CD271⁺ Cells Are Diagnostic and Prognostic and Exhibit Elevated MAPK Activity in SHH Medulloblastoma

Central nervous system tumors are among the most prevalent forms of childhood cancers and account for the highest number of cancer-related deaths in both American and Canadian children under 20 (National Center for Health Statistics, 2016; Canadian Cancer Society Statistics, 2017). Medulloblastoma is the most common primary malignant pediatric brain cancer and is currently divided into at least five molecular subgroups that exhibit different genomic aberrations, gene expression profiles and clinical outcomes; Wingless (WNT), Sonic Hedgehog (SHH)-TP53 wild-type, Sonic Hedgehog (SHH)-TP53 mutant, Group 3 and Group 4 (1).

SHH medulloblastomas have an intermediate prognosis; however, they also exhibit significant intertumoral heterogeneity including multiple very high-risk groups that account for a majority of treatment failures (2–5). For example, Zhukova and colleagues (4) demonstrated that TP53 mutations confer poor prognosis in patients with SHH medulloblastoma. Cavalli and colleagues (2) extended this finding by delineating 4 SHH subtypes within this medulloblastoma subgroup, SHHα, SHHβ, SHHγ, SHHδ, and determining that TP53 mutations are prognostic only in SHHα. The other SHH subtypes include two infant groups with distinct clinical outcomes (SHHβ, SHHγ) as well as older patients exhibiting TERT promoter mutations and better prognosis (SHHδ; ref. 2). This extensive heterogeneity has revealed a critical need for subtype-specific functionally validated biomarkers and targeted therapeutic strategies. Treatment with SHH pathway inhibitors, particularly smoothened (SMO) antagonists, showed initial promise; however, acquired drug resistance has led to relapse in both preclinical SHH medulloblastoma animal models and clinical trials (6–8). This is attributed to the mutational status of specific genes within the SHH pathway (8, 9) as well as compensatory molecular mechanisms that circumvent SHH signaling dependence in mouse models of the disease (9).

In addition to the genetic and molecular heterogeneity in medulloblastoma, cellular heterogeneity is also a major contributing factor to therapeutic resistance and tumor recurrence in brain tumors (10). Putative brain cancer stem cells or tumor-propagating cells (TPC) exhibit stem cell–like properties including the capacity for self-renewal and multi-lineage differentiation. They are thought to be responsible for tumor initiation, recurrence, and drug resistance in several types of brain cancers, including medulloblastoma (10). The cell surface markers CD133 and CD15 have been reported to select for TPCs in medulloblastoma (10–13). However, inconsistencies in the cellular phenotypes associated with these markers have complicated efforts to fully elucidate the molecular mechanisms contributing to cellular heterogeneity in medulloblastoma tumors. To address this issue, we previously sought to identify novel cell surface biomarkers that were differentially expressed between self-renewing and non-self-renewing SHH medulloblastoma cells (14). Using a high-throughput flow cytometry screening platform and gain-/loss-of-function studies, we demonstrated that the CD271/p75 neurotrophin receptor (p75NTR) is associated with SHH medulloblastoma stem/progenitor cells in vitro. Importantly, we also showed elevated CD271 expression specifically in SHH medulloblastoma (14).

CD271 plays many roles in the development of the nervous system and is a selection marker for TPCs in multiple cancers including esophageal squamous cell carcinoma (15), hypopharyngeal cancer (16), melanoma (17, 18), as well as head and neck squamous cell carcinoma (19). This receptor has also been shown to play an important regulatory role in glioblastoma TPC proliferation (20) as well as invasion (21, 22). Here, we employed complementary bioinformatics analyses of large patient datasets and cultured tumorspheres along with extensive validation in functional assays to fully characterize the role of CD271 in SHH medulloblastoma. CD271, in combination with the transcription factor orthodenticle homeobox 2 (OTX2) as an exclusion marker, is highly selective for SHH medulloblastoma. Higher CD271 expression is also associated with better prognosis in patients with SHH medulloblastoma. CD271⁻ and CD271⁺ subpopulations are distinct at the molecular and cellular levels both in vitro and in vivo. MAPK/ERK signaling regulates endogenous CD271 levels, stem/progenitor proliferation, survival and migration in vitro as well as survival in vivo. Our results underscore the clinical implications of cellular heterogeneity in SHH medulloblastoma and suggest that MEK1/2 inhibitors, either alone or in combination with other pathway antagonists, are potential treatment strategies for SHH medulloblastoma tumors.

Daoy cells were purchased from the ATCC and cultured as described previously (14). D283 and D341 cells were purchased from ATCC. D283 (23) exhibits features of Group 3 (24) and Group 4 medulloblastoma (25), while D341 is a Group 3 medulloblastoma cell line (26). D283 cells were grown as adherent cultures in EMEM + 10% FBS and as tumorspheres in neural stem cell (NSC) media as described previously (27). D341 were cultured as tumorspheres in StemPro NSC Serum-Free Medium (Life Technologies). UI226 cultures were originally established by the Central Nervous System Tissue Bank, Department of Neurosurgery, University of Iowa (Iowa City, IA) and adapted to cell culture in StemPro medium as described previously (14, 28). NanoString analysis identified UI226 cells as SHH medulloblastoma. All cell lines have been authenticated by STR profiling (ATCC) but not Mycoplasma tested.

For inhibitor treatments, tumorsphere assays were performed for Daoy and UI226 cells as described previously (14). D283 and D341 cells were dissociated and plated at 10 cells/μL in 24-well ultra-low attachment plates in NSC Medium and StemPro medium, respectively. Daoy-negative control, Daoy CD271 overexpression (OE), and UI226 cells were treated with PD98059 (1, 5, 10, 20, 50 μmol/L), selumetinib (1, 5, 10, 20, 50 μmol/L), vismodegib (1, 2, 5, 10, 20 μmol/L), or DMSO vehicle control. D283 and D341 cells were treated with PD98059 (1, 5, 10, 20, 50 μmol/L) or DMSO control. Tumorspheres were treated once at day 0, incubated for 5 days, counted, measured, then dissociated, replated, and treated with inhibitors again at day zero for secondary tumorsphere assays.

UI226 cultures were plated in 96-well ultra-low attachment round-bottom plates at 4 × 10³ cells/well as described previously (29). After 4 days, cells formed spheroids that were then overlain with collagen basement membrane (BD Biosciences). Medium was removed and then replaced with 100 μL collagen mixture (collagen type 1, DMEM, and NaOH based on the manufacturer's guidelines). Once the collagen mixture gelled, 100 μL medium was added along with PD98059, selumetinib (1, 5, 10, 20, and 50 μmol/L) or DMSO. Total aggregate migration was calculated by subtracting day 0 from day 3 diameter.

CD271 was stably overexpressed as described previously (14). Briefly, Daoy cells were infected with the pReceiver-Lv105 lentiviral construct (GeneCopoeia) containing a puromycin resistance gene. Lentifect Lentiviral Particles were used as a negative control and puromycin was used for stable selection.

UI226 tumorspheres were stained with CD271 and 7-aminoactinomycin D (7AAD; Beckman Coulter) and then FACS sorted on CD271 expression using a MoFlo XDP cell sorter (Beckman Coulter). Analysis was performed using FlowJo software (Tree Star Inc.). CD271⁺ and CD271⁻ cells were collected and RNA isolated using the Norgen RNA extraction kit (Norgen Biotek) according to manufacturer's instructions. RNA sequencing was performed by StemCore laboratories at the Ottawa Hospital Research Institute. Libraries were prepared from 500 ng of input total RNA with the Truseq RNA v2 library prep kit (Illumina). DNA libraries were prepared with unique barcodes and pooled; 75 cycles of single-end sequencing were performed on a high output flowcell with the NextSeq 500 (Illumina). RNA-seq reads were aligned to the human genome using HISAT2 (v.2.0.3; ref. 30), using an index built from the GRCh38 reference assembly, and the GENCODE v23 transcript annotation set. Overall mapping rates of 85.7%–88.3% were obtained across all samples. Mapped reads were assigned to genes from the GENCODE v23 annotation set using featureCounts (31). Between 69.7% and 75.3% of reads were unambiguously assigned to a single gene in each sample. Differential expression between conditions was analyzed in R using the DESeq2 package (v1.12.4; ref. 32). Per-gene read counts were loaded and filtered to retain only genes with at least five reads assigned in two samples. DESeq2 was used to calculate library normalization factors and dispersion estimates. Expression differences between CD271⁺ and CD271⁻ samples were calculated using a model that included terms for both the state (CD271⁺ or CD271⁻) and the replicate; inclusion of the replicate term resulted in identification of 3,433 genes with significantly different expression between the CD271⁻ and CD271⁺ conditions at a FDR of 5% (Benjamini–Hochberg correction applied to P values). Differentially expressed pathways were analyzed using Ingenuity Pathway Analysis (IPA) and significance (P < 0.05) was determined by a right-tailed Fisher exact test. Predicted activation/inhibition states of canonical pathways were based on a Z-score algorithm. GSEA (33) was used to explore which pathways, networks, and functional annotation classes are over-represented in genes that are expressed at higher or lower levels in CD271⁺ versus CD271⁻ samples. A list of protein-coding gene symbols was generated by filtering for the “protein_coding” label in the GENCODE “gene_biotype” field, ranked by log₂ fold change from high to low values. The ranked list was imported into the GSEA software and used to calculate GSEA enrichment scores for MSigDb gene sets (v5.2). Enriched gene sets were examined manually for pathways of interest.

BrdU incorporation was performed using a BD Pharmingen BrdU Flow Kit (BD Biosciences) according to manufacturer's instructions. Daoy CD271 OE tumorspheres, UI226 tumorspheres, and UI226 tumorspheres treated for 3 days with DMSO, PD98059, or selumetinib (20, 50 μmol/L), were pulsed with 1 mmol/L BrdU. After 5 hours of incubation, tumorspheres were dissociated and stained with CD271 antibody for 30 minutes in the dark. Cells were fixed, washed and stained according to manufacturer's guidelines. 7AAD was added and samples were analyzed using the MoFlo XDP cell sorter and FlowJo software.

Annexin V staining was performed using a phycoerythrin (PE) Annexin V Apoptosis Detection Kit (BD Biosciences) according to manufacturer's guidelines. UI226 tumorspheres were treated with 20, 50 μmol/L of PD98059, or selumetinib for 3 days, dissociated and then stained with Annexin V and 7AAD. Samples were analyzed using the MoFlo XDP cell sorter and FlowJo software.

Formalin-fixed, paraffin-embedded (FFPE) tissue from surgically resected medulloblastomas was obtained from the Clark Smith Brain Tumour Bank at the University of Calgary, the Hospital for Sick Children, and the University of Manitoba. Medulloblastoma subgroup information was determined previously by NanoString profiling (34). FFPE tissues were deparaffinized, antigen retrieval performed at 95–100°C for 20 minutes in citrate buffer pH 6.0, and cooled for 30 minutes. Slides were washed in 1× PBS, treated for endogenous peroxidase for 10 minutes, and washed in 1× PBS. Samples were blocked with (CD271): 3% lamb serum, (OTX2): 10% goat serum, (Ki67): 10% sheep serum in 1× PBS, and treated with primary antibody in (CD271):1% lamb serum in 1× PBS, (OTX2): 1% BSA in PBST (0.2% Triton X-100), or (Ki67): 1% sheep serum in 1XPBS overnight at 4°C: CD271 (1:400; #05-446; Millipore), OTX2 (1:500; #ab21990; Abcam), Ki67 (1:800; Cell Signaling Technology). Slides were treated with secondary antibody (1:500), sheep anti-mouse IgG (H+L) (Jackson ImmunoResearch) for 2 hours at room temperature. Slides were treated with 1:400 dilutions of streptavidin/HRP (Jackson ImmunoResearch) in 1× PBS for 30 minutes and developed using DAB. Slides were counterstained with hematoxylin. Coverslips were mounted with Permount (Thermo Fisher Scientific).

Protein was isolated from UI226 and Daoy cells dissociated from primary tumorspheres using 975 μL Lysis Buffer, 20 μL 50× protease inhibitor, and 5 μL orthovanadate. Forty micrograms of protein from sorted UI226 and Daoy CD271⁺ and CD271⁻ cells were separated by SDS-PAGE using 10% acrylamide gels. Protein was transferred using a semi-dry transfer method to nitrocellulose membrane (Bio-Rad) and washed as described previously (14). Membranes were blocked in 2.5% nonfat milk in TBST for 30 minutes, and then incubated for 1 hour at room temperature in primary antibody diluent with antibodies to GAPDH (#0411; Santa Cruz Biotechnology, 1:2,000), CD271 (#07-476; Millipore, 1:500), and pERK1/2/Total ERK1/2 (#4370S and #4695S; Cell Signaling Technology, 1:500). Membranes were washed 3× with TBST for 5 minutes before application of donkey anti-rabbit horseradish peroxidase secondary antibody (Bio-Rad; 1:5,000) for 1 hour at room temperature. Membranes were washed 3× with TBST for 5 minutes and then developed using SuperSignal West Pico.

Total RNA was extracted from sorted UI226 CD271⁻ and CD271⁺ cells using the Norgen RNA extraction kit (Norgen Biotek). First-strand cDNA was synthesized using the Superscript III First Strand Synthesis System (Life Technologies). The following qPCR conditions were used: 50°C for 2 minutes, 95°C for 2 minutes, and 40 cycles of 95°C for 15 seconds and 60°C for 30 seconds. qPCR was conducted using GoTaq qPCR Master Mix (Thermo Fisher Scientific) and analysis was performed on a Mx3000P Stratagene qPCR system. All values were normalized to GAPDH. Specific primer sequences for each gene evaluated are listed in Supplementary Table S1.

The University of Manitoba Animal Care Committee approved all procedures. Dissociated tumorspheres from Daoy CD271-negative control, Daoy CD271 OE, and UI226 were injected into the cerebellum of 5- to 7-week-old NOD SCID mice. Animals were injected with either 5 × 10⁴ cells (n = 6 Daoy control, n = 6 Daoy CD271 OE) or 2.5 × 10⁵ UI226 cells. For Daoy cells, animals were perfused with formalin after 12 weeks, and the brains extracted, placed in formalin for 2–7 days, and then prepared for IHC analysis as described previously (14). For UI226 cells, after 2 weeks of tumor growth, animals were randomly separated to either receive selumetinib (n = 8) or 0.5% hydroxypropyl methyl cellulose, 0.1% polysorbate 80 as the vehicle control (n = 10). Selumetinib was administered at 75 mg/kg, twice daily (once daily on weekends) via oral gavage. For vismodegib treatment, animals received either vehicle control (+6%DMSO) or 50 mg/kg vismodegib by oral gavage. After animals reached endpoint (20% weight loss, lethargy, and ruffled fur), they were perfused and/or brains extracted for histologic analysis as described previously (14).

For Daoy in vivo studies, mice were imaged on a 7T 21 cm Bruker Avance III NMR system with Paravision 5.0 (Bruker BioSpin) as described previously (14). For selumetinib studies in vivo, mice were anaesthetized with 4% isoflurane and maintained with a mask at 1.5%–2% isoflurane in oxygen. Mice were imaged using a MR Solutions cryogen free FlexiScan 7T MRI with a 17-cm bore. T1 images were acquired using a mouse head quadrature coil (MR Solutions) with the following parameters: FOV 20 × 20 mm, matrix size 512 × 252, slice thickness 0.5 mm. TR 1,000 ms and TE 12 ms. T2-weighted images were acquired with the following parameters: FOV 20 × 20 mm, matrix 256 × 245, slice thickness 0.5 mm, TR 4,000 ms, and TR 45 ms. Four averages were acquired in both cases.

CD271 and OTX2 expression were analyzed across 763 primary medulloblastoma samples, profiled on the Affymetrix Gene 1.1 ST array as described previously, normalized using the RMA method, and subgrouped using similarity network fusion (GSE85217; ref. 2). Differences across subgroups and subtypes were evaluated using ANOVA in the R statistical environment (v3.4.2). Using the R2 software (http://r2.amc.nl), we analyzed CD271 gene expression in primary SHH medulloblastomas and found genes highly correlated with CD271 expression with an r > 0.2 and corrected P value < 0.01. The specific pathways associated with the expression of CD271 in the R2 database were identified using the KEGG pathway finder option. These same genes were ranked by r value and a ranked GSEA analysis was performed using the oncogenic signatures (33, 35). The pathways that show significant (corrected P ≤ 0.01, χ² test) enrichment based on the genes coexpressed with CD271 were identified. Survival was measured from the time of initial diagnosis to the date of death or of last follow up. Survival distribution was estimated according to the Kaplan–Meier method using optimal cut-off selection and log-rank statistics in both the Cavalli and colleagues and Cho and colleagues datasets (2, 36). P values < 0.01 were considered to be statistically significant.

Data from in vitro and in vivo experiments were analyzed using Prism 5 software (GraphPad Software). Tumor size from our intracerebellar xenograft model was analyzed using an independent sample one-tailed t test with Welch correction. Tumor survival was assessed using a log-rank (Mantel–Cox) test. MEK inhibitor experiments were assessed using a one-way ANOVA followed by a Dunnett test for multiple comparisons. All analyses were checked for homogeneity of variances using a Brown–Forsythe test. All data were reported as a mean ± SEM. P values < 0.05 were considered significant.

To explore the possibility that CD271 serves as a novel SHH medulloblastoma diagnostic marker, we evaluated CD271 levels in 63 FFPE subgrouped medulloblastoma patient samples (Supplementary Table S2) using IHC analysis (Fig. 1A). The majority of SHH samples were CD271⁺ and exhibited a nodular CD271 staining pattern underscoring the cellular heterogeneity within these tumors. While all WNT and Group 4 tumors were CD271 negative, a portion of Group 3 tumors exhibited positive, albeit much less frequent, CD271 staining (Supplementary Table S2).

To address this, we also examined levels of the homeodomain transcription factor orthodenticle homeobox 2 (OTX2) in the medulloblastoma tumor samples by IHC (Fig. 1B), as OTX2 is expressed in all the subgroups except SHH (37). Nearly all WNT, Group 3, and Group 4 tumors expressed OTX2; while only 10% of SHH tumors exhibited detectable levels. In support of these findings, transcript levels of CD271 and OTX2 were analyzed across 763 subgrouped medulloblastoma patient samples. A striking inverse correlation of high CD271 with correspondingly lower OTX2 expression was observed across SHH tumors (Fig. 1C). Further breakdown of gene expression within the 4 SHH subtypes (2) revealed a CD271^high/OTX2^low pattern in SHH samples (Fig. 1D), particularly in the SHHδ subtype consisting of older patients with TERT promoter mutations and a more favorable prognosis (Fig. 1D). The inverse correlation between CD271 and OTX2 was highly significant not only within SHH (Fig. 1E) but also across all medulloblastoma tumours (Fig. 1F). These results prompted us to further investigate the prognostic relevance of CD271 and OTX2 across the medulloblastoma subgroups. Interestingly, low OTX2 and low CD271 were both significantly associated with poor outcome within SHH tumors (Fig. 1G–I); however, only CD271 was validated in a nonoverlapping dataset (Fig. 1H). When analysis was restricted to the high-risk SHHα subtype, both CD271 and OTX2 remain significant, with OTX2 showing quite a dramatic difference (Supplementary Fig. S1A and S1B). Collectively, these results demonstrate the potential utility of the CD271 cell surface receptor as both a combination diagnostic and prognostic biomarker in SHH medulloblastoma.

Our previous work revealed that CD271 is associated with the stem/progenitor cell state in SHH medulloblastoma (14). These results, combined with our current data demonstrating the clinical relevance of this receptor, prompted us to dissect the molecular profiles of CD271⁻ and CD271⁺ cells in SHH tumors. To this end, we evaluated the stability of these subpopulations by FACS sorting CD271⁻ and CD271⁺ cells from UI226 (low passage primary SHH medulloblastoma) tumorspheres (14, 28), followed by reestablishment of sorted cells as secondary tumorspheres. CD271⁻ and CD271⁺ subpopulations are unstable and undergo redistribution of their CD271 expression profile after 5 days (Fig. 2A and B). Expression of neural lineage markers by qPCR revealed no significant differences between CD271⁺ and CD271⁻ cells (Fig. 2C), further underscoring the notion that stemness is a dynamic rather than a fixed trait in many cancers (38). RNA sequencing was then performed on CD271⁻ and CD271⁺ cells immediately after sorting to define the molecular mechanisms contributing to these subpopulations (Fig. 2D–G). In total, 3,433 genes were significantly and differentially expressed between CD271⁻ and CD271⁺ cells. IPA analysis revealed that pathways associated with cell proliferation and survival including p53 signaling, MAPK/ERK, PI3K/AKT, and mitotic roles of polo-like kinase were among the most differentially expressed (Fig. 2E). GSEA supported these findings and demonstrated that genes/pathways associated with negative regulation of cell death, proliferation, and motility were significantly enriched in gene sets that were downregulated in CD271⁻ versus CD271⁺ subpopulations (Fig. 2F; Supplementary Tables S3–S5). In contrast, genes associated with the MYC pathway, spherical vs. adherent phenotypes and a glioblastoma proneural signature were enriched in gene sets that were upregulated in CD271⁻ versus CD271⁺ cells (Fig. 2G; Supplementary Tables S6–S8). We chose a subset of genes from these pathways that were most significantly up- or downregulated in CD271⁻ versus CD271⁺ to validate by qPCR. These results confirmed our RNA sequencing data and revealed downregulation of genes involved with cell proliferation/survival (CDK1, PLK1, IL32) and motility (TGFBI) in CD271⁻ versus CD271⁺ cells, while GRIK2, PLA2G4, MYC, DCX, OLIG1 were upregulated in CD271⁻ versus CD271⁺ cells (Supplementary Fig. S2A and S2B). Collectively, these results indicate that CD271⁻ and CD271⁺ cells are highly dynamic, yet molecularly distinct subpopulations within the heterogeneous tumorsphere microenvironment.

As signaling pathways associated with cell proliferation and survival such as Ras/MAPK were significantly upregulated in CD271⁺ cells (Fig. 2E), we wanted to further investigate these relationships in primary samples. To determine whether Ras/MAPK signaling is elevated in primary samples harboring high CD271 expression, we performed an analysis of genes highly correlated with high CD271 expression within SHH (r > 0.2, corrected P < 0.01) across 223 primary SHH medulloblastoma. Strikingly, a KEGG pathway analysis identified genes enriched in the Ras pathway as highly overrepresented in samples harboring high CD271 (P = 1.7 × 10⁻³). We confirmed this association using a ranked GSEA analysis, of genes highly correlated with CD271 (r > 0.2, corrected P < 0.01), and found several Ras-associated pathways and mTOR-related pathways significantly upregulated, confirming our sorted UI226 CD271⁺ and CD271⁻ subpopulations are representative of primary SHH tumors and are functionally distinct (Supplementary Tables S9–S11). Indeed, further interrogation of our RNA sequencing data revealed that genes associated with a RAS/MEK signature and the G₂–M checkpoint were significantly enriched in gene sets that were upregulated in CD271⁺ cells, consistent with CD271^high primary SHH tumors (Fig. 2H; Supplementary Tables S12 and S13).

To validate our transcriptome data at the functional level, BrdU incorporation was performed on sorted CD271⁺ and CD271⁻ UI226 cells from tumorspheres, and on our previously established stable CD271-overexpressing (OE) Daoy medulloblastoma cell line (14) as very few human SHH medulloblastoma cell models exist in culture for functional analyses. Both cell lines were used as CD271-expressing model systems. Consistent with our RNA sequencing findings, BrdU incorporation revealed a significant increase in the proportion of S-phase and G₂–M cells and a concomitant decrease in G₀–G₁ in the CD271⁺ subpopulation compared with CD271⁻ UI226 cells (Fig. 3A). Similar results were obtained for CD271^low/high cells FACS-sorted cells from Daoy CD271 OE tumorspheres, which also showed an increase in the frequency of S-phase and G₂–M phase cells (Fig. 3B).

We also performed intracerebellar transplants comparing growth and motility of Daoy CD271 OE relative to negative control tumorsphere cells in NOD SCID mice. Interestingly, Daoy CD271 OE cells generated significantly smaller, more localized tumors at the injection site (Fig. 3C and D); however, they also exhibited extensive subarachnoid space tumor cell spread, with additional nodules in the ventricles (including the lateral ventricle frontal horns) and acute hydrocephalus (shredding of white matter around ventricles; Fig. 3E and F). To evaluate proliferation between the two groups, we performed Ki67 staining on FFPE sections from Daoy-negative control and Daoy CD271 OE tumors (Fig. 3G and H). Interestingly, no differences in Ki67 were observed, as the smaller Daoy CD271 OE tumors at the injection site as well as the additional nodules/SAS spread were proliferating quite extensively. These results suggest that the in vivo microenvironment or niche elicits mainly a motility effect on Daoy cells stably overexpressing CD271. Taken together, these results support our RNA sequencing data and reveal functional differences in proliferation and cell motility between CD271⁻ and CD271⁺ cells in vitro and in vivo.

Our data revealed that genes associated with Ras/MAPK signaling were enriched in CD271⁺ cells from UI226 tumorspheres as well as primary SHH medulloblastoma tumors with high CD271 levels. As Ras/MAPK signaling has also been shown to be a druggable target in other brain tumors (39, 40), we chose to pursue this pathway in more detail. MAPK pathway activity was increased in sorted CD271⁺ relative to CD271⁻ cells from UI226 tumorspheres as demonstrated by elevated pERK1/2 (Fig. 4A). We also observed decreased pERK1/2 levels in our previously generated (14) stable CD271 knockdown UI226 tumorspheres relative to scramble controls that exhibit very high endogenous CD271 expression (Supplementary Fig. S2C). Importantly, we also observed pERK staining in a subset (30%) of SHH patient samples (Supplementary Fig. S3). Taken together, the analysis of gene expression and protein levels from primary SHH tumors, along with our gene expression analysis of SHH tumorspheres sorted for CD271 suggest that cells with activated MAPK signaling represent a clinically relevant subpopulation within these tumors.

To determine the effect of MAPK pathway activation on SHH medulloblastoma cell function, we utilized the well-characterized MEK1/2 inhibitor PD98059, and tested various concentrations (1, 5, 10, 20, 50 μmol/L) in tumorsphere assays (Fig. 4B). At 20 μmol/L and 50 μmol/L PD98059, UI226 cells exhibited a decrease in CD271 expression levels (Supplementary Fig. S4A) and a significant decrease in primary and secondary tumorsphere size, total cell number and viability (Fig. 4C–G). In support of these findings, we also observed a decrease in the percentage of S-phase cells by BrdU incorporation (Supplementary Fig. S4B) as well as an increase in Annexin V⁺/7AAD⁻ (dying) and Annexin V⁺/7AAD⁺ (dead) cells by flow cytometry (Supplementary Fig. S4C and S4D). Importantly, 20 μmol/L and 50 μmol/L PD98059 treatment had no significant effect on D283 and D341 Group 3 medulloblastoma tumorsphere number, cell count, or viability (Supplementary Fig. S4E–S4J). As Group 3 medulloblastoma tumorspheres are CD271⁻ (14), these results suggest that the effects of MEK1/2 inhibition may be specific to SHH medulloblastoma.

To strengthen our findings with PD98059, we also tested selumetinib (AZD6244), a potent, highly selective MEK1/12 inhibitor, in tumorsphere assays. Selumetinib has shown promise as a potential therapy for treatment of triple-negative breast cancer brain metastases (41) and glioblastoma (39) in preclinical animal models. Importantly, selumetinib crosses the blood brain barrier and is currently in clinical trials for treatment of pediatric refractory low-grade glioma (42). We utilized a range of concentrations shown to be effective in previous studies (1, 5, 10, 20, 50 μmol/L; ref. 43). As 3D cultures typically require higher drug concentrations than monolayer cultures, we chose to use a wider μmol/L range for testing in our SHH medulloblastoma tumorsphere assays. At all concentrations tested, we observed a decrease in pERK 24 hours after treatment (Fig. 5A). However, by day 5, pERK levels recovered following treatment with the lower concentrations consistent with previously published studies (44). Selumetinib treatment resulted in a 75%–80% reduction in CD271 levels (Supplementary Fig. S5A) as well as a dose-dependent decrease in primary and secondary tumorsphere size, total cell number, and viability (Fig. 5A–F). This was also supported by BrdU incorporation and Annexin V staining that showed a decrease in the frequency of cells in S phase (Supplementary Fig. S5B) and a corresponding increase in AnnexinV⁺/7AAD⁻ and Annexin V⁺/7AAD⁺ cells (Supplementary Fig. S5C and S5D).

As selumetinib is a more potent MEK1/2 inhibitor than PD98059 with significant changes observed even at the lowest doses, 1 and 5 μmol/L were also tested on our Daoy CD271 OE cells (Fig. 5G–K). Daoy control tumorspheres typically exhibit lower endogenous levels of CD271 (37). As such, we utilized our Daoy model to evaluate the effect of selumetinib in medulloblastoma tumors that exhibit lower (Daoy-negative control) and higher (Daoy CD271 OE) CD271 expression. While only 5 μmol/L selumetinib significantly reduced secondary tumorsphere size in Daoy-negative control cells, both 1 and 5 μmol/L significantly inhibited secondary tumorsphere size in the Daoy CD271 OE cells (Fig. 5H and I). Daoy CD271 OE cells also exhibited a decrease in cell number with 1 μmol/L and 5 μmol/L selumetinib as well as a decrease in viability with 5 μmol/L selumetinib (Fig. 5J and K).

In addition to cell proliferation and survival, our RNA sequencing data revealed significant expression differences in genes associated with cell motility between CD271⁺ and CD271⁻ cells, which was functionally validated in our xenograft model (Fig. 2F, 3C–F). Thus, we predicted that MEK inhibition would also affect cell migration. To this end, we generated aggregates of UI226 cells embedded in collagen and assessed migration after 3 days. While PD98059 did not significantly affect cell motility (Fig. 6A and B), we observed a strong, significant and dose-dependent decrease in cell migration following treatment with selumetinib (Fig. 6C and D). Similar results were obtained for Daoy CD271 OE aggregates with selumetinib significantly inhibiting cell migration in a dose-dependent manner (Fig. 6E and F).

Collectively, our functional data demonstrate the importance of MAPK signaling to SHH medulloblastoma proliferation, survival, and migration. The positive correlation between CD271 expression and RAS signaling in patient transcriptome datasets along with pERK staining in SHH medulloblastoma samples further underscore the clinical relevance of this pathway and provide a strong rationale for testing the effect of MEK/1/2 inhibitors in vivo.

As selumetinib significantly decreases CD271 levels, cell proliferation, survival and migration in vitro and is also known to be brain penetrant (39, 41), we tested the effect of this MEK1/2 inhibitor in our mouse xenograft model. Intracerebellar injections of 2.5 × 10⁵ UI226 tumorspheres cells were performed in NOD/SCID mice. Following 14 days of tumor growth, the animals were randomly divided into two groups with one receiving selumetinib (N = 10 at 75 mg/kg) and the other receiving the vehicle control (N = 8). Animals were treated twice daily (once daily on weekends) until endpoint was reached. While selumetinib-treated mice displayed relatively smaller tumors in the vermis at the same time point (Fig. 7A), control and treated tumors were phenotypically similar at survival endpoints, with large tumors in the vermis (Fig. 7A) accompanied by frequent extension into the fourth ventricle. Importantly, we observed a significant survival increase in the selumetinib-treated mice (Fig. 7B) with a median survival time of 55.5 days relative to 44 days for vehicle controls. Moreover, CD271 levels were downregulated in all selumetinib-treated tumors (Fig. 7C). In contrast, vismodegib, a well-known SMO inhibitor currently in clinical trials for the treatment of recurrent SHH medulloblastoma (8), had no significant effect on UI226 tumorspheres in vitro (Supplementary Fig. S6A–S6E) or survival in vivo (Supplementary Fig. S6F). These results support our in vitro studies and demonstrate the potential clinical utility of selumetinib for targeting CD271⁺ cells in a biologically relevant in vivo medulloblastoma tumor model.

In this study, we have established that CD271 is highly expressed across SHH medulloblastoma, suggesting both diagnostic and prognostic relevance, and within SHH medulloblastoma, CD271 identifies a functionally distinct subset of cells. Furthermore, we show that, CD271^high cells are characterized by elevated Ras/MAPK signaling and this population of cells can be targeted using MEK inhibition. Current personalized therapies for SHH medulloblastoma are lacking, as SMO inhibitors are not predicted to work in patients with very high-risk tumors and result in premature osseous fusion in those patients under age 10 limiting their utility. As such, targeting CD271⁺ cells through MEK inhibition represents a novel and rational therapeutic strategy for the treatment of SHH medulloblastoma.

CD271 may represent a very useful clinical marker, and our data suggest that CD271 immunopositivity and OTX2 immunonegativity can be used to robustly identify SHH tumors. Genome-wide DNA methylation or transcriptome profiling (5, 45) is the current gold standard for diagnosis of the medulloblastoma subgroups. However, Cavalli and colleagues, (2) recently demonstrated the existence of multiple subtypes within each subgroup by integrating data from both methods. While this work has revolutionized our understanding of the molecular heterogeneity in medulloblastoma, these methods are expensive and the vast majority of clinical settings still rely on IHC as a diagnostic method. As such, it is important to identify reliable and reproducible biomarkers that can be used on FFPE samples. GRB2-associated binding protein 1 (GAB1), an adaptor protein involved in multiple cell processes, has recently been shown to be an indicator of SHH medulloblastoma (46–48). However, the functional relevance of GAB1 in SHH medulloblastoma is not known (48). Here, we show that CD271 has diagnostic and prognostic utility, but more importantly, we have fully characterized the functional role and molecular profile of CD271⁺ cells. Indeed, CD271⁺ and CD271⁻ cells are distinct and coexisting subpopulations within SHH medulloblastoma tumors. Our working model is depicted in Fig. 7D. The CD271⁺ subpopulation is associated with an increase in cell proliferation and migration. Importantly, we identified novel pathways that are differentially expressed between these 2 subpopulations, underscoring the notion that CD271⁺ cells are functionally active druggable targets. Our pathway analyses in SHH patient samples revealed an enrichment of genes associated with MAPK signaling in CD271^high SHH medulloblastoma tumors and supports this group of tumors being a distinct and clinically relevant subset of SHH medulloblastoma.

To our knowledge, selumetinib has not been previously tested in human SHH medulloblastoma models. Selumetinib has been shown to decrease tumor growth in preclinical models of pediatric LGG with constitutively active BRAF (43, 49). Importantly, selumetinib showed significant activity in a phase I study of LGG and is currently in Phase II clinical trials for treatment of refractory LGG (42). This supports the notion that selumetinib is a brain penetrant and feasible treatment option in human clinical trials of SHH medulloblastoma. We observed strong and significant decreases in proliferation, survival and migration following selumetinib treatment in vitro. Combined with the in vivo data demonstrating a significant increase in survival following treatment with selumetinib alone, these results provide a rationale for pursuing selumetinib as a novel therapeutic strategy in SHH medulloblastoma. Interestingly, the MAPK signaling pathway has also recently been shown to drive SHH pathway inhibitor resistance (9). For example, Zhao and colleagues, demonstrated that MAPK/ERK pathway activation is increased in metastatic SHH medulloblastoma and this activation circumvents SHH pathway dependency specifically in mouse models of the disease (9). While we show that selumetinib treatment significantly extends survival in our intracerebellar transplant model, the mice still ultimately succumb to disease progression. Thus, future studies will identify candidates that act synergistically with selumetinib to further attenuate tumor growth in vivo.

In our study, CD271-overexpressing cells exhibit increased migration in vivo. This is an interesting finding and is supported by both the RNA sequencing data as well as the decrease in UI226 cell migration following selumetinib treatment in our 3D collagen assays. CD271 is associated with increased cell migration, invasion, and/or metastasis in several cancers (21, 22, 50, 51). Recent studies have even shown that CD271 regulates “phenotypic switching” in melanoma by decreasing proliferation while simultaneously increasing invasion through two completely independent mechanisms (52). These intriguing findings are not surprising given the complex nature of CD271 signaling in which effects on proliferation, motility, and survival are dictated by the (pro)neurotrophin ligand and coreceptor bound by CD271 as well as the availability of intracellular adaptor molecules (53). This context dependency may explain the association of CD271 with enhanced proliferation in tumorspheres in vitro and increased infiltrative spread in vivo. Thus, our strategy targeting SHH medulloblastoma cells harbouring the “CD271 signature” rather than CD271 signaling directly may yield the most consistent results. Indeed, a direct association between CD271 and MAPK signaling has been established. For example, Ceni and colleagues, (54) have shown that TRK-dependent activation of MEK/ERK signaling regulates cleavage and subsequent activation of CD271/p75NTR. Conversely, Perrone and colleagues, (55) and Murillo-Sauca and colleagues, (19) have both shown that CD271 stimulation leads to an increase in MAPK activity underscoring the complex, reciprocal regulation of these signaling pathways.

Our results demonstrate that reduced CD271 expression correlates with poor outcome across multiple independent SHH medulloblastoma datasets; however, these data were obtained exclusively from primary tumors. Evaluation of CD271 in the metastatic compartment or in matched primary recurrent tissue is limited due to a lack of available tissue and gene expression datasets obtained from recurrences and the metastatic compartment. Our findings reveal that CD271 is linked with proliferation, survival, and migration and that targeting CD271⁺ cells with MEK inhibitors affects these properties but not self-renewal. Thus, it appears that CD271 may be linked with a highly proliferative progenitor cell and not the more primitive stem cell phenotype, as we did not observe a difference in “stemness genes” (Fig. 2) in CD271⁺ relative to CD271⁻ cells. As previous studies have shown a strong correlation between expression of stem cell genes and poor prognosis, highly cycling, potentially more differentiated CD271⁺ cells may simply be more responsive to therapy and sensitive to cell death. This correlation with poor outcome is independent of subtype, suggesting that even in the aggressive SHHα harboring TP53 mutations, CD271⁺ cells may be more sensitive to treatment and that targeting these cells with MEK inhibition may allow a deescalation of current cytotoxic therapies.

We have utilized complementary bioinformatics data from both subtyped patient samples and low-passage primary cultures to identify novel diagnostic and prognostic roles for CD271 in SHH medulloblastoma. In addition, we have fully characterized the molecular profiles and the functional relevance of CD271⁺ and CD271⁻ SHH medulloblastoma cells in the tumorsphere environment. Importantly, the CD271⁺ subpopulation is functionally active and is successfully downregulated using the MEK inhibitor selumetinib both in vitro and in vivo thus opening new avenues for therapeutic targeting of SHH medulloblastoma cells exhibiting this novel cell surface biomarker.

No potential conflicts of interest were disclosed.

Conception and design: L. Liang, M.R. Del Bigio, V. Ramaswamy, T.E. Werbowetski-Ogilvie

Development of methodology: N. Tatari

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): L. Liang, L. Coudière-Morrison, N. Tatari, M. Stromecki, A. Fresnoza, M.R. Del Bigio, C. Hawkins, J.A. Chan, T.C. Ryken, M.D. Taylor, V. Ramaswamy

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): L. Liang, N. Tatari, C.J. Porter, M.D. Taylor, V. Ramaswamy, T.E. Werbowetski-Ogilvie

Writing, review, and/or revision of the manuscript: L. Liang, M.R. Del Bigio, C. Hawkins, T.C. Ryken, V. Ramaswamy, T.E Werbowetski-Ogilvie

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): L. Coudière-Morrison, N. Tatari

Study supervision: T.E. Werbowetski-Ogilvie

We thank Dr. Monroe Chan at the University of Manitoba Flow Cytometry Facility, Dr. Mike Jackson at the Small Animal Imaging Core, University of Manitoba, Dr. Richard Buist at The Magnetic Resonance Microscopy Center, University of Manitoba as well as Rhonda Kelley and Shawn Blum from Vet Services at the University of Manitoba for technical support. This work was funded by operating funds from the Canada Research Chairs Tier II Program and an Operating Grant from The Canadian Institutes of Health Research (to T.E. Werbowetski-Ogilvie). V. Ramaswamy is supported by the American Brain Tumor Association, the Collaborative Ependymoma Research Network, Garron Family Cancer Centre, Brain Tumor Foundation of Canada, b.r.a.i.n.child, and Meagan's Walk.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.