Intratumoral heterogeneity of MYC drives medulloblastoma metastasis and angiogenesis

Abstract

Background

Intratumoral heterogeneity is crucially involved in metastasis, resistance to therapy, and cancer relapse. Amplifications of the proto-oncogene MYC display notable heterogeneity at the single-cell level and are associated with a particularly dismal prognosis in high-risk medulloblastomas (MBs). The aim of this study was to establish the relevance of interclonal cross-talk between MYC-driven and non-MYC-driven MB cells.

Methods

We used fluorescence in situ hybridization, single-cell transcriptomics, and immunohistochemistry, in vitro isogenic cell models, non-targeted proteomics, mass spectrometry-based metabolite quantification, HUVECs tube formation assay, and orthotopic in vivo experiments to investigate interclonal cross-talk in MB.

Results

We found that the release of lactate dehydrogenase A (LDHA) from MYC-driven cells facilitates metastatic seeding and outgrowth, while secretion of dickkopf WNT signaling pathway inhibitor 3 from non-MYC-driven cells promotes tumor angiogenesis. This tumor-supporting interaction between both subclones was abrogated by targeting the secretome through pharmacological and genetic inhibition of LDHA, which significantly suppressed tumor cell migration.

Conclusion

Our study reveals the functional relevance of clonal diversity and highlights the therapeutic potential of targeting the secretome to interrupt interclonal communication and progression in high-risk MB.

Graphical Abstract

Key Points.

• Secreted lactate dehydrogenase A (LDHA) from MYC-driven medulloblastoma (MB) cells facilitates metastatic dissemination.

• Release of dickkopf WNT signaling pathway inhibitor 3 from non-MYC-driven MB cells enhances angiogenesis.

• Pharmacological and genetic targeting of LDHA suppresses MB metastasis formation.

Importance of the Study.

Remarkable intratumoral heterogeneity (ITH) has been recognized in medulloblastoma (MB). However, the functional consequences of ITH including subclonal amplification of the MYC oncogene in high-risk MB have not been comprehensively elucidated. We show that MYC-driven subclones promote invasiveness through secretion of lactate dehydrogenase A (LDHA), while non-MYC-driven subclones stimulate angiogenesis through release of dickkopf WNT signaling pathway inhibitor 3. Tumor migration was suppressed via pharmacological and genetic targeting of LDHA. The recognition of the importance of this secretome-mediated network between MYC-driven and non-MYC-driven subclones improves our understanding of aggressive Group 3 MB and the design of therapeutic interventions.

Intratumoral heterogeneity (ITH) has long been recognized as a potential factor contributing to metastasis dissemination and therapy resistance. Next-generation sequencing analyses have permitted higher resolution and more rapid analysis of individual cancer genomes at the single-cell level, which fosters the understanding of the biological and clinical impact of ITH.^1–4 While the origins and drivers of ITH have become a major focus of interest, the roles of intratumoral interactions between different subclones and their clinical implications remain poorly understood.

Medulloblastoma (MB) comprises the most common malignant pediatric brain tumor.^5,6 Distinct biological subgroups, namely wingless-related integration site (WNT), Sonic Hedge Hog (SHH), Group 3 (G3), and Group 4 (G4), indicate notable intertumoral heterogeneity.^7,8 Recent single-cell sequencing studies revealed profound clonal ITH^1,4,9 in addition to previously described spatial ITH.¹⁰ Subclonal amplification and overexpression of the proto-oncogene and transcription factor MYC are hallmark characteristics of high-risk G3-MB^7,8,11 with particularly dismal prognosis.¹² The amplification of MYC is typically restricted to a subpopulation (0.5% to 79% of the tumor cells) within these tumors.^13,14 During normal development, MYC is heterogeneously expressed in epiblasts, and epiblast cells with high MYC expression eliminate those with lower MYC expression.¹⁵ In contrast, in G3-MB, non-MYC-driven subclones are not outcompeted by aggressive MYC-driven subclones; this suggests that MYC-driven and non-MYC-driven cells are required and that both subclones might interact to cooperatively drive aggressive growth.

To better understand the mechanisms mediating cellular heterogeneity and subclone communication in MYC-driven MB, we used an in vivo orthotopic injection model and in vitro co-culture experiments to show that ITH promotes aggressiveness and that the functional interaction between MYC-driven and non-MYC-driven cells is mediated by the secretome. We identified and verified lactate dehydrogenase A (LDHA) is a protein secreted from MYC-driven cells, and dickkopf WNT signaling pathway inhibitor 3 (DKK3) is a protein secreted from non-MYC-driven cells. Furthermore, we showed that secreted LDHA increases the conversion of pyruvate to lactate in non-MYC-driven cells, which is sufficient to drive invasiveness. Using tube formation assays and in vivo experiments, we found that non-MYC-driven cells stimulate tumor growth by secreting DKK3 to promote tumor angiogenesis. Finally, we demonstrated that pharmacological targeting of LDHA significantly suppresses MB cell migration.

Materials and Methods

Key Resources

Experimental models, antibodies, chemicals, recombinant proteins, commercial assays, software, and algorithms are summarized in Supplementary Table S1.

Acquisition of Patient Samples

Patient samples were collected following written informed consent and study approval was obtained by the internal review board at the Necker Hospital for Sick Children (Paris, France, IRB: DC-2009).

Cell Culture and Viral Production

Cells were cultured according to Good Cell Culture Practice. In general, early passages of genetically manipulated cells (P1-P4) were used. Cells were mixed immediately prior to orthotopic injection. Lentivirus production was carried out by transfecting HEK293T with standard packaging vectors. Stable cell lines were sorted using flow cytometry (fluorescence activated cell sorting [FACS]) or selected via antibiotics. Concentration for selection agents was determined using a kill curve (Supplementary Table S2).

Fluorescence In Situ Hybridization

Vysis IntelliFISH was conducted according to the manufacturer’s instructions for chromosome 8 (8q11) and locus-specific 8q24 (MYC) probes.

Immunohistochemistry

Paraffin-embedded specimens were subjected to hematoxylin/eosin (H&E) and immunohistochemistry staining for MYC and CD34 according to standard methods.

Quantitative Real-Time PCR

Using the SV total RNA isolation system extracted RNA was reverse transcribed with M-MLV transcriptase. The resulting cDNA was subjected to amplification using GoTaq qPCR master mix. Primer sequences are shown in Supplementary Table S3.

Propidium Iodide Staining and Cell Cycle Analysis

Cells were fixed at -20°C with 70% ethanol. Fixed cells were incubated in PBS containing 150 units/ml of DNase-free RNase for 30 min at 37°C. After centrifugation, cells were resuspended in 200 µl of 50 µg/ml of propidium iodide (PI) for 2 hours in the dark. PI fluorescence was analyzed via FACS.

Live Cell Proliferation Assay

Cells were seeded in an E-Plate and the impedance signals were recorded using the xCELLigence.

Cell Doubling Time Calculation

Cell numbers were counted using Vi-CELL BLU. The doubling time was calculated as follows:

$${\displaystyle \begin{array}{l}{\displaystyle \frac{\mathrm{D}\mathrm{u}\mathrm{r}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}\times \mathrm{l}\mathrm{o}\mathrm{g}(2)}{{\displaystyle \mathrm{l}\mathrm{o}\mathrm{g}\left(\mathrm{F}\mathrm{i}\mathrm{n}\mathrm{a}\mathrm{l}\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{c}\mathrm{e}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}\right)-\mathrm{l}\mathrm{o}\mathrm{g}(\mathrm{I}\mathrm{n}\mathrm{i}\mathrm{t}\mathrm{a}\mathrm{l}\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{c}\mathrm{e}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n})}}}\end{array}}$$

Harvest of Conditioned Medium and Exosome Isolation

Conditioned medium (CM) collection was conducted as previously described.¹⁶ Exosomes for the spiking experiment were sedimented by ultracentrifugation at 150,000 × g for 2 hours. Exosomes for proteomic analysis were isolated using ExoQuick-TC PLUS. Exosome-free CM was generated using the Capturem Exosome Isolation kit.

Wound Healing Assay by Culture-Inserts

Ibidi culture-inserts were used according to the manufacturer’s recommendations. The percentage of wound closure from 0 hour was calculated using imageJ. Cell migration was calculated as follows:

$${\displaystyle \begin{array}{l}{\displaystyle \left(\frac{{(\mathrm{P}\mathrm{r}\mathrm{e}\text{-}\mathrm{m}\mathrm{i}\mathrm{g}\mathrm{r}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n})}_{\mathrm{a}\mathrm{r}\mathrm{e}\mathrm{a}}-{(\mathrm{M}\mathrm{i}\mathrm{g}\mathrm{r}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n})}_{\mathrm{a}\mathrm{r}\mathrm{e}\mathrm{a}}}{{(\mathrm{P}\mathrm{r}\mathrm{e}\text{-}\mathrm{m}\mathrm{i}\mathrm{g}\mathrm{r}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n})}_{\mathrm{a}\mathrm{r}\mathrm{e}\mathrm{a}}}\right)\times 100}\end{array}}$$

Transwell Cell Invasion Assay

The Corning Matrigel invasion chambers were used according to the manufacturer’s instructions. Briefly, 2.5 × 10⁴ cells were resuspended in 500 μL serum-free medium. The bottom chambers contained complete medium or CM.

Live Cell Migration Assay

2 × 10⁴ cells were seeded in a CIM-Plate 16 and migration was monitored for 72 hours using xCELLigence.

Liquid Chromatography-Mass Spectrometry Analysis

Sample preparation, Liquid chromatography-mass spectrometry (LC-MS) setting, and data processing were performed as described.¹⁶

Immunoblotting Analysis

Protein electrophoresis, immunolabelling, and subsequent chemiluminescence detection were performed as described by the manufacturer (Pierce). Anti-DKK3, anti-LDHA, and anti-MYC primary antibodies were used.

Vector Construction for Overexpression of MYC, LDHA, and DKK3

LeGO-iG2 and LeGO-iV2 plasmids (gifts from Boris Fehse)¹⁷ were used. The human MYC gene was amplified from pcDNA3.3_c_MYC (a gift from Derrick Rossi).¹⁸ BamHI and EcoRI restriction sites were introduced in PCR primers (Supplementary Table S4). The digested PCR product and plasmid were ligated using LigaFast rapid DNA ligation. LDHA overexpression plasmid was made by subcloning LDHA gBlocks fragment into the EcoRI and Mscl sites of LeGO-iV2. HaloTag-DKK3 overexpression plasmid was made by subcloning HaloTag-DKK3 gBlocks fragment into BamHI and NotI sites of LeGO-iG2.

Staining of Actin Filaments, Nucleus, and Secretome

Cells were incubated with CM-OE-LDHA-iV2 or CM-OE-HaloTag-DKK3 at 37°C for 24 hours. For labeling HaloTag-DKK3, growth medium was replaced with medium containing the Oregon Green ligand. Cells were stained with 4′,6-diamidino-2-phenylindole for nucleus and Alexa Fluor 594-Phalloidin for cytoskeleton, as described by the manufacturer.

CRISPR Activation and Inhibition

pLVhU6-sgRNA hUbc-dCas9-KRAB-T2a-Puro (a gift from Charles Gersbach)¹⁹ and lenti-sgRAN(MS2)_puro (a gift from Feng Zhang,)²⁰ were used. The guide RNAs (Supplementary Table S5) were designed on the basis of the CRISPR Design service engine. Complementary sgRNAs were annealed and ligated to BsmBI digested plasmids.

Measurement of Pyruvate and Lactate

The cell pellet extract was prepared for LC-MS measurement as described.²¹

Tube Formation Assays

Commercially available human umbilical vein endothelial cells (HUVECs) were used within five passages. The endothelial cell tube formation was evaluated at the end of 18 hours of incubation with human recombinant DKK3 peptide (rhDKK3) or CM. Tube formation was assessed in three randomly selected fields of view per well and tube length was quantified using the Angiogenesis Analyzer for ImageJ.

Vessel Quantification

Vessel number was identified by CD34 staining. Counting rules were defined in which every CD34-positive object, no matter how small, should be counted, except suspected CD34-positive monocytes, macrophages, and tumor cells. Furthermore, if interconnected, stained objects were considered as a single vessel.

Establishment of Xenografts

Six- to eight-week-old NOD.Cg-Prkdc^scidIl2rg^tm1Wjl/SzJ mice were purchased. In accordance with Directive 2010/63/EU (Decreto-Lei No. 113/2013), animal procedures were approved by the institutional animal welfare body (ORBEA-iMM). Mice were injected intracranially (250,000 cells/3 µL/mouse) in the right cerebellar hemisphere.

Human scRNA-seq Data Processing

Cells that lacked CNVs and/or clustered with normal reference populations were excluded. Downstream analyses were performed using the Seurat framework in R, including Uniform Manifold Approximation and Projection (UMAP), which was calculated using Seurat’s RunUMAP function and visualized using the DimPlot command.

Statistical Analysis

Cell indices for real-time proliferation and migration were calculated by the RTCA Software. Dose-response curves were generated using nonlinear regression (log(concentration of inhibitor) vs response). Biostatistics data were analyzed using Prism8. Numerical data were expressed as a mean ± SEM. The P-values were determined by unpaired t-tests. P < .05 was considered as statistically significant.

Results

ITH of MYC Amplification, mRNA, and Protein Expression

G3-MB has increased MYC expression compared to normal brains and other subgroups (published data sets^11,22,23Supplementary Figure S1A). To explore the cellular diversity of MYC amplification in G3-MB, we performed FISH assays using paraffin-embedded primary tumor sections of MB patients. A mixture of MYC-amplified and non-amplified tumor cells (Figure 1A) was detected. Subsequently, highly variable MYC expression within G3-MB patients (Figure 1B) was demonstrated using two sets of recently published single-cell sequencing data.^1,4 In addition, using immune cells and oligodendrocytes as references, UMAP analysis showed cancer cells clustered by patient, whereas normal cells clustered together regardless of patient source (Supplementary Figure S1B). The variable expression of MYC was present across other subgroups (Supplementary Figure S1C and SD) and non-tumor cells. Immunohistochemistry showed markedly different MYC protein levels between tumor cells within the same tumor. MYC-positive and MYC-negative tumor cells were clearly detected in tumor sections of G3-MB patients using brown or blue nuclear staining (left, Figure 1C), respectively, but were not detected in tumor sections from SHH-MB patients (right, Figure 1C). Collectively, these results indicate that MYC amplification, mRNA expression, and protein levels are heterogeneous among tumor cells within G3-MB.

Fig. 1.

Subclonal heterogeneity of MYC amplification and expression patterns in Group 3 (G3) medulloblastoma (MB). (A) FISH analysis demonstrating MYC non-amplified and MYC-amplified tumor cells in a primary tumor section of a G3-MB patient. Green signals correspond to MYC signals and red signals correspond to a chromosome 8 centromere probe. Cell nuclei are stained with DAPI. (B) Relative MYC mRNA expression at the single-cell level shows variable MYC expression in tumors of G3-MB patients from two separate data sets.^1,4 The red line indicates the overall mean of MYC mRNA expression. (C) Immunohistochemical staining for MYC protein expression (in brown) in primary tumor section of patients with a G3-MB (left) or a SHH-MB (right) (scale bars, 50 µm). Note heterogeneity of nuclear MYC protein expression in the G3-MB tumor cells while the SHH-MB tumor cells lack MYC immunostaining. Sections are counterstained with hemalum.

Orthotopic Injection of Mixed MBMYC− and MBMYC+ Cells Resulted in Shorter Survival in Comparison to Injection of MBMYC− Cells

To model the oncogenic effects of a heterogeneous tumor cell population, two MB cell lines (ONS76 and UW228-3) with comparatively low MYC expression were selected (Supplementary Figure S2A) and engineered to overexpress MYC (Supplementary Figure S2B and C). Cells with stable overexpression (MB^MYC+) showed a significantly increased growth rate compared to control cells (MB^MYC−) (Supplementary Figure S2D and E). We engrafted these isogenic cells into animals as either a single homogenous population or in a mixed ratio of MB^MYC−:MB^MYC+ (1:1). Mice orthotopically injected with mixed cell populations showed significantly shorter survival compared to mice engrafted with only MB^MYC− cells. There was no significant survival difference between mice engrafted with only MB^MYC+ cells and those receiving mixed cell populations (Figure 2A). Tumors derived from co-injected mixed cell populations showed significantly higher MYC expression compared to controls (Figure 2B). However, in co-injected mice, we noticed heterogeneous MYC protein expression patterns (Figure 2B). Between 57% and 70% of cells in ONS76 co-injected mice, and between 52% and 78% of cells in UW228-3 co-injected mice have high MYC protein expression ((Supplementary Figure S2F).

Fig. 2.

Orthotopic co-injection of MB^MYC− and MB^MYC+ cells correlates with poor survival and increased spinal subarachnoid dissemination in mice. (A) Kaplan-Meier survival curves for murine orthotopic models (n = 5/group) injected with MB^MYC− cells, MB^MYC+ cells, or 1:1 mixed cells. The P-values were determined by logrank test. (B) H&E and immunohistochemistry indicating MYC protein expression in mice bearing tumors after injection of MB^MYC− (left) cells, MB^MYC+ (middle) cells, and 1:1 mixed (right) cells (scale bars, 50 µm). (C) H&E showing MB cells localizing in the spinal leptomeninges (scale bars, 50 µm). (D) Animal numbers presenting with subarachnoid dissemination. For each experimental condition, spinal cords of five mice were harvested and divided into six segments each. Each segment was microscopically investigated for subarachnoidal tumor. Presence of subarachnoidal tumor spread was scored for each section as either “tumor cell positive” or “tumor cell negative”.

G3-MB frequently disseminates via the cerebrospinal fluid to the brain and spinal cord.⁵ Thus, we examined spinal cord samples (Figure 2C) and observed a marked increase in the incidence of leptomeningeal dissemination (LMD) from baselines of 20% (n = 1 of 5) (ONS76^MYC− alone) or 0% (UW228-3^MYC− alone) to 100% (n = 5 of 5), in mice engrafted with mixed populations of MB^MYC− and MB^MYC+. There was no significant difference between mice that received mixed injections compared to homogenous MB^MYC+ injections (Figure 2D). Next, we showed that MYC mRNA expression was strikingly similar in the primary tumor and the metastases of the same patient (published dataset²⁴Supplementary Figure S3A). These data support a model in which cellular heterogeneity, specifically interactions between non-MYC-driven and MYC-driven subclones, might stimulate the metastatic propensity of non-MYC-driven cells.

CM Derived From MYC-Overexpressing MB Cells Promotes Migration

To model the LMD-promoting effect of intercellular communication, we used in vitro co-culture systems to determine cell migration and invasiveness. In the wound healing assay, co-cultured MB^MYC− and MB^MYC+ cells showed faster-wound closure compared to MB^MYC− cells alone (Figure 3A). Counting the number of invaded cells in the Boyden chambers established that co-cultured cells exhibited increased invasiveness (Supplementary Figure S3B). We noticed that the gap was closed by mCherry-labeled MB^MYC− cells and GFP-labeled MB^MYC+ cells (Figure 3A). Together, these data show that MB^MYC− and MB^MYC+ cells migrate cooperatively, so cell-cell interactions might be crucial in promoting this behavior.

Fig. 3.

Secretome from MB^MYC+ cells promotes cell migration of non-MYC-driven medulloblastoma cells. (A) Wound healing assay to assess migration. Photographs were taken at 12–18 h after wound induction (left) (scale bars, 50 µm). The red color indicates mCherry-labeled MB^MYC− cells and the green color indicates GFP-labeled MB^MYC+ cells. The detected migration was shown as a percentage of wound closure from 0 h (right). (B) Real-time migration of non-MYC-driven cells in the presence of CM derived from MYC-overexpressing and corresponding control cells. (C) Real-time migration of non-MYC-driven cells in the presence of CM derived from MB^MYC+ or exosome-depleted CM from MB^MYC+. (D) Real-time migration of non-MYC-driven cells in the presence of CM derived from MB002 or exosome-depleted CM from MB002. Experiments were repeated three times.

To study whether factors secreted by MB^MYC+ cells influence the migration of MB^MYC− cells, we performed real-time cell migration assays using CM. MB^MYC− cells treated with CM-MB^MYC+ showed enhanced migration compared to cells treated with CM-MB^MYC− (Figure 3B). Short-term boiling denatures most proteins.²⁵ Therefore, our finding that boiled CM showed almost no ability to increase cell migration (Supplementary Figure S3C) demonstrated that secreted proteins from MB^MYC+ cells most likely stimulated migration of MB^MYC− cells.

Identification and Validation of Differentially Secreted Proteins in CM From MBMYC+ and MBMYC⁻Cells

To address problems associated with identification of candidate proteins directly within CM,²⁶ we used exosomes for candidate identification. We removed exosomes from CM-MB^MYC+ and CM of patient-derived MYC amplified cells (MB002)²⁷ using capture spin columns. We then compared the migratory behavior of MB^MYC− cells cultured with CM or exosome-depleted CM. Cells cultured with exosome-depleted CM showed a significant reduction in migration (Figure 3C and D). Furthermore, we isolated exosomes from CM using ultracentrifugation and added isolated exosomes from the “donor” cells to “recipient” cells. We observed a substantial increase in cell motility in “recipient” cells incubated with exosomes isolated from CM-MB^MYC+ compared to exosomes from CM-MB^MYC− (Supplementary Figure S3D). These data show that exosomes from MB^MYC+ cells stimulate migration of MB^MYC− cells and can thereby be used to identify functional secreted proteins.

To identify the MYC-dependent secretome, isolated exosomes from MYC-overexpressing ONS76 and corresponding control cells were subjected to a proteomic analysis. We detected a total of 629 proteins, and 352 proteins were quantified. According to a survey of the exosome database Exocarta,²⁸ the majority of identified proteins (95%) were exosomal. Among these exosomal proteins, 25 were significantly more and 15 were significantly less abundant in ONS76^MYC+ compared to ONS76^MYC− (Figure 4A). To obtain validation candidates, we performed an additional experiment using CM from non-MYC-driven (ONS76, Daoy, UW228-3) and MYC-driven (Med8A, D341, D283) MB cells. In both independent proteomics analyses, LDHA, a key glycolytic enzyme, and DKK3 were identified as significant candidates (Supplementary Tables S6 and S7). Using an integrated data set,²⁹ we showed that mRNA expression of LDHA, as an enriched protein from the secretome of MB^MYC+, strongly correlated with mRNA expression of MYC (Supplementary Figure S4A) and LDHA protein levels were significantly increased in G3-MB (Supplementary Figure S4B).^8,30 Additionally, survival analysis showed that increased LDHA mRNA expression is associated with decreased overall survival (Supplementary Figure S4C). As an enriched protein from the secretome of MB^MYC−, DKK3 was the candidate with the highest fold change resulting from the comparison of ONS76^MYC− and ONS76^MYC+ (Supplementary Figure S5A). The mRNA expression levels of DKK3 were inversely correlated with those of MYC (Supplementary Figure S5B),²⁹ and DKK3 protein levels were significantly lower in G3-MB (Supplementary Figure S5C).⁸ Based on these observations, LDHA and DKK3 were selected for further validations.

Fig. 4.

Proteomic analysis identifies DKK3 and enzymatically active LDHA as secreted proteins. (A) Significant exosome proteins are displayed as black dots (cutoff: P < 0.05 and fold-change > 1.5). DKK3 is labeled orange and LDHA is labeled blue. (B) Immunoblotting quantification (above) by normalization to total loading proteins (Supplementary Figure S6A). Immunoblotting of DKK3 and LDHA (below). (C) Confocal z-stack of single UW228-3 cell: cell nucleus (DAPI/blue), uptaken LDHA (Venus/green), and cytoskeleton (Alexa Fluor-594-Phalloidin/red). (D) LC-MS analysis of LDHA enzyme activity of wildtype cells treated with CM-MB^MYC− or CM-MB^MYC+. (E) Immunoblotting of LDHA expression in CM from control or LDHA transcriptionally activated cells (left). Immunoblotting quantification (right) by normalization to total loading proteins (Supplementary Figure S6F). (F) LC-MS analysis of LDHA enzyme activity of wildtype cells treated with CM derived from control or from LDHA transcriptionally activated cells. Experiments were repeated three times.

To verify our MS-based protein identification, immunoblotting analysis was employed to determine protein levels of secreted DKK3 or LDHA and normalized to the total loading protein (Supplementary Figure S6A). We confirmed increased secretion of DKK3 in MB^MYC− cells and increased secretion of LDHA in corresponding MB^MYC+ cells (Figure 4B).

Uptake of Secreted and Enzymatically Active LDHA Stimulates Migratory Propensity of non-MYC-Driven Cells

We generated a fluorescent fusion protein overexpression system (Supplementary Figure S6B and C) to monitor the intercellular traffic of LDHA in UW228-3. Freshly collected CM was filtered and applied directly to non-labeled cells. To confirm the intracellular localization of LDHA taken up by “recipient” cells, confocal microscopy was employed, which allowed for three-dimensional segmentation of intracellularly localized LDHA (green fluorescence), cell nucleus (blue fluorescence), and cytoskeleton (red fluorescence) (Figure 4C).

To test whether the LDHA was taken up into cells still retains enzyme activity, an LC-MS-based method was used for quantification of cellular metabolites. Since LDHA catalyzes the conversion of pyruvate to lactate, the ratio of lactate to pyruvate provides a measure of its enzymatic activity. After 24 hours treatment with CM, pellets of MB^MYC− cells were analyzed. Strikingly, compared with CM-MB^MYC− treated cells, CM-MB^MYC+ treated cells showed enhanced turnover from pyruvate to lactate (Figure 4D). This finding not only demonstrated the aberrant secretion of LDHA from MB^MYC+ cells, but also showed that the secreted LDHA taken up by cells is enzymatically active. We then generated LDHA overexpressing MB^MYC− (ONS76 and UW228-3) cells using a CRISPR/Cas9-based transcriptional activation system (Supplementary Figure S6D and E). This effective transcriptional activation remarkably induced the presence of extracellular LDHA protein (Figure 4E, total protein was calculated using TGX gel (Supplementary Figure S6F)). Due to the uptake of LDHA, MB^MYC− cells treated with CM from the corresponding LDHA overexpressing cells showed distinctly enhanced conversion of pyruvate to lactate (Figure 4F). In summary, we established that MB^MYC+ cells release LDHA and extracellular LDHA retains its enzyme activity; LDHA can then be taken up by MB^MYC− cells, thereby promoting the Warburg effect.

LDHA catalyzes the last step of glycolysis. Aerobic glycolysis is considered a metabolic signature for invasive cancer.³¹ Notably, overexpression of LDHA increased cell migration (Figure 5A) and invasion of MB cells (Figure 5B). In contrast, knockdown of LDHA (Supplementary Figure S6G and H) decreased MB cell invasion (Figure 5C). Conditioned media from LDHA overexpressing MB cells corroborated this effect (Figure 5D and E). Furthermore, we tested whether pharmacological inhibition of LDHA with a specific LDHA inhibitor (GSK 2837808A)³² reduces migratory properties of MB cells. Growth of MB^MYC+ cells was inhibited by GSK 2837808A with lower IC50 compared to MB^MYC− cells (Supplementary Figure S7A). Notably, treatment of mixed MB^MYC+ and MB^MYC− cells with low concentration of GSK 2837808A significantly reduced cell motility (Figure 5F). Importantly, suppression of cell migration was not due to reduced overall cell number since, that is, there was no change in cell growth under the concentration of treatment (Supplementary Figure S7B and C).

Fig. 5.

Secreted LDHA stimulates migratory propensity of non-MYC-driven medulloblastoma cells and pharmaceutical inhibition of LDHA effectively represses cell migration. Real-time migration (A) and invasion (B) of control and LDHA transcriptionally activated cells. (C) Relative cell invasion of MB002 and HD-MB03 upon LDHA knockdown. Real-time migration (D) and invasion (E) of MB^MYC− cells in the presence of CM from control or from LDHA transcriptionally activated cells. (F) Real-time migration of MB^MYC− and MB^MYC+ mixed cells (1:1) treated with LDHA inhibitor GSK 2837808A compared to DMSO control treatment. Experiments were repeated three times.

Together, these results show that secreted LDHA from MB^MYC+ cells promotes metastatic propensity and invasiveness of MB^MYC− cells and hence constitutes a possible pharmacological target.

DKK3 Promotes Tumor Angiogenesis

To determine the growth dynamics of MB^MYC− and MB^MYC+ subpopulations in vitro, we monitored the change in cell populations under direct co-culture conditions. Notably, in the MB^MYC+ and MB^MYC− co-cultures continuous overgrowth of MB^MYC+ cells was detected (Figure 6A). Thus, MYC-dependent proliferation resulted in subclonal expansion of MB^MYC+ cells, a commonly observed phenomenon when in vitro cell lines are established from MBs with subclonal MYC amplification.²⁷ Significant differences between in vivo and in vitro could be explained as in vitro 2D culture systems fail to comprehensively imitate the in vivo architecture and microenvironment.³³ Since angiogenesis is required for invasive tumor growth and metastasis,³⁴ we next compared the difference in angiogenesis between tumors from mice engrafted with MB^MYC−, MB^MYC+, or mixed cells. Highlighting the vessels by staining their endothelial cells for the endothelial cell marker CD34 (Figure 6B) showed the average densities of the CD34-positive microvessels in MB^MYC−, MB^MYC+, and mixed groups were 48 ± 4.3, 23 ± 1.5, and 39 ± 3.6, respectively. This result suggests that MB^MYC− cells may promote tumor angiogenesis.

Fig. 6.

Secreted DKK3 promotes tumor angiogenesis in medulloblastoma. (A) FACS measurement of GFP-positive MB^MYC+ cells. GFP-positive cells presented as percentage of total cells (GFP- and mCherry-positive cells). Experiments were repeated three times. (B) Immunohistochemical staining of CD34 indicated microvessel density in tumors from mice injected with control, MB^MYC+ or mixed cells (left, scale bar, 50 µm), and average vessels per high-power field (HPF, right). The quantification was done by counting microvessel in three to six random fields (20×) of tumor sections from five mice. (C) Confocal z-stack: cell nucleus (DAPI/blue), uptaken DKK3 (Oregon-Green/green), and cytoskeleton (Alexa Fluor-594-Phalloidin/red) are indicated. (D) HUVEC tube formation assay with the addition of denatured rhDKK3 or rhDKK3 protein (left). The quantification of relative tubule length (right). Experiments were repeated five times. (E) Immunoblotting analysis of DKK3 in CM from control or DKK3 knockdown cells (above). Immunoblotting quantification (below) by normalization to total loading proteins (Supplementary Figure S8E). Experiments were repeated three times. (F) Relative tubule length from HUVEC tube formation assay in the presence of CM from control or DKK3 knockdown cells. Experiments were repeated three times. (G) The quantification of average vessels per high-power field (HPF) from immunohistochemical staining of CD34 in tumors from mice injected with MYC-overexpressing UW228-3 mixed with corresponding control cells or MYC-overexpressing UW228-3 mixed with DKK3 knockdown UW228-3. The quantification was done by counting microvessel in three to six random fields (20×) of tumor sections from six mice. (H) Kaplan-Meier survival curves for murine orthotopic models (n = 10/group) injected with MB^MYC+ + MB^MYC− cells (black) and MB^MYC+ + DKK3-KD-MB^MYC− cells (orange). The P-values were determined by logrank test. ns means nonsignificant.

DKK3 displays angiogenic activity and is involved in remodeling the tumor vasculature.³⁵ To monitor secretion and uptake of DKK3, we investigated DKK3 fusion protein overexpression in UW228-3 (Supplementary Figure S8A and B). Presence of green fluorescent dots around the nuclei of wildtype cells indicated that DKK3 released from “donor” cells was taken up by “recipient” cells (Figure 6C). To evaluate the pro-angiogenic effects of DKK3, we plated HUVECs on matrigel, exposed them to rhDKK3 protein. The addition of rhDKK3 to the medium promoted HUVEC tube formation (Figure 6D). We then achieved stable transcriptional repression of DKK3 in MB^MYC− cells using dCas9-based CRISPR interference (Supplementary Figure S8C and D). This effective suppression of cellular DKK3 protein expression led to nearly 100% decrease of DKK3 in the CM (Figure 6E, total protein was calculated using TGX gel (Figure S8E)). DKK3-depleted CM-treated HUVECs showed a significantly reduced endothelial cell tube formation compared to control CM-treated HUVECs (Figure 6F and Supplementary Figure S8F). These findings indicate that MB^MYC− cells stimulate angiogenesis via the secretion of DKK3. We performed in vivo orthotropic injection with mixed cells, in which MYC-overexpressing UW228-3 cells were mixed with either corresponding control cells or DKK3-depleted control cells. Measurement of vascularization using CD34 staining demonstrated that knockdown of DKK3 significantly reduced microvessel density (Figure 6G and Supplementary Figure S8G), and extended animal survival (Figure 6H). In MB patients, vascularization-related survival could be demonstrated using a multitude of angiogenesis-related factors (Supplementary Figure S8H). Thus, our results demonstrate that secreted DKK3 from non-MYC-driven MB cells promotes neoangiogenesis in high-risk MB.

Discussion

ITH is critical for treatment resistance, cancer progression, and disease relapse,³⁶ and is highly prevalent in MB.^1,3,9 In-depth biological understanding of the specific driving forces and functional interactions between different subclones within tumors holds great potential for the development of more effective cancer therapies.³⁷ A recent study elegantly demonstrated that the analysis of extracellular vesicles and intercellular communication may improve identification of high-risk MBs across subgroups.^38,39 G3-MBs harbor a number of genetic alterations,⁵ including ITH of MYC amplification and expression, a property that challenges our understanding of the underlying tumor biology and our ability to achieve effective therapeutic responses. Here, we reveal that the secretome of MYC-driven and non-MYC-driven cellular subclones synergistically drives MB progression. Our data support a model in which LDHA from MYC-driven cells facilitates metastatic seeding and outgrowth, while non-MYC-driven cells, in turn, support tumor angiogenesis through the release of DKK3.

Our model recapitulates the aggressive features observed in G3-MB patients by showing that mice injected with mixed cells have significantly reduced survival and increased LDM compared to those injected with non-MYC-driven cells. In vitro, the increased migration of non-MYC-driven cells was observed after incubation with the CM from MYC-driven cells. Thus, our results demonstrate that the secretome of MYC-driven cells stimulates metastasis formation of non-MYC-driven cells.

Our proteomic approach identified LDHA as a protein secreted by MYC-driven MB cells. LDHA, as an MYC target, is utilized by cancer cells to bypass oxidative phosphorylation by reducing pyruvate to lactate,⁴⁰ which is a recognized fuel for tumors.⁴¹ Abnormal expression of LDHA has dramatic implications for glioma development by promoting migration and invasion in high-grade glioma.⁴² We demonstrated that cellular uptake of extracellular LDHA promotes glycolysis and invasiveness in non-MYC-driven MB cells. Thus, the secretome causes metabolic reprogramming.

Intriguingly, we have shown that tumors from mice co-injected with MYC-driven and non-MYC-driven cells have increased angiogenesis compared to mice injected with only MYC-driven cells. We identified DKK3 as a secreted protein from “innocuous” non-MYC-driven cells. High expression of DKK3 is associated with aggressive breast, colorectal and ovarian cancers.⁴³ Additionally, DKK3 displays angiogenic activity and is involved in remodeling the tumor vasculature.³⁵ Recent publications have demonstrated that DKK3 is capable of directly differentiating human fibroblasts into functional endothelial cells.⁴⁴ We demonstrate the pro-angiogenesis capabilities of DKK3 simply by adding purified rhDKK3 to the extracellular milieu of HUVEC. A significant decrease in vascularization was also identified in tumors from mice co-injected with MB^MYC+ cells and DKK3-depleted MB^MYC− cells compared to co-injection of MB^MYC+ and MB^MYC− cells. MYC overexpression triggers tumor development, and, due to rapid tumor growth, all mice were dead within 15 days. Regardless, a significant but modest improvement in median survival was noted in mice that were co-injected with MB^MYC+ cells and DKK3-depleted cells. Both in vitro and in vivo data point to a previously unknown role of secreted DKK3 in the stimulation of MB angiogenesis.

G3-MB is responsible for the majority of deaths among MB patients. The evidence from our study that secretome-mediated interactions of subclones stimulated G3-MB growth and metastasis suggests that cancer treatment strategies may need to be adapted to target these interactions. LDHA is considered a safe therapeutic target and loss of LDHA protein has only relatively mild symptoms of exertional myopathy.⁴⁵ As recently shown inhibition of LDHA significantly reduces the growth of MYC-driven MBs but has little effect on normal cerebellar cells.⁴⁶ We showed that inhibition of LDHA by a small-molecule inhibitor more effectively reduced the proliferation of aggressive MYC-driven cells than non-MYC-driven cells. More interestingly, treatment of mixed MB cells using a low dose of the LDHA inhibitor significantly reduced cell migration. These findings suggest that inhibition of LDHA may have potential for patients with MYC-driven MB, however, this hypothesis needs to be proven in clinical studies.

In summary, we identified a secretome-mediated network between MYC-driven and non-MYC-driven subclones in aggressive G3-MB. This network actively maintains the heterogeneity of MYC, which is not only a substrate for MB evolution but also promotes and is required for continued tumor development and progression. Further understanding of the cooperative relationship between MYC-driven and non-MYC-driven subclones in tumor growth and progression may support the development of effective therapeutic strategies.

Funding

This work was supported by grants from José-Carreras Foundation (DJCLS 21R/2019, U.F., M.R.), Elterninitiative Kinderkrebsklinik e.V. (M.R., N.Q.), “Förderverein Löwenstern” to A.B., the Gert-und-Susanna-Meyer foundation (M.R., J.B.), Deutsche Forschungsgemeinschaft (RE 2857/2-1 to M.R., RE938/4-1 to G.R., KFO 337 to M.R., CRC/TRR 205, TRR205 to G.E.), the German Cancer Aid (111537 to G.R.), the Research Commission of the Medical Faculty, Heinrich Heine University Düsseldorf (FOKO 2015-46 to M.R. and N.Q., FOKO 2019-06 to N.Q.), the German Childhood Cancer Foundation (DKH 2021.20 to M.R. and N.Q.), Fundação Millennium bcp and Fundação Amélia de Mello (to C.C.F.), and FOKO 2021-44 to M.R.

Conflict of interest statement. No conflicts of interest were declared.

Authorship statement. N.Q., G.R., A.B., G.E., C.C.F., and M.R. designed the experiments, analyzed the data, and wrote the manuscript. E.P., M.L., D.P., C.C., R.C., F.-D.M., L.B., S.G., J.B., N.B., J.G., J.C.B., A.S., J.T.B., K.S. and J.F. carried out experiments. B.M., U.F., G.R., and A.B. helped with data interpretation. G.R. evaluated immunohistochemistry. S.P. O.A., provided FFPE slides and inputs.