1 lox/lox ;GFAP cre medulloblastoma. Subcutaneous injections of 1x10 6 CD24+ and CD24-purified cell populations recapitulated tumour formation in vivo, with resulting tumours appearing as a highly vascularised tumour mass of nucleated cells (Fig 2A and 2B). Statistical analysis revealed that CD24+ cells gave rise to tumours faster than CD24-cells at 80 days post injection (black arrow Fig 2A; p = 0.027), but with similar total tumour incidences after 300 days (average CD24 + 72,57+/-24.45 days; CD24-99.25+/-14.92 days; p = 0.53). Based on these findings we postulated that the CD24-TIC may represent a rarer stem-like cell compared to a more common CD24+ tumour-initiating neural progenitor, and that with transplantations of decreasing cell concentrations, the CD24-TIC could theoretically be diluted to a point where insufficient TICs are present to induce tumour formation. To test this hypothesis, we transplanted smaller concentrations of CD24 cell populations into the subcutaneous flanks of SCID mice. Injections of 4x10 5 CD24+ primary Ptch1 lox/lox ;GFAP cre medulloblastoma cells resulted in 6 tumours from 15 transplantations (95.17+/-19.51 days) while equivalent numbers of CD24-cells gave 0 tumours following 9 injections (p = 0.021, Fig 2C). Reverse transcriptase PCR and quantitative CD24 identifies Patched1 deleted medulloblastoma-initiating cells PCR analysis of the resulting CD24+ tumours confirmed Ptch1 deletion and identified no significant difference in Gli1 and Nmyc expression levels compared with Ptch1 lox/lox ;GFAP cre medulloblastoma (Fig 2D, 2E and 2F). IHC/immunofluorescence analysis of CD24+ and CD24-cell-initiating tumours identified diffuse CD24 expression throughout both tumour populations (Fig 2G and 2H) that did not co-label with GFAP or Olig2+ (Fig 2I-2L). This data suggests that CD24 expression is correlated with higher tumour initiation frequencies, but is not a definitive marker of Ptch1 deleted medulloblastoma TICs.

CD133 is of limited utility as a marker to identify stem-like cells within Ptch1 deleted medulloblastoma
Given its published identification as a putative marker of stem/progenitor cells, we investigated the tumour propagating potential of CD133 and its co-labelling potential with CD24. CD133 expression was similar between P7 wild type GCPs (10.25% +/-1.83) and P17+ Ptch1 lox/lox ; GFAP cre medulloblastoma (7.

CD24 expression is high on human Group 3, Group 4 and SHH medulloblastoma compared with WNT tumours and normal cerebella
Having shown that the expression of CD24 is significant in our murine model we sought to ascertain its relevance in human medulloblastoma and its four subgroups: WNT, SHH, Group 3 and Group 4, whose stratification is based on tumour location and genetic inception [62]. While CD24 expression has previously been shown to be elevated in human medulloblastoma [51] there have been until now no studies examining the expression of CD24 between the subgroups. We identified elevated CD24 levels in all human medulloblastoma subgroups compared to normal cerebellum and normal brain tissue controls in 8 non-overlapping, independent gene expression-profiling studies (p<0.001, Fig 5A and 5B) [5,42,44,45,[63][64][65]. Significantly higher CD24 expression levels were observed in SHH (p<0.0001), Group 3 (p = 0.0049) and Group 4 (p<0.0001) medulloblastoma subgroups compared with WNT subgroups (Fig 5A). A reduction in CD24 expression was observed in Group 3 compared to SHH (p = 0.0004), while no differences were identified between Group 4 and SHH (p = 0.59). Compared with fetal cerebellum, minor increases in CD24 expression were observed in SHH (p = 0.04) and Group 4 (p = 0.03), while expression was significantly reduced in the WNT subgroup (p = 0.04; Fig 5B). In contrast, all subgroups showed significantly higher expression profiles compared with adult cerebella tissue (p<0.0001 all subsets; Fig 5A and 5B). These findings were further corroborated on a protein level with elevated CD24 protein expression levels in Group 3, Group 4 and SHH medulloblastoma compared to the WNT medulloblastoma subgroup (Fig 5C). Co-labelling with CD15 identified little CD24+/CD15+ co-expression in any of the four medulloblastoma subgroups (Fig 5D). CD24-/CD15+ expression was predominantly found on WNT tumours, while CD24+/CD15-expression was correlated with SHH, Group 3 or Group 4 designation (Fig 5D). Survival analysis identified a correlation in patient survival with CD15 expression, but not with CD24 (p = 0.008; Fig 5E). Together these results indicate that CD24 expression is elevated on non-WNT subgroups of human medulloblastoma and may serve as a selective marker in the treatment of Group 3, Group 4 and SHH medulloblastoma subgroups.

Discussion
The cancer stem cell hypothesis posits that a distinct and identifiable subpopulation of cells play a key role in tumour initiation, progression and tumour resistance. Identifying this population may be an important step in not only developing target therapies, which may lead to better outcomes in cancer treatment, but also identify patients with tumours that may be responsive to therapy. To this end, a significant amount of research has focused on the CD24 identifies Patched1 deleted medulloblastoma-initiating cells expression of cell surface markers as a means to identify the putative cancer stem cells. Here we report a cell surface CD antibody screen on a Ptch1 deleted model of medulloblastoma [33], in which we compared expression profiles with developing post-natal day 7 wild type granule cell precursors and adult wild type cerebella. Of the antibodies tested, CD15, CD24 and CD133 have previously been reported in human medulloblastoma [25,26,29,50,51]. CD15 and CD133 expression have been identified on TICs within murine Ptch1 deleted medulloblastoma [25], but until now CD24 expression had not been extensively investigated in any murine model of medulloblastoma.
As a putative tumour-propagating cell marker, elevated CD24 expression has been documented in lung, breast, ovarian and brain cancers [52][53][54][55][56], and has been affiliated with high Shh/Ptch1 pathway activity in both colorectal cancer and medulloblastoma formation [57,58]. CD24 expression has been identified in zones of secondary neurogenesis including the dentate gyrus, the subventricular zone (SVZ) and the rostral migratory stream, areas known to contain neural stem cells [59]. In our screen, CD24+ cells were highly positive for the stem/progenitor marker Sox2, whose expression has been shown to label stem-like cells in the murine brain [66], and has been shown to be required for SHH-associated medulloblastoma formation [67]. Despite this, CD24 is not known as a putative neural stem cell marker but rather shown to regulate the proliferation of neuronal precursor cells within the SVZ [60]. Histological and FACS analysis identified CD24+ cells to have a more glial progenitor-like identity, suggesting that CD24 expression may be correlated with the expansion of granule cell precursors, but not with the transition of stem cells to granule cell precursor fate. Both CD24+ and CD24-cells were capable of initiating tumorigenesis although CD24+ cells gave rise to tumours at a faster rate and with lower numbers of transplanted cells, implying that CD24 expression may be correlative with tumour propagation. Interestingly, the vast majority of cells in tumours resulting from the CD24-injections were CD24+ implying that the CD24-cell is capable of giving rise to CD24+ cells in vivo. Rietze et al. identified that CD24 low /peanut-agglutinin low cells isolated from the adult neurogenic niche were almost exclusively neural stem cells [20], suggesting a possible neural stem cell fate for the CD24-cells within Ptch1 deleted medulloblastoma. While Sox2 expression was not identified on our screen of CD24-cells, our findings suggest that the TIC of origin in the CD24-population is a potential rarer, stem-like cell. CD24 alone could not isolate putative stem cells, but may label cells transitioning from a Sox2 fate to a neural progenitor or radial glial cell fate, cell types frequently observed within medulloblastoma [25,68].
While CD24 does not appear to label stem-like cells in our model, CD133 (prominin-1) positive stem/progenitor-like cells have been identified in the murine cerebellum [28,38]. In our Ptch1 model of medulloblastoma we observed a CD133 expression profile similar to previous reports that have identified CD133+ cells in glioblastoma and medulloblastoma [23,38,50,[69][70][71]. While CD133+ cells have been reported to recapitulate neural tumour formation [29,50], results corroborated in this study, less is known about the ability of CD133 negative cells to induce tumours. Previous studies have reported that tumours derived from xenotransplanted CD133-glioma cells have been shown to contain large proportions of CD133+ cells, CD24 identifies Patched1 deleted medulloblastoma-initiating cells suggesting that while CD133 expression is not necessary for oncogenesis it may be important for tumour progression [70]. In this study no distinct expression profiles were observed with CD133 and CD24 co-staining, indicating that CD24 expression was not correlative with CD133 expression when identifying stem-like cells in Ptch1 deleted medulloblastoma. In contrast, we observed a differential expression profile when CD24 was co-labelled with the carbohydrate adhesion molecule CD15. CD15 has been identified as a possible marker of neural stem cells [21], and in combination with CD24 and CD29, has been shown to demarcate embryonic neural stem cells, neural crest cells and neurons [72]. Recently CD15 has been identified as a marker of medulloblastoma-initiating cells in Ptch1 +/mice [25,30] and its expression is correlated with a poor prognosis in human medulloblastoma patients [25,30]. Read et al. proposed that CD15 identified a medulloblastoma initiating cell irrespective of whether it was a true cerebellar stem cell or a more committed granule cell precursor [25]; results corroborated in this study. Nevertheless, despite its success as a marker of TICs, and in a possible corollary of the situation for CD133, CD15 is not a putative stem-cell marker. A subset of CD15 + cells do not form neurospheres in vitro [25], and its expression on TICs has been shown to diminish when used on serially passaged neurospheres [21,73]. In addition, it has been reported that in contrast to CD24, which labels a more neuron-like precursor, CD15 expression was associated with a more neural stem-like cell identity [72]. Based on these findings, we hypothesised that CD15 alone was not sufficient to purify the putative tumour-initiating granule cell precursor population within Ptch1 deleted medulloblastoma. Co-labelling with CD15 and CD24 allowed for the identification of a more mitotic and tumorigenic CD15+/CD24 + cell population within Ptch1 deleted medulloblastoma, illustrating that the CD24 TIC population could be further purified with CD15. Interestingly, rare tumours did arise from the CD15+/CD24-and CD15-/CD24-populations indicating that within the CD24-population resides a TIC that cannot be selected for by CD15.
The expression of cell surface receptors on human medulloblastoma and glioblastoma have played an important role in understanding the correlation between receptor activity and patient survival. In gliomas, studies have identified a correlation between CD antibody expression and tumour severity. Specifically, CD133 and CD15 expression have been correlated with reduced patient survival, associated with late stage glioma formation [25,26,29,30,50,74]. In medulloblastoma, where the severity of the tumours is more difficult to deduce based on genetic inception, less is known about the correlation of CD antibody expression and patient survival. CD24 and CD15 expression has been reported to be up-regulated on human medulloblastoma but little is known about their expression and correlation with patient survival on a subgroup level [25,51]. Here we report for the first time that CD24 is universally up-regulated across all medulloblastoma subgroups compared to normal brain tissue. CD24/CD15 coexpression identified a correlation with high CD15 expression and low CD24 expression on WNT subgroups of medulloblastoma, with high CD24 expression and low CD15 expression correlated with SHH, Group 3 and Group 4 classifications, suggesting CD24 as a marker of non-WNT medulloblastoma. In contrast to Read et al [25], who report a correlation of better patient survival with low CD15 gene expression, we show that the expression of CD15 was correlated with better patient survival, a result likely associated with the WNT subgroup profile which generally have a better survival rate than non-WNT medulloblastoma [75].
Collectively these results confirm the complexity of identifying TICs in medulloblastoma. Where correlations can be made from CD antibody expressions on Ptch1 deleted murine models, human subgroup analyses can show conflicting results. Ongoing work is required to elucidate the role and function of CD15, CD24 and CD133 in medulloblastoma with the hope of utilising these markers to successfully eliminate TICs. Based on the observations reported in this study we hypothesise that the medulloblastoma TIC is a rare stem/granule progenitor-like cell that cannot be identified by CD15, CD24 or CD133 alone, but together may enhance the ability to target these cells within different human medulloblastoma subgroups.