Niche-derived soluble DLK1 promotes glioma stemness and growth

Tumor cell behaviors associated with aggressive tumor growth such as proliferation, therapeutic resistance, and stemness are regulated in part by soluble factors derived from the tumor microenvironment. Tumor-associated astrocytes represent a major component of the glioma tumor microenvironment, and astrocytes have an active role in maintenance of normal neural stem cells in the stem cell niche, in part via secretion of soluble Delta-like Non-Canonical Notch Ligand 1 (DLK1). We found that astrocytes, when exposed to stresses of the tumor microenvironment such as hypoxia or ionizing radiation (IR), increased secretion of soluble DLK1. Tumor-associated astrocytes in a glioma mouse model expressed DLK1 in perinecrotic (hypoxic) and perivascular tumor areas. Glioma cells exposed to recombinant DLK1 displayed increased proliferation, enhanced sphere and colony formation abilities, and increased levels of stem cell marker genes. Mechanistically, DLK1-mediated effects on glioma cells involved increased and prolonged stabilization of Hypoxia-Inducible Factor 2alpha (HIF-2alpha), and inhibition of HIF-2alpha activity abolished effects of DLK1 in hypoxia. Forced expression of soluble DLK1 resulted in more aggressive tumor growth and shortened survival in a genetically engineered mouse model of glioma. Together, our data support DLK1 as a soluble mediator of glioma aggressiveness derived from the tumor microenvironment. Graphical abstract Highlights Astrocytes secrete DLK1 after exposure to hypoxia or irradiation Soluble DLK1 promotes stemness in glioma, in part by increasing HIF-2alpha stabilization. High levels of soluble DLK1 are associated with tumor aggressiveness and lethality.


INTRODUCTION
Glioblastoma recurrence following standard-of-care treatment with radiotherapy, surgery, and chemotherapy invariably gives rise to incurable lesions, and the median survival following diagnosis remains dismal at 15-20 months despite recent advances in our understanding of glioblastoma at a molecular level (Huse & Holland, 2010).
Evidence from human patient samples and murine models of brain tumors suggest that inherent therapeutic resistance within a subset of tumor cells with stem cell characteristics may be the primary source of recurrent tumors (Bernstock et al., 2019;Dirkse et al., 2019). While the origin and fate of such cells remain controversial, glioblastoma cell phenotypes are highly plastic, and non-stem-like cells can acquire characteristics of stem cells as a result of microenvironmental interactions with the extracellular matrix, with growth factors, or as a result of altered oxygenation or pH (Dirkse et al., 2019;Hambardzumyan & Bergers, 2015;Jawhari, Ratinaud, & Verdier, 2016;Ryskalin et al., 2019). Indeed, tumor cells with stem cell properties appear to be spatially restricted to specific microenvironments such as the perinecrotic (hypoxic) and perivascular niches, suggesting that these niches may control residing tumor cell phenotypes (Hambardzumyan & Bergers, 2015;Jawhari et al., 2016;Majmundar, Wong, & Simon, 2010).
Delta Like Non-Canonical Notch Ligand 1 (DLK1) is a transmembrane protein in the Notch family of ligands, that is capable of signaling in a Notch-dependent andindependent manner depending on cellular context (Baladrón et al., 2005;Ceder, Jansson, Helczynski, & Abrahamsson, 2008;Falix, Aronson, Lamers, & Gaemers, 5 2012; Ferrón et al., 2011;Huang et al., 2019;Li et al., 2014). Expression of DLK1 is increased with tumor grade in glioma, and its signaling has been associated with various tumor cell properties such as proliferation, invasion, and stemness (Grassi, Pantazopoulou, & Pietras, 2020;Yin et al., 2006). The mechanisms underlying these effects on tumor cell behavior remain poorly understood, but likely include signaling from the extracellular, soluble domain of DLK1 (Wang & Sul, 2006). Indeed, soluble DLK1 secreted from astrocytes was recently shown to be a critical component of the subventricular zone (SVZ) neural stem cell niche (Ferrón et al., 2011). Astrocytes represent a prominent cell type in the brain tumor microenvironment (Mega et al., 2020), and recent studies revealed that DLK1 is one of the top upregulated genes in tumor-associated astrocytes of high-grade vs. low-grade gliomas (Katz et al., 2012).
The regulation of DLK1 expression is poorly understood, but some elements of the brain tumor microenvironment such as hypoxia have been shown to drive DLK1 expression (Begum, Kim, Lin, & Yun, 2012;Kim, Lin, Zelterman, & Yun, 2009).
Here, we sought to investigate the role of soluble DLK1 in the high-grade glioma tumor microenvironment. We found increased secretion of DLK1 from tumorassociated astrocytes subjected to stresses of the tumor microenvironment, such as hypoxia and ionizing radiation (IR). Soluble DLK1 increased proliferation and stemness of glioma cells, and promoted tumor growth in a genetically engineered mouse model of glioma. Together, our findings suggest that soluble DLK1 is a nichederived mediator of aggressive tumor growth in brain tumors.
Soluble DLK-S was cloned into the RCAS vector and mice were co-injected with a 1:1 mix of DF1 cells expressing PDGFB and DLK-S or empty RCAS as indicated. Each litter was allocated to one experimental group. Mice were monitored daily and sacrificed upon displaying brain tumor symptoms. All procedures were approved by the Malmö-Lund Ethical Committee (permit M186-14). The sample number was determined based on the law of diminishing returns with the resource equation method (total number of animals -total number of groups >10). A total of 6 pups were excluded due to non-tumor symptoms during week 0-3, the final numbers were n=26 for PDGFB and n=24 for DLK-S.

Immunofluorescence
Whole brains were embedded in OCT (ThermoFisher) and frozen in pre-cooled isopentane. 5 µm thick cryosections were air-dried for 30 min, then fixed in ice-cold acetone or 4% PFA and permeabilized using 0.3% Triton X-100 in PBS (Sigma).

Colony and sphere formation assays
Mechanical dissociation with Accutase (ThermoFisher) was used to prepare single cell suspensions. Cells were counted using a hemocytometer. For colony formation assay, 350 cells were seeded in 5 cm dishes coated with polyornithine (Sigma) and laminin (Biolamina). U3082MG, U3084MG and U3065MG cells were cultured for 14 days while PIGPCs for 8 days, under the indicated conditions, then washed in PBS and fixed using 4% PFA. Cells were stained using 0,01% crystal violet/H2O. Wells were washed gently in water, then air-dried for 24 hours. Images were acquired with a Fujifilm LAS 3000 Imager.
Sphere formation assay was performed with the hanging-drop method. 10 cells in 35 µl drops were seeded on the lid of a 48 well plate and grown under the indicated conditions for 2 weeks. For secondary sphere assay, primary spheres were pooled, pelleted, dissociated with Accutase and reseeded at the indicated conditions. Wells with spheres were manually counted and images were acquired with a Zeiss AX10 inverted microscope.

Real-Time Quantitative PCR
The

Data availability
The data that support the findings of this study are available from the corresponding author upon reasonable request.

DLK1 is expressed and secreted by tumor-associated astrocytes in the glioma microenvironment
To test whether tumor-associated astrocytes could be a source of DLK1 in the glioma tumor microenvironment, we generated PDGFB/shp53-induced murine gliomas using the RCAS/tv-a system as previously described (Grassi et al., 2020), then costained tumors for the astrocyte marker GFAP and DLK1. While the bulk of the tumor cells appeared negative for DLK1 expression, DLK1 signal was detected in areas of GFAP staining both in perinecrotic and perivascular tumor areas, and both in GFAP positive and negative cells, suggesting DLK1 expression in astrocytes and tumor cells in these areas (Fig. 1A-B). 15 The use of an independent antibody directed against the N-terminal, soluble domain of DLK1 also showed specific signal in the perinecrotic tumor areas (Supp. Fig. 1) and co-staining of GFAP and DLK1 (Fig. 1C-D).
DLK1 gene expression was previously reported to be upregulated in isolated GFAP+ tumor-associated astrocytes of high-grade glioma compared to those of lower-grade tumors (Katz et al., 2012), further confirming DLK1 expression in astrocytes in this model system. We speculated that this DLK1 induction could be mediated by microenvironmental factors. As hypoxia is one major microenvironmental reality of high-grade glioma compared to low-grade glioma (Jawhari et al., 2016;Jensen, 2009), we cultured human fetal astrocytes under normoxic or hypoxic conditions for up to 10 days. We also subjected astrocytes to a single dose of irradiation to mimic another physiological response to a therapeutic intervention relevant to high-grade glioma. Both astrocytes subjected to growth in hypoxia and those subjected to irradiation displayed increased DLK1 levels secreted into the culture media, as measured by an ELISA assay ( Fig. 2A). The higher levels of DLK1 in the media were sustained for the entire 10 day period in the case of astrocytes cultured in hypoxia, whereas irradiated astrocytes displayed a peak secretion at 2 days post-treatment with levels returning to baseline after 9 days ( Fig. 2A). Together, these data support that astrocytes may secrete DLK1 into the tumor microenvironment in glioma.

Soluble DLK1 promotes glioma cell proliferation, survival and self-renewal
We next examined what effect soluble DLK1 may have on glioma cells. 16 We first performed a media transfer experiment by treating human glioblastoma cell lines maintained in serum-free, stem cell-promoting conditions with media from astrocytes cultured in normoxic (ACM CTRL) and hypoxic (ACM 1% O 2 ) conditions for up to 9 days. In 2 out of 3 cell lines, media from hypoxic astrocytes induced a significant increase in the proliferation rate (Fig. 2B). Since astrocyte conditioned media contains many other different factors that may influence cancer cell growth, secreted part, and once again, the two DLK-responding cell lines showed significant increase in their proliferation (Fig. 2F). Taken together, these data demonstrate that soluble DLK1 is able to induce glioma cell proliferation, irrespectively of its origin.
As the use of the recombinant protein allows for better control of DLK1 concentrations, we then moved forward with this approach. We first generated a dose-response curve in all 3 human glioblastoma cell lines and in PDGFB-induced glioma primary cultures (PIGPCs) derived from the glioma mouse model. All the cell lines, with the exception of the non-responding U3065MG, showed a dose 17 dependent increase in cell proliferation, with a plateau obtained at 200 ng/ml DLK1 and EC50s in between 25 and 35 ng/ml (Fig 3A).
In line with these findings, all cell lines that responded to DLK1 in the proliferation assay also increased their colony formation ability in a dose-dependent manner when exposed to sub-maximal soluble DLK1 concentrations similar to those obtained in hypoxic astrocytes ( Fig. 3B-C).
Furthermore, soluble DLK1 strongly enhanced the self-renewal ability of responsive glioma cell lines, as measured by the serial sphere-formation assay ( Fig. 4A-B), and induced a significant increase in the stemness markers OCT4, NANOG and SOX2 (Fig.   4C). Since DLK1 has been reported to influence the Notch pathway (Baladrón et al., 2005;Huang et al., 2019), we tested if these effects were Notch-dependent.
Luciferase experiments performed at different time points revealed no significant alterations in Notch activity in any tested cell lines (Fig. 4D).

DLK1-effects are mediated in part by HIF-2alpha
Because DLK1 secretion was increased by astrocytes under hypoxic conditions, and because of known previous links between DLK1 expression and function to hypoxia (Kim et al., 2009) (Fig. 5A-B). This increased HIF-2alpha expression was reflected in a stronger hypoxic response, as 2/3 human glioblastoma lines and PIGPCs displayed increased activation of hypoxia-responsive elements (HREs) at 72 h of culture in hypoxia with DLK1 stimulation, as measured in an HRE-luciferase assay (Fig. 5C). Moreover, analysis of PDGFB/shp53-induced murine gliomas revealed that both DLK1 and HIF-2alpha were strongly expressed and showed a significant co-localization only in the perivascular and perinecrotic niches ( Fig 5D).
As HIF-2alpha is a known driver of stem cell characteristics in glioma and other tumor forms (Das et al., 2019;Johansson et al., 2017;Pietras et al., 2008;Pietras, Johnsson, & Påhlman, 2010;Yan et al., 2018), we next tested whether effects of DLK1 on glioma cell behavior were mediated by HIF-2alpha. The treatment with the specific HIF-2alpha inhibitor PT2385 at concentrations with significant effects on HIF-2alpha protein (Persson et al., 2020;Wallace et al., 2016) (Supp. Fig 2) was able to revert the DLK1-induced increase in hypoxia response (Fig. 6A). Similarly, while stimulation of glioma cells with soluble DLK1 boosted the increase in the expression of the stem cell marker genes NANOG, OCT4 and SOX2 in hypoxic cells, addition of PT2385 blocked this specific DLK1 effect in all cell lines tested (Fig. 6B). Moreover, PT2385 decreased the colony formation ability of glioma cells exposed to soluble DLK1 in hypoxia (Fig. 6C). Notably however, PT2385 treatment did not significantly affect DLK1-induced gene expression in normoxia. Together, these data suggest that DLK1 promotes the glioma stem cell character in part via HIF-2alpha stabilization.

DLK1 promotes aggressive glioma growth in vivo
To test the effects of soluble DLK1 on glioma growth in vivo, we generated a mouse model for the overexpression of soluble DLK1 together with PDGFB using the RCAS/tv-a system (Fig. 7A). Co-injection of RCAS-PDGFB with RCAS-DLK-S (soluble) resulted in more aggressive tumors as compared to RCAS-PDGFB with empty vector control, as measured by survival time following injections (Fig. 7B). Evaluation of Ki67 expression revealed a significant increase in cell proliferation in murine DLK-S tumors as compared to controls (Fig. 7C, D), thus confirming the in vitro data 3A). In agreement with our in vivo data, analysis of the human TCGA GBMLGG dataset (Ceccarelli et al., 2016) revealed that tumors expressing high levels of DLK1 were significantly more aggressive than those with low levels of DLK1 (Fig. 7E), presumably as a result of the higher DLK1 levels reported in high-grade glioma.

DISCUSSION
An increasing focus on cancer stemness has revealed parallels between normal neural stem cell regulation and cancer stem cell characteristics in brain tumors (Dirks, 2010;Lathia, Mack, Mulkearns-Hubert, Valentim, & Rich, 2015). Control over tumor cell phenotypes by specific, local microenvironments within a tumor, for example, is reminiscent of the way that normal tissue stem cells reside within and rely on their niche to maintain the stem cell character (Dirkse et al., 2019;Hambardzumyan & Bergers, 2015). It is likely that some of the same mechanisms involved in neural stem cell maintenance in the vascular niche of the SVZ may also 20 be involved in maintaining stemness of brain tumor cells located in a perivascular niche. We describe one such example here: soluble DLK1 secreted from astrocytes appears to be involved in stemness maintenance both of normal neural stem cells and glioma cells, as shown here. An association between DLK1 expression and aggressive tumor growth in glioblastoma has previously been established (Kim et al., 2009). By generating a mouse model for testing effects of soluble DLK1 overexpression specifically in the context of glioblastoma, we show that the previously reported association between DLK1 expression and tumor grade in glioma (Grassi et al., 2020, Yin et al., 2006   Statistical analysis: independent experimental replicates are as follow, A n=6, except U3065MG where n=4, C n=4, Statistical significance was determined by one-way ANOVA, followed by Bonferroni post hoc test.
In the whole figure significance is represented as * p<0.05 and ** p<0.01 vs.