ABSTRACT
Oligodendrogliomas are lower-grade, slow-growing gliomas that are ultimately fatal. Although driver mutations are known, the mechanisms underlying their signature slow growth rates are poorly understood. We found evidence for intra-tumoral interactions between neoplastic and non-neoplastic cells in oligodendroglioma tissues. To further study these cell interactions, we used two patient-derived oligodendroglioma cell lines of lower and higher aggressivity. Both oligodendroglioma cell lines released extracellular vesicles that had cytotoxic effects on non-neoplastic and neoplastic cells, but each had distinct vesicular proteomes. Consistent with extracellular vesicles mediating growth inhibitory effects in oligodendrogliomas, higher expression levels of several extracellular vesicle biogenesis genes (SMPD3,TSG101, STAM1) correlates with longer survival in oligodendroglioma patients. Furthermore, SMPD3 overexpression slows oligodendroglioma cell growth in culture. Conversely, SMPD3 knockdown enhances oligodendroglioma proliferation in vitro, in murine xenografts, and in human cerebral organoid co-cultures. Oligodendroglioma-derived extracellular vesicles thus mediate tumor cell microenvironmental interactions that contribute to low aggressivity.
INTRODUCTION
Gliomas are a heterogenous group of primary glial brain tumors that are composed of a mixture of neoplastic glial cells (‘tumor cells’) and non-neoplastic ‘stromal cells’ that include non-neoplastic glia, neurons, and a variety of other inflammatory and vascular cells1, 2. Gliomas are classified based on their molecular and histological features3, 4. The most aggressive and most frequently studied is glioblastoma multiforme (GBM), a high-grade astrocytoma with 5-year survival rates of less than 5%5. In contrast, lower-grade gliomas (World Health Organization– WHO stage II/III tumors), which include oligodendroglioma (ODG), are slower growing tumors, with median survival times between 9-14 years6. Glioma subtypes each have a unique constellation of mutations, that in ODG typically includes isocitrate dehydrogenase 1 (IDH1) or IDH2 mutations, chromosomal 1p/19q co-deletion, and mutation of Capicua (CIC), a transcriptional repressor, in the retained 19q allele 7, 8, 9, 10. While tumor cells carry driver mutations that are responsible for the cells neoplastic transformation, non-neoplastic cells in the tumor microenvironment form part of the ecosystem that sustains tumor cell proliferation and growth11. Understanding how glioma cells interact with cells in the microenvironment is thus essential to understanding disease progression.
Extracellular vesicles (EVs) are important mediators of intercellular communication. EVs package lipids, protein, DNA and RNA in a lipid bilayer that protects cargo from degradation in the extracellular space and facilitates membrane fusion and delivery of bioactive material to neighboring cells12. Small EVs (sEVs; 40-200nm), also known as exosomes, are generated via different biogenic enzymes, including endosomal sorting complex related transport (ESCRT)-dependent and -independent pathways12. ESCRT-independent biogenesis is mediated by sphingomyelin phosphodiesterase 3 (SMPD3), encoding neutral sphingomyelinase 2 (nSMase2), which produces ceramide required for exosome generation and budding13. EVs are secreted by most if not all cells in the brain, including oligodendrocytes14, astrocytes15, neurons16, as well as cancerous cells17. There are now several examples of glioma cells interacting with cells in the tumor niche via EVs18, 19. Strikingly, while EVs from higher-grade gliomas are generally growth promoting, the few studies conducted with ODG suggest that ODG-derived EVs may be cytotoxic20, 21.
Here, we further queried how vesicular factors contribute to the slow growth properties of ODG, using patient using patient tissue samples, patient-derived ODG cell lines9, and publicly available molecular and clinical glioma data. We found evidence that ODG cells exert growth inhibitory effects non-cell autonomously in part through the secretion of cytotoxic vesicular factors. Accordingly, higher expression levels of several EV biogenic genes (SMPD3, TSG101, STAM1) correlates positively with longer survival in low-grade glioma patients. We focused on SMPD3 and found that it is a critical regulator of cytotoxic EV biogenesis in ODG. SMPD3 negatively regulates ODG growth in cultured cells, in mouse xenografts, and in human cerebral organoid co-cultures. We conclude that ODG cells negatively modulate tumor growth in part through the secretion of vesicular factors that have homotypic and heterotypic cytotoxic effects on cells in the tumor niche.
RESULTS
Oligodendroglioma cells exert non-cell autonomous effects within the tumor niche
To determine whether lower-grade glioma cells interact non-cell autonomously with neighboring non-neoplastic cells (heterotypic interactions) in the tumor microenvironment, we analyzed surgical resection specimens from five IDH-mutant, 1p/19q-codeleted ODG patients (Table 1). We imaged tumor sections in the more confluent tumor core and in peritumoral infiltrating regions, where we reasoned that tumor cells could influence non-neoplastic reactive cells. To identify tumor cells, sections were immunolabeled with an antibody to IDHmR132H (hereafter IDHm)22. As expected, co-staining with IDHm and OLIG2, an oligodendrocyte lineage marker highly expressed in ODG23, showed high co-localization of IDHm and OLIG2 in all samples (Fig. 1A-E).
We then asked whether IDHm-negative non-neoplastic cells in the core and peritumoral regions were influenced by the tumor by assessing proliferation, a rare event in non-cancerous adult brains24. Co-labeling with IDHm and Ki-67, a pan-proliferative marker, showed that a surprisingly large fraction of Ki-67+ cells (83-88%) were IDHm-negative (Fig. 1F-J). Proliferation of non-neoplastic cells in the tumor mass suggested that tumor cells could secrete growth factors or other signals that modulate the behavior of neighboring cells. Consistent with the presence of non-cell autonomous signaling in the microenvironment, we detected MEK/ERK pathway activation (as evidenced by pERK staining) in IDHm-negative cells (Fig. S1A-E, Table S1).
Although correlative in nature, these findings are consistent with the idea that IDH mutant ODG cells may communicate with neighboring stromal cells (neural, vascular, inflammatory) in the tumor niche in a non-cell autonomous fashion (Fig. 1K).
ODG secretomes have distinct bioactive effects but both include cytotoxic EVs
To mediate contact-independent communication, tumor cells and neighboring non-neoplastic stromal cells secrete a variety of bioactive molecules (e.g. growth factors, cytokines, interleukins) that are enriched in the tumor microenvironment25. In vitro, bioactive factors are found in tumor conditioned media (CM), as soluble factors and/or in EVs12. To further understand how ODG cells interact with cells in the tumor microenvironment, we used two different patient-derived cell lines established from IDH-mutant, 1p/19q co-deleted anaplastic ODGs, termed BT088 and BT054 cells9. Both cell lines can self-renew and grow in vitro, but only BT088 cells, isolated from a higher grade III tumor, retain the ability to form tumors after xenografting in immunocompromised mice9. Consistent with reported differences in tumor growth in vivo, using real-time live cell imaging, we found that BT088 cells have a 1.76-fold shorter doubling time than BT054 cells when grown in vitro (Fig. 2A, Table S1).
To ask whether BT088 and BT054 ODG cells might influence the growth/survival of neighboring cells in the tumor niche, we tested the bioactivity of secreted factors on two cell types - embryonic neural stem cells (eNSCs) and ODG cells themselves (Fig. 2B). While eNSCs are not present in the adult tumor niche, they were chosen as a surrogate cell type as their growth is highly sensitive to external cues26. Embryonic day (E) 12.5 NSCs isolated from the cortex were first plated at clonal density in fresh media (FM), or CM collected from the two ODG cell lines. After 7 days in vitro (DIV), neurosphere number (measure of activated NSCs), neurosphere size (aggregate measure of proliferation and apoptosis) and live cell number (measure of cell survival) were quantitated (Fig. 2C-F). Compared to FM, NSCs grown in BT088-CM gave rise to more neurospheres that were larger in size (Fig. 2C-I, all counts in Table S1). However, the larger spheres had markedly fewer live cells (Fig. 2J, Table S1). In contrast, BT054 CM did not alter neurosphere number, and it had an inhibitory effect on both neurosphere size and live cell number compared to FM (Fig. 2K-N). Thus, while BT088 CM initially promotes NSC revival from quiescence and NSC proliferation, cell progeny have decreased viability. In contrast, BT054 CM in aggregate is not pro-proliferative, but like BT088 cells, BT054 cells also secrete cytotoxic factors.
The bioactive effects of ODG CM could be due to soluble and/or EV-enclosed factors. To dissect this activity further, we repeated the neurosphere assay using CM in which EVs were removed by sequential centrifugation (CM-EV). E12.5 NSCs grown in BT088 CM-EV formed the same number of neurospheres as NSCs grown in BT088 CM, however, there was a further increase in neurosphere size, and a striking increase in live cell number (Fig. 2C-J; Table S1). Similarly, E12.5 NSCs grown in BT054 CM-EV formed the same number of neurospheres as NSCs grown in BT054 CM, and there was an increase in neurosphere size, although live cell number was not altered (Fig. 2K-N; Table S1). Thus, BT088 CM contains pro-proliferative factors that are largely soluble, and cytotoxic factors that are mainly vesicular. In contrast, BT054 CM contains soluble pro-proliferative factors, the activity of which is masked by cytotoxic factors that are both soluble and vesicular.
Next, to directly assess EV bioactivity, EVs were added to FM (FM+EV). BT088 EVs added to FM reduced neurosphere number, neurosphere size and live cell number compared to FM (Fig. 2C-J, Table S1). In contrast, BT054 EVs added to FM did not alter neurosphere number, while neurosphere size and live cell number were both reduced compared to FM (Fig. 2K-N, Table S1). Thus, BT088 and BT054 cells both produce EVs that carry cytotoxic factors.
Finally, to determine whether the cytotoxic nature of BT088 EVs translated to other cell types, we directly examined BT088 EV effects on BT088 cell growth (i.e. homotypic activity) using a similar tumorsphere assay (Fig. 2B). Notably, BT088 EVs had a similar cytotoxic effect on BT088 cells themselves, reducing tumorsphere size and live cell number after 5 DIV (Fig. 2O-S), suggesting that ODG cells may limit their own growth/survival by EV-mediated autocrine/paracrine effects.
Taken together, these studies confirm that while ODG CM from both BT088 and BT054 cell lines contains pro-proliferative soluble factors, in addition to EV-enclosed factors that induce cell death, which in BT054 cells, overshadows the pro-proliferative effects. These differences in the bioactive nature of the BT054 and BT088 secretomes may help to explain their different growth rates (Fig. 2A) and tumorigenicity9.
ODG EVs induce proliferation followed by apoptosis
The ability of BT088 CM to induce the formation of more and larger neurospheres suggested that pro-proliferative effects could occur first, followed by cell death. To interrogate the timeline of events, we performed live cell imaging using phase area confluence as a surrogate measure of cell number (Fig. 2T). NSCs plated in either BT088 CM or CM-EVs increased in number after 5 DIV at a faster rate than NSCs grown in FM or FM+EV (Fig. 2U). However, after 5 DIV, while NSCs continued to grow exponentially in CM-EV, cell number started to decline in CM (Fig. 2U). There was also a decrease in cell number in FM+EV compared to FM, which supports the notion that EVs carry cytotoxic factors (Fig. 2U).
We further confirmed the cytotoxic nature of BT088 EVs by performing live cell imaging after incorporating a fluorescent cytotoxic dye (‘Cytotox’) into NSC cultures. A decline in cellular health increases cell permeability thereby permitting entry and intercalation of Cytotox into affected cells. After 3 DIV, NSCs exposed to CM and FM+EVs incorporated Cytotox at a higher rate than cells grown in FM or CM-EV, validating the cytotoxic nature of the EVs (Fig. 2V; Fig S1F-Q). Finally, to determine whether BT088 EVs exerted a pro-apoptotic effect on NSCs, AnnexinV/Propidium Iodide (PI) co-labeling of late apoptotic cells was assessed via flow cytometry27 (Fig. S1R). After 3 DIV, NSCs grown in FM included 3.8% AnnexinV+PI+ late apoptotic cells, whereas nearly 52.7% of NSCs grown in FM+EV were double+ apoptotic cells (Fig. S1S). Similarly, NSCs cultured for 3 DIV in BT088 CM included 55.6% AnnexinV+PI+ late apoptotic cells, and this number declined to 35% when grown in CM-EV (Fig. S1S).
In summary, BT088 CM has an initial pro-proliferative effect that is followed a few days later by an increase in EV-mediated cell death, suggesting that ODG EVs induce apoptosis of NSC progeny.
ODG cells secrete bioactive EVs mainly in the 40-200 nm exosome size range
To determine whether the isolated ‘EVs’ tested above were truly vesicular, we characterized their morphology and protein content. First, we used scanning electron microscopy to assess cellular topography, revealing the presence of budding vesicles on the surfaces of both BT054 and BT088 cells (Fig. 3A,B). Next, to assess vesicular size and composition, EVs were isolated from CM using sequential centrifugation12. Using transmission electron microscopy, we demonstrated that sedimented particles isolated from ODG cells had a lipid bilayer-enclosed nanoparticle morphology characteristic of EVs (BT088 EVs shown, Fig. 3C,C’). To calculate the size and number of isolated EVs, we used nanosight tracking analysis (NTA). BT054 and BT088 cells both produced EVs that were largely within the exosome size range (i.e. 40-200 nm 28, Fig. 3D,E, Table S1), but BT088 cells produced twice as many EV particles (Table S1). Finally, as a first assessment of molecular phenotype of ODG EVs, using nanoscale-flow cytometry we revealed that 13.7±2.0% of BT088 EVs expressed CD9, a marker of a subset of EVs29 (Fig. 3G).
To further assess the molecular nature of ODG EVs, we focused on BT088 EVs, as they were more numerous. Using Western blotting, we compared protein content in the crude cell lysate (CL) and in the EV pellet. Compared to CL, EVs were enriched in EV-associated markers30, including Alix, CD9 and Cetp (Fig. 3H). Flotillin1, a common EV marker, was also detected in EVs, but was also present at high levels in CL due to its association with the cell membrane31 (Fig. 3H). Finally, we confirmed the relative purity of the EV preparations by probing for proteins associated with other organelles, including the endoplasmic reticulum (Calreticulin, Calnexin), mitochondria (Vdac), Golgi bodies (GM130) and peroxisomes (Pex5), all of which were detected in the CL, as expected, and at negligible levels in EVs (Fig. 3I).
As the sequential centrifugation method of EV isolation is sedimentation-based, it can also isolate non-vesicular components28. To determine whether non-vesicular material was in the ‘EV’ pellet, we used density gradient ultracentrifugation for size fractionation. BT088-EVs isolated by sequential ultracentrifugation were loaded onto a discontinuous Optiprep™ gradient. After ultracentrifugation, eight fractions were collected from the density gradient, and analyzed by Western blotting and NTA. EV associated markers (Alix, CD9, Cetp) were detected in fractions 4 and 5, with a density of 1.08-1.09 g/cm3, while negative EV markers (Vdac, Calnexin) were absent in these fractions (Fig. 3J). Using NTA, we confirmed that EV particles in layer 4 were predominantly in the exosome size range (Fig. 3F). Of note, Cetp was also detected in low density fractions 1-3 (0.95-1.07 g/cm3), which contain non-vesicular low-density lipoproteins to which Cetp associates32. Thus, the BT088-EV pellet collected by ultracentrifugation includes EVs mainly in the exosome size range but also some microvesicles and non-vesicular material, as previously shown for other cell types28.
EVs deliver molecular content to neighboring cells through membrane fusion, a delivery method not available to non-vesicular material. To determine whether BT088 cells secrete bioactive cargo in EVs that can be taken up by and influence recipient cells, we used a Cre recombinase-based fluorescent reporter assay previously used to report EV-mediated Cre transfer33. BT088 ‘donor’ cells were stably transduced to express Cre recombinase and GFP, while NIH-3T3 ‘recipient’ cells were transduced with a dual BFP-dsRed Cre reporter33. BFP is expressed in recipient cells prior to Cre (mRNA/protein) transfer from donor-to-recipient cells, which undergo Cre mediated excision of a STOP cassette to then allow expression of dsRed. We aggregated BT088-GFP-Cre donor cells with NIH-3T3 BFP-dsRed recipient cells (Fig. 3K). After 3 DIV, we detected RFP+ cells that were GFP−, indicating that they did not arise from cell fusion (Fig. 3L-O).
BT088 cells thus secrete bioactive cargo to recipient cells, and from our in vitro assays, we suggest that this includes EVs and other soluble factors.
Proteomic profiling reveals ODG cells secrete distinct vesicular cargo
To characterize the potential bioactive content of ODG EVs, we carried out proteomic analyses of BT088 and BT054 vesiculomes by LC-MS/MS (Fig. 4A). Given the known differences in biologic behavior of the two cell lines, it was unsurprising that the protein content differed between the two EV pools, with 390 proteins detected in BT088 EVs and 186 proteins in BT054 EVs, of which only 72 proteins were common between the two vesiculomes (minimum 2 out of 3 individual replicates; Fig. 4A’). Among the shared proteins, both BT088 and BT054 EVs contained commonly associated EV proteins (e.g. Alix, CD63; Table S328, 34), validating the vesicular nature of our preparations.
BT088 and BT054 vesiculomes were enriched in proteins associated with several biological processes, including metabolic, developmental, immune system, growth and cell death (Fig. 4B). In both BT088 and BT054 vesiculomes, several identified proteins were assigned to cell death pathways, some of which were higher in BT088 EVs (e.g. RHOA, RPL11; Fig. 4C,D), others elevated in the BT054 vesiculome (e.g. HTRA1 (Fig. 4E), and some common to EVs from both cell types (e.g. CLU; Fig. 4F). Notably, of the proteins associated with proliferation, such as PKM, HSPB1, HSP90AA1, and HSP90AB1 (Fig. 4G-J), all were detected at higher levels in the BT088 versus BT054 vesiculome, in keeping with differences in the growth rates of these two tumor cell lines9.
While our data suggested that cytotoxic signals largely overshadow pro-proliferative signals in ODG-derived EVs, BT088 and BT054 tumor cells continue to grow and thrive. To understand how ODG cell proliferation is supported, we queried the proteome for growth regulators, as proteins are loaded into EVs non-specifically, and serve as an informative readout of cellular state28. Two proteins detected in both BT088 and BT054 EVs stood out - SRI (Sorcin) and MFGE8 (Lactadherin) (Fig. 4K,L). SRI expression is elevated in several cancers, including breast, hepatocellular, and gastric35, and it is required to maintain VEGF expression, which is involved in angiogenesis, tumor invasion, and metastasis35, 36. Similarly, MFGE8 promotes VEGF- dependent neovascularization in endothelial cells37. In line with the viewpoint that EVs are non-specific carriers of cellular protein, and that EV content implies signaling status 28, we inferred that VEGF signaling may be activated in these two ODG cell lines. To assess the importance of VEGF signaling in supporting ODG tumor growth, we treated BT088 cells with Foretinib, a VEGF-receptor inhibitor38, at five different concentrations (50 nM-500 nM) and cell growth was assessed over five days using live cell imaging. Strikingly, in all conditions, BT088 cell growth resembled BT054 growth (Fig. 2A), with increase in BT088-cell doubling times (from 4-9 fold) at all concentrations, compared to BT088 cells grown in DMSO (Fig. 4M). Foretinib treatment of BT088 cells also increased cytotoxicity at all assessed doses and inhibited growth of BT088 cells (Fig. 4N).
BT088 and BT054 vesiculomes are thus distinct from each other, revealing the heterogeneity of ODG intracellular signaling and the potential pathways that may trigger cell death in neighboring cells.
Lower SMPD3 expression is associated with poor prognosis in oligodendroglioma patients
Given that BT088 and BT054 EVs both had cytotoxic effects, we queried whether EV generation was itself associated with ODG biologic behavior. We analyzed data from the cancer genome atlas (TCGA) database to compare expression of genes involved in ESCRT-dependent (TSG101, STAM1) 12, and ESCRT-independent (SMPD3)13 exosome synthesis in a lower grade glioma patient cohort, correlating gene expression with patient survival. Strikingly, lower TSG101, STAM1, and SMPD3 expression levels correlated with shorter survival times compared to patients with higher expression of these genes (Fig. 5A-C; p<0.0001, Table S1), consistent with the notion that exosome synthesis may play a role in limiting lower grade glioma growth. Given the better prognosis of lower grade glioma patients with high SMPD3 expression, we focused further studies on SMPD3, which encodes for nSMase2, the major sphingomyelinase in the brain39.
We next used the TCGA data to compare SMPD3 expression levels in normal brain versus lower- and higher-grade gliomas. SMPD3 expression levels were highest in normal brain, followed by lower-grade glioma and higher-grade GBM (Fig. 5D). We further dissected the lower-grade glioma dataset into IDH-mutant astrocytoma and IDH-mutant ODG and found that astrocytoma patients, which have a shorter survival time (median survival=9-13 years6, 36), presented with lower SMPD3 transcript levels compared to ODG patients, which have a longer survival time (median survival= 12-14 years40, 41; Fig. 5E; p<0.0001) – suggesting that higher SMPD3 levels may reduce tumor growth. Furthermore, within each disease type, patients with lower SMPD3 expression in their tumors had shorter survival times compared to those with higher SMPD3 expression (Fig. 5F, p=0.0058 for astrocytoma; and Fig. 5G, p <0.0001 for ODG;Table S1).
These data, together with our findings from our EV cell culture experiments (Fig. 2), suggest that higher SMPD3 expression levels may limit ODG (and potentially also IDH-mutant astrocytoma) growth to a certain degree by triggering an increase in cytotoxic exosome production.
SMPD3 inhibits ODG proliferation in vitro
To assess whether SMPD3 might play a role in regulating ODG growth, we first confirmed that the encoded protein, nSMase2, was expressed in BT088 and BT054 ODG cells, demonstrating co-immunolabeling with Sox10, a marker of glial-like tumor cells42 (Fig. S2A,B). Previous studies indicated that SMPD3 overexpression enhances EV production43, which our data suggested should inhibit ODG cell growth. To test this prediction, we used a doxycycline-inducible lentiviral system to generate stable SMPD3-GFP and control-GFP BT088 cell lines (Fig. 6A,A’). We confirmed that 72 hrs post-doxycycline exposure, nSMase2 expression was induced (Fig. 6B,B’). We also used nanoscale flow cytometry to confirm that SMPD3 overexpression increased the release of CD9+ EVs from SMPD3-GFP cells, revealing a 2.3-2.9 fold increase compared to control cells, and SMPD3-GFP cells not treated with doxycycline (Fig. 6C,C’, Table S1).
We next monitored the growth of control-GFP and SMPD3-GFP BT088 cells using live cell fluorescent imaging of adherent cells (Fig. 6D-G’). As a proxy measure of cell growth, we monitored the cumulative area covered by GFP+ cells normalized to seeding day (day 0) (Fig. 6H). Control-GFP cells and SMPD3-GFP cells not exposed to doxycycline grew exponentially, with doubling times in the range of 78-84 hrs. In contrast, doxycycline dosed SMPD3-GFP cells failed to proliferate in the first 1.5 days, after which there was a decline in the cumulative GFP area, suggestive of cell death (Fig. 6H, Table S1). We also monitored the growth of Control-GFP and SMPD3-GFP BT088 cells grown in 3D suspension cultures, which can replicate more of the cell-cell interactions observed in vivo, and found that SMPD3 overexpression similarly inhibits tumor cell growth (Fig. S2C-G). Finally, to assess the potential cytotoxicity of elevated SMPD3 expression levels, we incorporated the cytotoxic dye ‘Cytotox’ in the culture media, and assessed the cumulative area covered by Cytotox+ cells. Doxycycline dosed SMPD3-GFP cells increased Cytotox accumulation compared to all control conditions (Fig. 6I-M, Table S1).
Taken together with the cytotoxic nature of BT088 EVs on NSC and BT088 cell growth, we conclude that elevated SMPD3 expression likely inhibits BT088 cell growth through the production of EVs, although we cannot rule out the potential for SMPD3 having additional cell autonomous effects.
Knockdown of SMPD3 promotes proliferation of ODG cells in vitro
We next asked the converse question, which is whether reduced SMPD3 expression levels might promote ODG cell growth. We used an shRNA approach to stably knockdown SMPD3 expression in BT088 cells. Lentiviral constructs that expressed GFP and one of four shSMPD3 variants (A-D) or an shScrambled (shScr) control sequence were transduced into BT088 cells to create five GFP-tagged BT088 cell lines (Fig. 7A). To assess the efficacy of shSMPD3 knockdown, we performed Western blotting, revealing a >50% decrease in nSMase2 levels for shSMPD3 B-D variants, but not variant A (Fig. 7B,B’). shSMPD3 variants B-D also all reduced the number of secreted CD9+ EV particles compared to shScr control (Fig. 7C,C’, Table S1). In addition, isolated EV pellets from the CM (from shSMPD3 B-D variants) had lower levels of Alix, an EV marker (Fig. S3A).
To assess the effects of SMPD3 knockdown on BT088 cell growth, we used live cell fluorescent imaging of adherent cell cultures (Fig. 7D-H). Compared to shScr control cells, BT088 cells expressing shSMPD3 variants B-D cells grew more rapidly after 5 DIV, with doubling times of ~160-192 hrs compared to ~800 hrs for shScr control (Fig. 7H, Table S1). The growth stimulatory properties of SMPD3 knockdown was similarly observed when BT088 cells were grown in 3D suspension cultures for 9 DIV (Fig. S3B-F). To further validate these data, we also performed manual tumorsphere counts and diameter measurements of BT088 cells expressing control shScr versus shSMPD3-B (Fig. S3G-K). After 10 DIV, SMPD3 knockdown increased tumorsphere number (Fig. S3I), tumorsphere size (Fig. S3J), and live cell number (Fig. S3K), replicating the live cell imaging data.
The increase in cell number observed after shSMPD3 knockdown could be due to an increase in proliferation and/or a decrease in cell death. Proliferation must have increased to produce more cells after shSMPD3 knockdown and, accordingly, all BT088 cell lines incorporated BrdU, indicative of active DNA synthesis in S-phase of the cell cycle (Fig. S3L-O). However, to determine whether SMPD3 knockdown also reduced normal levels of cell death in BT088 cells, we assessed ‘Cytotox’ dye incorporation in BT088 cells expressing shScr versus shSMPD3 knockdown constructs, revealing more dye incorporation in the first days of culture by the shSMPD3 lines (Fig. 7I-M). Thus, knockdown of SMPD3 expression in BT088 cells not only enhances cell proliferation, but also reduces cell death rates to increase tumor cell number.
As a final independent measure of the effects of SMPD3 knockdown on BT088 cell growth, we used a pharmacological approach, treating BT088 cells with GW4869, a competitive inhibitor of phosphatidylserine, a phospholipid that binds nSMase2 and is required for enzyme activation44, 45. Exposure of BT088 cells to 1μM GW4869 for 48h reduced CD9+ EV production by 2-fold compared to DMSO control cells (Fig. S3P,Q). Treatment of BT088 cells with GW4869 also increased cell proliferation using live cell imaging but only at later stages (Fig. S3R-T). To confirm that the knockdown of SMPD3 was stimulatory for BT088 growth, we also used a higher concentration of GW4869 (18 μM), which increased tumorsphere number (Fig. S3U-W) and size (Fig. S3X) after 5 DIV.
Taken together, we have genetic and pharmacological evidence that lowering SMPD3 expression levels increases ODG cell proliferation and survival, in keeping with the reduced survival times associated with lower SMPD3 expression in ODG patients.
SMPD3 knockdown facilitates ODG growth in vivo
BT088 cells engineered to express GFP form tumors by 6 months after orthotopic xenograft into the cerebral cortices of NOD scid Gamma (NSG) mice (Fig. 8A). In the xenografts, the neoplastic (human) cells can be identified by the expression of human nuclear antigen (HNA) whereas mouse cells are negative for HNA (Fig. S4A,B). As expected, the main tumor masses were comprised predominantly of densely packed HNA+ tumor cells, while a smaller number of HNA+ tumor cells infiltrated at the periphery (Fig. S4A). We further confirmed that that engrafted HNA+ BT088 cells continued to express nSMase2 (Fig. S4A,B).
We then used this model to further evaluate the impact of SMPD3 expression on tumor growth in the context of the complex in vivo environment. We made stable SMPD3 knockdown (KD) and shScr control BT088 cell lines (Fig. S4C-G). shScr and SMPD3-KD cells were orthotopically xenografted into NSG mice (N=8 per cohort), and animals were monitored over six months (Fig. 8A). shScr and SMPD3-KD BT088 survival curves were significantly different from one another (Fig. 8B; Mantel-Cox log rank test p=0.0074, Table S1). At endpoint, all mice were confirmed to have tumors in sections consisting of hypercellular masses that disrupted the normal cortical architecture. By day 154, however, all mice xenografted with SMPD3-KD BT088 cells had been sacrificed at the humane endpoint after showing terminal symptoms, whereas two mice xenografted with shScr BT088 cells were still alive after 6 months (181 days), our experimental endpoint.
To further characterize the tumor masses associated with BT088 shScr and SMPD3-KD cells, we co-immunostained sections with HNA along with the oligodendroglial lineage marker Olig2, the astrocytic marker GFAP, the vascular marker isolectin, and the proliferative cell marker Ki-67 (Fig. 8C-J). Both BT088 shScr and SMPD3-KD cells co-expressed HNA and Olig2 (Fig. 8C,D), as expected, and many HNA+ tumor cells were Ki-67+ in both tumor types (Fig. 8E,F). However, the two tumor types differed as SMPD3-KD tumors appeared to be more highly vascularized than control shScr tumors, with dense isolectin staining (Fig. 8G,H), phenocopying the increase vascular endothelial proliferation that is associated with higher-grade gliomas in patients46. In addition, GFAP+ cells bearing morphologic features of reactive astrocytes (hypertrophy of soma and processes, resulting in coarser and enlarged processes, and more abundant cytoplasm) were more abundant in SMPD3-KD tumor margins (Fig. 8I,J), suggestive of enhanced reactive changes 47.
SMPD3 knockdown in BT088 cells thus results in the formation of more biologically aggressive tumors that display some phenotypic features of higher-grade III ODG48.
SMPD3 knockdown increases ODG invasiveness and growth in human cerebral organoids
Tumor xenografting in immunocompromised mice is a powerful method to examine how genetic or pharmacological manipulations impact tumor burden, but species-specific features of tumor growth cannot be examined. To test whether SMPD3-KD BT088 cells grew faster in a human context, we used cerebral organoids (COs) generated from human embryonic stem cells49, 50. Human COs have now been used in a few studies for brain tumor modelling and provide an excellent readout of tumor infiltration and growth progression51. To generate 30-day old COs, we followed a modified Lancaster protocol (Fig. S4H, 9A)52. After 30 DIV, COs were either cultured alone (Fig. 9B-B’’’) or together with BT088 cells for an additional 7 DIV (Fig. 9A,C-E’’’). To assess SMPD3 knockdown effects, we co-cultured 30 DIV COs with stable BT088 cell lines engineered to express GFP and shScr (Fig. 9A,C-C’’’), shSMPD3-B (Fig. 9A,D-D’’’) or shSMPD3-D (Fig. 9A,E-E’’’) (knockdown validated; Fig. 7).
After 7 DIV, we first examined the neural identity of co-cultured COs by performing SOX2 immunostaining, which labeled neural rosettes in all COs (Fig. 9B’-E’). Notably, SOX2+ rosettes tended to concentrate in the CO periphery, a positional preference that we validated by counting SOX2+ cells in seven zones from periphery to the core (zone width =50 μm; Fig. 9A’,F). The negative slope of the lines of best-fit suggested that there was a biased distribution of neural rosettes in the CO periphery, and as the slopes were similar for shScr, shSMPD3-B and -D co-cultures, the addition of ODG cells did not alter neural cell organization (Fig. 9G; Table S1). Next, we co-stained the COs with turbo-GFP (tGFP) to label infiltrating tumor cells. tGFP+ tumor cells were detected in all CO co-cultures (N=3 for each condition; Fig. 9B’’-E’’). Strikingly, for shSMPD3-B and shSMPD3-D BT088 lines, more tGFP+ cells were detected in the periphery, where SOX2+ neural rosettes localize, compared to shScr controls (Fig. 9H), as revealed by the significant difference in slopes of the lines of best-fit (Fig. 9I; Table S1). Moreover, SMPD3 KD BT088 lines had more tGFP+ cells in the CO compared to shScr, indicative of a growth advantage (Fig. 9J, Table S1).
Thus, the CO-tumor co-culture system recapitulated the more proliferative phenotype exhibited by shSMPD3 BT088 cells in vitro and in mouse xenografts.
DISCUSSION
ODG tumors are slow growing, but the molecular mechanisms underlying their indolent growth is poorly understood. Here we report that ODG cells produce cytotoxic EVs that could allow them to interact and communicate with other tumor cells and non-neoplastic brain cells in the glioma microenvironment. Furthermore, we identify SMPD3 as a critical regulator of EV biogenesis in ODG tumors, revealing that lower SMPD3 expression levels are associated with a less favourable prognostic outcome both in patients and in murine xenografts (Fig. 9K). Prior studies had demonstrated that EVs isolated from G26/24 ODG cells exerted a cytotoxic effect on neurons and astrocytes in vitro, acting via Fas ligand (Fas-L) and tumor necrosis factor-related apoptosis-inducing ligand (TRAIL), respectively20, 21. However, the in vivo relevance and underlying biogenetic pathways were not addressed. Moreover, it is important to acknowledge that G26/24 is a mouse glioma cell line that was originally classified as ODG based on histological resemblance53. G26/24 cells have not been shown to model the human disease genetically, and their relevance to human ODG remains unclear. Furthermore, while TRAIL and Fas-L mediate the cytotoxic effects of G26/24-EVs20, 21, these factors were not present in BT088- and BT054-derived EVs, demonstrating key differences between mouse and human ODG EVs, while also highlighting the heterogenous nature of the ODG vesiculomes.
In aggregate, our data suggests that the secretion of cytotoxic EVs by ODG tumor cells may contribute to the slow growing nature of these tumors. Notably, SMPD3 also exhibits growth inhibitory functions in other cancers, mediated either by its role in EV biogenesis or by its role in ceramide production54. Early studies using a rat C6 glioma cell line found that Smpd3 produces ceramide, which promotes apoptosis by activating PKCδ signaling55. Homozygous deletion of Smpd3 also induces osteosarcoma in a rodent model56. In line with these studies, we report here that SMPD3 inhibits ODG cell growth, possibly through paracrine and autocrine effects. Along with SMPD3, we also found a correlation between high expression levels of ESCRT-dependent exosome biogenesis genes57 (TSG101, STAM1) and improved patient survival in low-grade glioma patients, suggesting that EV biogenesis via several pathways may play an important role in these tumors. It is important to note, however, that increasing EV biogenesis is not always favorable, and depends on the tumor type. For instance, knockdown of Rab27a/b, which also blocks EV secretion, was shown to inhibit tumor growth in other brain tumor models58, 59. Rab27a/b knockdown in astrocytes, which reduces EV secretion, blocks the metastasis of breast cancer cells to the brain58. Moreover, knockdown of Rab27a/b in an astrocyte-derived glioma cell line blocked glioma growth in vivo in mouse xenografts 59. A likely cause of these different effects is that EV cargo is well known to differ based on the cell of origin60. In high-grade GBM tumors secrete EVs that contain many growth factors (e.g. EGFRvIII) that could promote glioma growth and progression61, 62.
While we showed that ODG EVs have cytotoxic effects, it is likely not the EVs themselves that are cytotoxic, but rather, their enclosed cargo. Even within two different ODG lines, we found distinct differences between the vesiculomes that may help to explain differences in the growth of BT088 and BT054 cells. The identification of SRI (Sorcin) and MFGE8 (Lactadherin) in ODG cell EVs is of interest, as it suggested that VEGF signaling is increased, and indeed, we could block BT088 cell growth with Foretinib in vitro. VEGF signaling is typically linked to shaping the tumor vasculature, including in GBM 63, but our data is suggestive of a potential autocrine role for VEGF in supporting ODG growth, as has been suggested in GBM 64 and hepatocellular carcinomas 55. Future studies in which ODG cell xenografted animals are treated with Foretinib would help assess the potential of this drug as a therapeutic, since in vitro assays are not always translated in an in vivo setting, as we have highlighted. Furthermore, the analysis of ODG EV-associated factors such as CLU, previously implicated in invasion65, may help to understand how these tumor cells infiltrate the brain in vivo.
In summary, the mechanisms underlying neural-ODG cell interaction are complex with the tumor microenvironment playing a crucial role. Our studies indicate that ODG cells release EVs that serve as messengers of predominantly cytotoxic cargo to neighbouring cells. We further report that SMPD3 expression levels correlate with ODG survival, and that SMPD3 acts in part through its ability to regulate EV generation and secretion. Given that ODG EVs in the microenvironment contribute towards a slow growth phenotype, drugs that increase EV production could have therapeutic potential. While most high-throughput screens have focused on identifying small molecules that block EV secretion, which could be useful for GBM tumors66, there have also been recent screens for small molecules that induce EV secretion67. Interestingly, N-methyldopamine and norepinephrine activate nSMase2 to increase EV production in mesenchymal stem cells (MSCs) without increasing cell number, a strategy that is being developed to enhance the regenerative potential of MSC EVs67. Testing whether these drugs increase ODG EV production and the associated cytotoxicity would be of interest in the future. However, one must keep in mind that it is also possible (likely) that the interaction of ODG EVs with non-neoplastic cells in the tumor niche may be reciprocated by EV secretion by brain cells that are not transformed, which together would help to create a unique microenvironment for every tumor. Careful analyses of drug efficacy would thus require 3D culture systems that mimic cell-cell interactions, such as our human-CO/tumor cell co-culture assay, or xenografts into immunocompromised mice, which remains the gold standard for pre-clinical data.
METHODS
Patient-derived tumor tissues and cells and study approval
IDH mutant ODG and astrocytoma patient biopsies were obtained from the pathology archives at the Calgary Laboratory services and Clark Smith Brain Tumor Bank at the University of Calgary9. Samples were formalin fixed and embedded in paraffin. Approval for use was obtained from Calgary Laboratory Services and the Calgary Health Region Ethics Board (University of Calgary Conjoint Health Research Ethics Board to JAC (HREB #2875 and #24993). BT088 and BT054 cell lines were collected under approved protocols from the Health Research Ethics Board of Alberta to JAC (HREBA.CC-16-0762 and HREBBA.CC-16-0154). Culture of BT088 and BT054 cells was approved by the Sunnybrook Research Ethics Board (REB) to CS (PIN: 301-2017). Culture of human ESCs received approval from the Canadian Institutes of Health Research (CIHR) Stem Cell Oversight Committee (SCOC) to CS and was approved by the Sunnybrook REB (PIN: 264-2018).
The Cancer Genome Atlas (TCGA) survey
Kaplan Meier plot was generated from lowest and highest quartiles of all low-grade glioma patients for all assessed genes. Normal brain cortex (GTEX), GBM, low-grade glioma, IDH mutant astrocytoma, and IDH mutant ODG (TCGA) datasets were downloaded from UCSC’s Xena Browser (https://xenabrowser.net/) and analyzed to compare SMPD3 expression levels, which was correlated with overall patient survival.
Animals
CD1 outbred mice used for neurosphere assays were purchased from Charles River Laboratories (Senneville, QC). Embryos were staged using the morning of the vaginal plug as embryonic day (E) 0.5. Sex was not considered due to the difficulty in assigning sex at embryonic stages. Animal procedures were approved by the Sunnybrook Research Institute Animal Care Committee (20-606) in compliance with the Guidelines of the Canadian Council of Animal Care.
BT088 and BT054 cell culture
BT088 and BT054 cells9 were grown in Human Neurocult proliferation media (Stem Cell Tech; # 05751) containing human epidermal growth factor (hEGF, 20 ng/ml, Peprotech; AF-100-15), human fibroblast growth factor 2 (hFGF2, 20 ng/ml, Wisent; # 511-126-QU), heparin (2 μg/ml; Stem Cell Tech, #07980), and Antibiotic-Antimycotic solution (0.1%; Wisent; # 05751). Tumorspheres were dissociated and passaged using Accutase (Stem Cell Tech; # 07920). Neurosphere media was DMEM (Wisent; #319-005-CL):F12 (ThermoFisher Scientific; #31765-035) (3:1), with hFGF2 (20ng/ml), hEGF (20ng/ml), B27 supplement (2%; ThermoFisher Scientific; #17504044), Antibiotic-Antimycotic solution (0.1%), Cyclopamine (0.5 μM; Sigma; #C4116), and Heparin (2 μg/ml). BT088 cells were cultured either as adherent cells on Poly-D-Lysine:Laminin coated tissue culture plates, or in suspension on non-coated flasks/plates. For BT088 tumorsphere assays, dissociated BT088 cells were seeded at 8000 cells/ml in BT088 media (FM), or in BT088 FM+EVs. BT088 cells grew undisturbed for 5 DIV, and tumorspheres were then imaged.
Small molecule inhibitors
Small molecules were added as follows: GW4869 (stock concentration=1 mM): FM was supplemented with 1 or 18 μl/ml DMSO (control) or 1 μl/ml of 1 μM or 18 μl of 18 μM GW4869 (SigmaAldrich). Foretinib (stock concentration=0.1 mM): FM was supplemented with 5 μl/ml DMSO (control) or 0.5, 1, 2, 4 or 5 μl/ml Foretinib (GSK1363089) at 50, 100, 200, 400, or 500 nM.
Mouse NSC isolation and culture
Dorsal telencephalons (cortices) from E13.5 embryos were dissected and dissociated in 0.125% trypsin (Wisent) at 37°C for 10 mins. Dissociated cells were seeded at 8000 cells/ml in neurosphere media (FM), or in BT088 or BT054 CM, CM-EV, or FM+EVs. NSCs grew undisturbed for 7 DIV, and neurospheres were then imaged.
AnnexinV-PI Apoptosis assay
NSCs were cultured in FM, BT088 CM, CM-EV, and FM+EV for 3 DIV. Cells were stained with FITC-labeled Annexin V and PI using the FITC Annexin V Apoptosis Detection Kit with PI as per the manufacturer’s instructions (BD Biosciences; 640914) for 20 min at 25°C and analyzed by flow cytometry.
Conditioned media
1×106 BT088 or BT054 cells were seeded in 11 ml of fresh media and CM was collected after 24 hrs. CM was centrifuged at 300 x g for 5 min to remove cells, followed by 2000 x g for 10 min to remove cellular debris and 10,000 x g for 30 min to remove protein aggregates and smaller debris. CM was then sequentially ultra-centrifuged at 100,000 x g at 4 ° C for 2 hrs in a Beckman Coulter Optima L-100 XP Ultracentrifuge using an SW41-Ti rotor and polycarbonate centrifugation tubes (Beckman Coulter; #331372). The EV pellet, and supernatant (CM-EV) were used as indicated. The EV pellet was rinsed with Phosphate Buffered Saline (PBS) (ThermoFisher Scientific; # 14190144) and centrifuged at 100,000 x g, 4 ° C for 1 hr and resuspended in 50 μl PBS prior to use.
Incucyte live cell imaging
Cell growth and death were monitored using an Incucyte S3 Live cell imaging system (Essen BioScience). BT088 and BT054 cell growth rates were monitored using NucLight Rapid Red Reagent for nuclear staining of the cells (as per manufacturer’s instructions; IncuCyte; #4717). Cells were suspended in media supplemented with 2 μl/ml of the reagent prior to cell seeding. Cell growth was quantified by monitoring area covered by RFP+ objects. All other cell growth studies monitored total phase area confluence or area covered by GFP+ cells. For cell death assays, the media was supplemented with 0.25 μl/ml of Cytotox dye/well (as per manufacturer’s instructions; red #4632 and green #4633) prior to cell seeding. Phase contrast and Red/Green fluorescent imaging was carried out at designated intervals (cell growth studies: every 12/24 hrs; cell death studies: every 4/12 hrs), and at 10x/20x magnification. A minimum of 9 images were taken per well at each time point. Quantification of cell proliferation and cell death was performed using the analyser algorithm built in the Incucyte application. Mean values of total phase area (normalized to day 0) ratio (U) and Cytotox+ object area (normalized to day 0) ratio (V) were plotted, comparing between days 0 to 7.
CO-BT088 co-cultures
Feeder-free H1 hESCs (WiCell) were cultured on Matrigel in TeSR™-E8™ kit for hESC/hiPSC maintenance (StemCell Tech; #05990). hESCs were used to generate COs using media included in the STEMdiff Cerebral Organoid Kit (StemCell Tech; #08570) and STEMdiff Cerebral Organoid Maturation Kit (StemCell Tech; #08571), with some modifications. Briefly, hESCs were plated in 96-well round-bottom ultra-low attachment plates at 9,000 cells/well in embryoid body (EB) seeding medium. Dual SMAD inhibitors (2μM Dorsomorphin; StemCell Tech; #72102, and 2 μM A83-01; StemCell Tech; #72022) until day 5. Newly formed EBs were transferred to 24-well plates containing StemCell Tech CO induction medium. On day 9, EBs with optically translucent edges were embedded in matrigel and deposited into 6-well ultra-low adherent plate with StemCell Tech expansion medium. From day 5 to day 13, media was supplemented with 1 μM CHIR-99021 (StemCell Tech; #72052) and 1 μM SB-431542 (StemCell Tech; #72232) to support formation of well-defined, polarized neuroepithelia-like structures. On day 13, embedded EBs exhibiting expanded neuroepithelia as budding surfaces were transferred to a 12-well spinning bioreactor (Spin Omega52) containing maturation medium in a 37°C incubator. For BT088 co-culture, on day 30, COs were individually transferred to a 24-well plate containing Neurocult NS-A proliferation media (Catalog # 05751, StemCell Tech) with freshly added hFGF2 (20 ng/ml), hEGF (20ng/ml), and heparin (2μg/mL). Subsequently, 10,000 GFP+ BT088 cells were added to each well. Plates were incubated for 24 hrs without agitation and on the next day tumor-bearing COs were washed with PBS once and maintained in maturation media on an orbital shaker at 37°C for 7 more days. On day 8, COs were fixed in 4% paraformaldehyde (PFA) for 45 min, transferred into 30% sucrose overnight, snap frozen in OCT for cryosectioning.
Pellet assay
BT088 cells expressing GFP and Cre were mixed with NIH-3T3 cells transfected with BFP-loxP-dsRed in a 5:1 ratio. Cells were centrifuged, and the pellet was placed on a cell culture membrane. The membrane was floated on DMEM media in a 6-chamber dish. The cells were incubated at 37°C for 3DIV after which the membrane was embedded in a cryopreservative, OCT, and frozen gradually on dry ice. 10 μm thick sections were obtained by sectioning the OCT block.
Density gradient Ultracentrifugation
OptiPrep™ (Iodixanol 60% stock solution; StemCellTech; #07820) was diluted with a homogenization solution (0.25M sucrose in 10mM Tris HCl pH 7.5) to generate a discontinuous density gradient of 40% (2.5ml), 20% (2.5ml), 10% (2.5 ml), and 5% (2 ml). The solutions were carefully pipetted in an ultracentrifuge tube and left undisturbed for >1hr. EVs (filtered through a 0.2 μm filter and resuspended in 500 μl PBS) isolated from BT088 CM by sequential centrifugation after the first 100,000 x g spin were loaded onto the gradient. Samples were centrifuged at 100,000 x g for 18 hrs. Post centrifugation, 1 mL fractions were pipetted out carefully from the top. Fractions were mixed with PBS and centrifuged at 100,000 x g for 4 hrs. EV pellets were resuspended in 50 μl PBS.
Scanning Electron Microscopy
BT088 and BT054 cells were cultured on 13 mm coverslips (EM Biosciences) precoated with Poly-O-Lysine – Laminin in a 24 well plate. The samples were then fixed (2% glutaraldehyde in 0.1M sodium cacodylate buffer, pH 7.3; for >2 hrs), rinsed and dehydrated. Samples were mounted on stubs, gold sputter-coated, and imaged with FEI/Philips XL30 scanning electron microscope at 15 kV. Imaging was performed in the Nanoscale Biomedical Imaging Facility, SickKids Research Institute.
Transmission Electron Microscopy
Before grids were prepared, carbon-coated Cu400 TEM grids were glow discharged for 30 s (Pelco EasiGlow, Ted Pella Inc.). Then 4 μL of BT088 exosome solution was applied to the grid for 60 seconds before wicking away excess solution. The grid was washed three times with 4 μL of distilled water. The grid was stained with 4 μL of 2% uranyl acetate solution for 30 seconds. Excess uranyl acetate solution was wicked away. The grids were air-dried. Imaging was performed on a Thermo Fisher Scientific Talos L120C TEM operated at 120 kV using a LaB6 filament. Imaging was performed in the Microscopy Imaging Laboratory, University of Toronto.
Nanosight tracking analysis (NTA) and Nano-flow cytometry
NTA was performed using the Malvern NanoSight NS300 at the Structural & Biophysical Core Facility, University of Toronto. EV pellets were collected by sequential centrifugation, resuspended in 200 μl PBS, diluted 1:50 and passed through the Nanosight chamber. NTA data acquisition settings were as follows: camera level 13, acquisition time 3 × 30 s with detection threshold 12. Data was analyzed with the NTA 3.2 Dev Build 3.2.16 software. For nanoscale flow cytometry, EVs (1 μl in 18 μl sterile water) were incubated with CD9 antibody (1 μl; Santa Cruz; #sc9148) for 30 mins at room temperature. Post incubation, EVs were stained with Alexa Fluor 647 Far red secondary antibody (1 μl of 1:20 antibody solution; Invitrogen) for 20 mins. Stained EVs were diluted in 500 μl sterile water and quantified on the Nanoscale Flow Cytometer (Apogee Flow Systems Inc). Representative scatterplot of BT088 EVs plotted for 638-Red (detecting Alexa 647 bound CD9+ particles) and Long angle light scatter (LALS; for size distribution) EVs were defined as size events greater than 100 nm.
Molecular Cloning
SMPD3 Gain of function
Conditional overexpression of SMPD3 was achieved with a doxycycline inducible lentiviral vector pCW57-MCS1-P2A-MCS2 (GFP) (Addgene plasmid # 80924). We cloned SMPD3 (SMPD3 Human Tagged ORF Clone, Origene Cat#: RG218441, RefSeq-NM_018667.2) and, after the P2A site, Luc2 (pcDNA3.1(+)/Luc2=tdT (Addgene plasmid # 32904)) into this vector.
SMPD3 Loss of function
For all in vitro SMPD3 knockdown experiments, we used NSMase2 (SMPD3) Human shRNA Plasmid Kit (Origene, Locus ID 55512), which included what we termed shSMPD3 variants A-D and shScr. For xenograft experiments, we used a piggyBac shRNA (GFP) construct (SB #PBSI505A-1) as the backbone. The vector backbone was linearized with BamH1 and EcoR1 and the following annealed oligonucleotides were cloned into the site:
shSmpd3:5’
pGATCCCCCTCATCTTCCCATGTTACTTCAAGAGAGTAACATGGGAAGAT GAGGGACGCGTG3’(sense) and 5’pAATTCACGCGTCCCTCATCTTCCCATGTTACTC TCTTGA AGTAACATGGGAAGATGAGGGG 3’ (antisense); and for shScrambled: 5’pGATCCATTCACTTATCCGCCTCTCCTTCAAGAGAGGAGAGGCGGATAAGTGAATC TCGAGG3’(sense), and 5’pGAATTCCTCGAGATTCACTTATCCGCCTCTCCTCTCTTGAA GGAGAGGCGGATAAGTGAATG-3’ (antisense). The shSmpd3 construct targeted mouse Smpd3 sequence, such that the mouse shRNA target sequence has a 3 bp mismatch upon alignment with human SMPD3 sequence, but it effectively knocked down human SMPD3 (Fig. S4A).
Transduction and transfection
SMPD3 Gain of function
To generate lentiviral particles for SMPD3-P2A-Luc2 GFP and Luc2 GFP the lentiviral vector was packaged in LentiX HEK293T cells using the packaging plasmids psPAX2 (Addgene Plasmid #12260) and pMD2.G (Addgene Plasmid #12259). The resulting lentiviral particles were concentrated by ultra-centrifugation and used to transduce BT088 cells. To induce SMPD3-Luc2 or Luc2 only (control) expression, transduced GFP+ cells were treated with 2 μg/mL doxycycline hyclate (D9891, Sigma).
SMPD3 Loss of function
BT088 cells were transduced with 4 SMPD3 human shRNA lentiviral particles (A,B,C,D) and Lenti shRNA Scramble control particles (pGFP-c-shLenti; TL301492V; Origene). Transduced GFP+ cells (shSMPD3-GFP variants B,D and shScrambled-GFP) used in CO co-culture studies were sorted using the BD FACS ARIA III. GFP+ cells were selected and collected in Neurocult proliferation media and plated. For in vitro tumorsphere assays and generating BT088 cells for xenograft studies, BT088 cells were nucleofected with each shSmpd3/shScrambled vector mixed with Super PiggyBac Transposase expression vector (SBI, Cat#PB210PA-1) in 1: 3, PiggyBac: transposase ratio. 2×106 dissociated BT088 cells were suspended in 20 μl of P3 reagent with the DNA mix (final concentration=12 μg). Nucleofection was performed using the 4D nucleofector (Lonza) in nucleofector strips using the program CZ167. For shSmpd3 knockdown experiments, HEK 293 LentiX cells were transfected with Super PiggyBac Transposase expression vector, hSMPD3-GFP construct, and shScr/shSMPD3 (internal control)/ or mouse shSmpd3-GFP construct using Lipofectamine 3000 (ThermoFisher).
Tumor xenografts
shSmpd3-GFP and shScr-GFP BT088 were injected into the right cerebral hemispheres of 8-10-week old female NOD scid Gamma mice (NSG; NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ; N=8) (Princess Margaret animal breeding facility). Coordinates for implantation were AP-1.0, ML 2.0, and DV 3.0. The xenografted cells developed into tumors, for which the mice were monitored daily. Upon development of terminal symptoms, mice were sacrificed at relative end points. Two control mice did not show any terminal symptoms post engraftment and were humanely sacrificed after 180 days.
Western blotting
Cell or EV pellets were lysed in lysis buffer with protease (1X protease inhibitor complete, 1 mM phenylmethylsulfonyl fluoride) and phosphatase (50 mM NaF, 1 mM NaOV) inhibitors. 3-10 μg of lysate was run on 10% SDS-PAGE gels for Western blot analysis. Primary antibodies included: Flotillin1 (Cell signalling; #3253), CD9 (Santa Crutz; #sc9148), nSMase2 (Abcam; #ab85017), Cetp (Abcam; #ab2726), Alix (Cell Signaling; #2171S), GM130 (BD Biosciences; #610822), Calnexin (Abcam; #ab22595), Calreticulin (Abcam; #ab2907), Pex5 (Novus Biologicals; #NBP1-87185), VDAC (Cell Signaling; #4661), and Actin (Abcam; #ab8227). Densitometries were calculated using ImageJ. The average values of normalized expression levels were plotted.
Tissue Processing and Immunostaining
ODG patient tumor sections were stained using the Opal™ 4-Color Manual IHC Kit (NEL820001KT), as per manufacturer’s instructions. In case of tumor xenograft studies, upon reaching end-time mice were sacrificed and perfused with PBS and 4% PFA solution. Mouse brains were collected and fixed at room temperature for 30 mins. The samples were then rinsed with PBS and immersed in 20% sucrose overnight at 4°C and embedded in OCT. 10 μM cryosections were collected on Superfrost Plus slides (Fisher). For immunostaining, sections were washed and permeabilized in Phosphate Buffered Saline with 0.1% Triton X100 (PBT) followed by blocking with 10% normal horse serum/PBT (blocking solution) for 1 hr at room temperature. Sections were then incubated with primary antibodies overnight at 4°C. Sections were washed with PBT and incubated with secondary antibody for 1 hr at room temperature. Sections were then washed and counterstained with DAPI diluted in PBT at room temperature. Sections were washed in PBS and mounted with coverslips using AquaPolymount (Polysciences). Primary antibodies included: pERK (Cell Signaling; #CS4370S); IDH1 R132H (Dianova; #DIA-H09); Ki67 (Abcam; #ab16667); GFAP (Millipore; #mab360); Olig2 (Abcam; #ab109816), Sox2 (Abcam; #ab97959) HNA (Millipore; #MAB1281), nSMase2 (Abcam; #ab85017), CD63 (Abcam; #ab59479); turboGFP (Origene; #TA150041), BrdU (Abcam; #ab6326). Secondary antibodies included: Alexa 568 donkey anti-rabbit, Alexa 488 donkey anti-rabbit, Alexa 488 donkey anti-mouse (all from Invitrogen), and were diluted in PBT.
Mass Spectrometry
Mass spectrometry was performed on BT088 and BT054 EVs isolated by sequential ultracentrifugation at the SPARC Biocentre-Mass Spectrometry facility at SickKids Research Institute. Mass spectrometry analysis information is provided in Table S3. Scaffold data analysis was performed by applying NCBI annotations to all proteins, removing proteins that matched the search term “keratin” and which did not have “Homo sapiens” under the taxonomy heading. Protein filtering thresholds were set at 99.0%, with a minimum number of 2 peptides and a peptide threshold of 95%. For analysing the EV samples, Cytoscape program and ClueGO plugin were used.
Image analysis
Images were captured with a Leica DM IL LED or DMRXA2 optical microscope using LasX software. ImageJ software was used for image analysis. For astrocytoma and ODG patient tumor analysis, Single channel TIFF images with pERK/Ki-67/OLIG2 staining from all sample sets were transformed to binary format with mean intensity as the selecting parameter. A fixed minimum and maximum threshold value were determined for each set of images to ensure correct thresholding of pERK/Ki-67/OLIG2 staining. Images were analyzed by adjusting the size filter option to count cells with pERK or Ki-67 staining giving the total number of pERK+/ Ki-67+/ OLIG2+ cells in each data set. Single channel images with pERK/Ki-67/OLIG2 staining were merged with IDH1 R132H (IDHm) single channel images. Cells displaying co-localization of pERK/Ki-67/OLIG2 with IDHm was manually counted. For primary neurosphere and tumorsphere assays, neurosphere/tumorsphere sizes were measured using the ruler measurement tool in ImageJ. For cerebral organoid-tumor co-culture assessment, zone-wise cumulative GFP intensity and total number of Sox2+ cells were analyzed using ImageJ. The freeline tool was used to mark the periphery of the organoid. Seven zones (width=150 pixels or 50 μm) were constructed mapping the shape of each organoid, spanning from organoid periphery (zone1) to the core (zone7). Cumulative GFP intensity and Sox2+ cell counts were assessed per zone.
Statistical analysis
A minimum of three biological replicates were carried out for all assays. Statistical analysis and graphs were generated using GraphPad Prism 6 software. Student’s t-test was used when comparing two groups, while One-way ANOVA with TUKEY post corrections were used when comparing groups of more than two. All data expressed as mean value ± standard error of the mean (SEM). In all experiments, a p value <0.05 was taken as statistically significant, *p < 0.05, **p < 0.01, ***p < 0.005, and ****p < 0.001.
AUTHOR CONTRIBUTIONS
AB: conceptualization, data curation, formal analysis, investigation, methodology, visualization, validation, writing – original draft, writing – review and editing
LA: conceptualization, data curation, formal analysis, investigation, methodology, visualization, validation, writing – review and editing
VC: conceptualization, data curation, formal analysis, investigation, methodology, visualization, validation, writing – review and editing
LV: data curation, formal analysis, software
OP: data curation, formal analysis, software
MJC: data curation, formal analysis, investigation
AES: data curation, formal analysis, investigation
TO: formal analyses, software, validation
YT: formal analysis, methodology
STA: formal analysis, software
RI: methodology, investigation
DZ: methodology, investigation
SS: formal analysis, methodology
LCC: formal analysis, methodology
BK: formal analysis, methodology
TF: methodology, investigation
HSL: resources, supervision, writing – review and editing
CMM: resources, supervision, writing – review and editing
MB: resources, supervision, writing – review and editing
VAW: resources, supervision, writing – review and editing
JAC: resources, supervision, validation, writing – review and editing
CS: funding acquisition, conceptualization, project administration, resources, supervision, validation, writing – original draft; writing – review and editing
COMPETING FINANCIAL INTERESTS
The authors declare no competing financial interests.
Supplementary Fig. 1. Oligodendroglioma EVs induce apoptosis in neural stem cells. A-E. ODG patient biopsies co-immunostained with IDHm and pERK (A-D’’). Blue is DAPI counterstain. Yellow arrows indicate IDHm− cells that are pERK+. Percentage of pERK + IDHm− and pERK + IDHm+ cells (E) in oligodendroglioma patients. Regions marked with white dotted boxes in A,C were digitally magnified (4 times) and presented in B-B’’,D-D’’. F - Q. Representative images of Cytotox incorporation in NSCs exposed to FM, CM, CM-EV, FM+EV at day 0,5,7 from one independent experiment. R. NSCs were cultured in FM, BT088 CM, CM-EV, and FM+EV for 3 DIV. Cells were stained with FITC-labeled Annexin V and propidium iodide (PI) and analyzed by flow cytometry. A typical flow cytometry dot plot from one representative experiment is given for each sample. Cell death was determined by quadrant gating: Q1 (necrotic cells); Q2 (late apoptotic cells); Q3 (early apoptotic cells); Q4 (live cells). S. Mean values of Annexin V+ PI+ cell percentages (Q2) were plotted. Bars represent means ± s.e.m‥ *, p < 0.05; **, p < 0.01; ***, p < 0.005. Scale bars: 200 μm.
Supplementary Fig. 2. High SMPD3 expression inhibits oligodendroglioma cell growth in a suspension system. A,B. Immunostaining of BT088 (A) and BT054 (B) cells with Sox10 and nSMase2. Merged images of Sox10 (green) and nSMase2 (red) with DAPI as counterstain in blue. C-G. To monitor effects of SMPD3 overexpression in cells in a suspension system, GFP expressing cells (GFP-BT088 cells: Ctrl/Ctrl-GFP; SMPD3-GFP-BT088 cells: SMPD3/ SMPD3-GFP) were seeded on uncoated plates and growth was monitored by monitoring total area covered by GFP+ cells. Representative images of cell growth at day 14, from one independent experiment, is presented (C-F). Mean values of cumulative GFP+ area (normalized to day 0) ratio were plotted comparing between days 0 to 14 (G). Bars represent means ± s.e.m‥Scale bars: 200 μm.
Supplementary Fig. 3. SMPD3 knockdown is growth enhancing and recapitulated by GW4869, a pharmacological inhibitor. A. Western blot for Alix protein of shScr and shSMPD3-A,B,C,D BT088 cell-derived EVs. B-F. BT088 cells transduced with shScr and shSMPD3-B,C,D constructs were cultured for 9 days on uncoated plates. Growth was monitored by monitoring total area covered by GFP+ cells. Representative images of cell growth at day 9, is presented (B-E). Mean values of total GFP+ area normalized to day 0) ratio were plotted comparing between days 0 to 9 (F). G-K. BT088 cells transfected with shScr-GFP and shSMPD3-GFP were cultured for 10 days. Representative images of BT088 shScr-GFP (G) and shSMPD3-GFP (H) tumorspheres after 10 DIV. Quantitation of tumorsphere number (I), tumorsphere size (J), and live cell number (K). L-O. BT088 cells transduced with shScr and shSMPD3-B,C,D constructs were cultured for 48 hrs and fixed post a 30 min BrdU pulse. Merged images of fixed cells immunostained with BrdU (red), GFP (green), and DAPI counterstain in blue. P-T. BT088 cells treated with DMSO control (Ctrl) and with GW4869 (1 μM) for 14 DIV. Total CD9+ EV particle determination using nanoscale flow cytometry (P,Q). Representative images of Ctrl (R,S) and GW4869 (1 μM) treated cells after 14 DIV. Mean values of total phase object count (relative to day 0) ratio were plotted spanning from day 0 to day 14 (T). U-X. BT088 cells treated with Ctrl (U) and 18 μM GW4869 (V) for 5 DIV. Quantitation of the number of tumorspheres (W) and the size of tumorspheres (μm) (X) generated after 5DIV. Bars represent means ± s.e.m‥ *, p < 0.05; **, p < 0.01; ***, p < 0.005; Student’s t test, and One-way ANOVA when testing more than two groups. Scale bars=200 μm in (B-E); 100 μm in (G,H); 50 μm in (L-O).
Supplementary Fig. 4. Confirming SMPD3 knockdown of xenografted oligodendroglioma cells. A,B. BT088 GFP tumor xenograft sections immunostained with nSMase2 (red) and Human Nuclear Antigen (HNA; green), and DAPI as counterstain in blue (A). Region marked in white box in (A) is magnified (4X) and presented in (B). C. Western blot for whole HEK cell lysates co-transfected with shScr or SMPD3 knockdown (KD) construct and SMPD3 construct. Expression of nSMase2 was assessed for knockdown in SMPD3 expression. Three biological replicates of each sample set were loaded in varying amounts (5X/1X) to show efficiency of knockdown. D-G. Representative images of BT088 cells transfected with shScr (D,D’), and SMPD3 KD (E-E’) cultured for 10DIV (BF, Brightfield). Quantitation of GFP+ tumorsphere numbers (F) and sphere diameter (μm) (G) after 10 DIV. H. hESC cerebral organoid generation protocol. 30day cerebral organoids were co-cultured with BT088 cells (shScr, shSMPD3-B/D) for 7 DIV. Bars represent means ± s.e.m‥ Scale bars: 100 μm.
ACKNOWLEDGEMENTS
We thank Dr. Gregory Cairncross and Dr. Sam Weiss (University of Calgary) for providing the BT088 and BT054 tumor cell lines for our study. We thank Ali Darbandi (TEM), Lindsey Fiddes (SEM) and Greg Wasney (NTA) for technical help. This project was supported by an International Development Research Centre (IDRC 108875) to CS, Canada First Research Excellence Fund (CFREF) Medicine by Design Cycle 2 to CS and CM, and a Cancer Research Society grant to CS and JAC. CS holds the Dixon Family Chair in Ophthalmology Research.
Footnotes
↵* co-senior authors