An affinity for brainstem microglia in pediatric high-grade gliomas of brainstem origin

Abstract Background High-grade gliomas (HGG) in children have a devastating prognosis and occur in a remarkable spatiotemporal pattern. Diffuse midline gliomas (DMG), including diffuse intrinsic pontine gliomas (DIPG), typically occur in mid-childhood, while cortical HGGs are more frequent in older children and adults. The mechanisms behind this pattern are not clear. Methods We used mouse organotypic slice cultures and glial cell cultures to test the impact of the microenvironment on human DIPG cells. Comparing the expression between brainstem and cortical microglia identified differentially expressed secreted proteins. The impact of some of these proteins on DIPGs was tested. Results DIPGs, pediatric HGGs of brainstem origin, survive and divide more in organotypic slice cultures originating in the brainstem as compared to the cortex. Moreover, brainstem microglia are better able to support tumors of brainstem origin. A comparison between the two microglial populations revealed differentially expressed genes. One such gene, interleukin-33 (IL33), is highly expressed in the pons of young mice and its DIPG receptor is upregulated in this context. Consistent with this observation, the expression levels of IL33 and its receptor, IL1RL1, are higher in DIPG biopsies compared to low-grade cortical gliomas. Furthermore, IL33 can enhance proliferation and clonability of HGGs of brainstem origin, while blocking IL33 in brainstem organotypic slice cultures reduced the proliferation of these tumor cells. Conclusions Crosstalk between DIPGs and the brainstem microenvironment, in particular microglia, through IL33 and other secreted factors, modulates spatiotemporal patterning of this HGG and could prove to be an important future therapeutic target.

High-grade gliomas (HGGs) constitute 8%-10% of brain tumors in children and usually cause death within 2 years of diagnosis. 1 Pediatric gliomas can arise in the supratentorial cerebral cortex or in infratentorial regions, such as the cerebellum, brainstem, midbrain, thalamus, and spine. 2 Approximately half of pediatric high-grade gliomas (pHGGs) arise in midline locations, specifically in the thalamus and the pons, and are classified as diffuse midline gliomas (DMGs; pontine DMG is also known as diffuse intrinsic pontine glioma [DIPG]). While pontine and thalamic DMGs typically occur in mid-childhood, pediatric cortical gliomas occur in older children and young adults, and HGGs of later adulthood occur mainly in the frontotemporal lobes. 2 Recent studies make it increasingly apparent that supratentorial and infratentorial pHGGs differ from one another and from their adult counterparts. In particular, the predominant oncogenic mutations vary depending on location and on the patient's age. A striking example is the lysine to methionine mutation on residue 27 in histones H3.1 (H3.1K27M) and H3.3 (H3.3K27M) observed in approximately 80% of pontine DMG cases in younger children, while tumors with H3.3G34R/V mutations usually occur in the cerebral hemispheres of adolescents and young adults. [2][3][4] One possible explanation for this remarkable spatiotemporal pattern of DMGs may be a local advantage of the microenvironment of the infratentorial regions in a restricted time frame that fosters the development of tumors carrying these specific mutations. This assumption is consistent with the limited ability to induce tumorigenic transformations in older mice when using genes associated with DIPG mutations. 5 A preference for pontine DMGs toward the young brainstem would imply that specific cells at a restricted time point in this location may be more suitable for the formation, progression and survival of such tumors. If specific cell types in the brainstem indeed give an advantage to the formation of DIPG tumors, such cells should have differential anatomical and developmental properties. Consistent with this idea, accumulating evidence indicates that various cells in the CNS such as astrocytes, oligodendrocytes, and microglia show some anatomical and functional differences. [6][7][8][9][10] Evidence for interaction and crosstalk between microglia, astrocytes, and vascular cells in adult forms of glioma has been previously demonstrated. [11][12][13] Pediatric brain tumors were also recently shown to be influenced by their environment. 14,15 In this case, activated neurons in the microenvironment were shown to promote glioma growth.
In this work, we tested the possibility that pontine DMGs of brainstem origin are more attuned to brainstem surroundings, and subsequently aimed to find an anatomically specific DIPG-microenvironment crosstalk contributing to this compatibility.

Materials and Methods
Animals: C57BL/6 mice were obtained from ENVIGO (Rehovot, Israel). Animal handling adhered strictly to national and institutional guidelines for animal research and was approved by the Ethics Committee of the Hebrew University. Mice both males and females between days 2 and 4 after birth were used.
Antibodies: a list of antibodies is in Supplementary methods.
Immunostaining and Western blot: explanation is found in Supplementary methods.

Patient-derived tumor cell cultures
All cultures were maintained as neurospheres in tumor stem medium (TSM) consisting of 50% Neurobasal(-A)/50% DMEM/ F12, supplemented with B27 TM (-VitA) and growth factors as described previously 14 and in Supplementary methods.

GFP + or luciferase DIPG cell line preparation.-Each
patient-derived HGG cell line was infected with Lenti-GFP/Luciferase viral supernatant and allowed to recover for 1 week. GFP-positive cells were isolated for purity by fluorescence-activated single cell sorting (FACS) (BD FACS Aria) and returned to culture. Luciferase-positive cells were isolated via Puromycin (1 μg/ml) selection. Validation of the sorting was done using flow cytometry analysis and by observing the cell fluorescence under the microscope (with at least a 90% positive GFP rate).

Importance of the Study
Pediatric gliomas can arise in the supratentorial cerebral cortex with a median diagnosis age of 13 years, while diffuse midline gliomas (DMGs), present a median diagnosis age of 7 years, and appear most frequently in the pons. DMGs have a devastating prognosis and are currently untreatable. We tested the contribution of the brainstem microenvironment to the observed spatiotemporal pattern of pediatric high-grade gliomas.
Our results support a model in which the brainstem microenvironment, particularly microglia, offers more supportive conditions for diffuse intrinsic pontine gliomas (DIPG) cells. By using the differences in gene expression between brainstem and cortical microglia, we were able to identify factors that may contribute to the DIPG preference toward the brainstem. In this study, we focused on the factor interleukin-33 and provided evidence to support its contribution to the propagation of DIPGs in this specific anatomical location.

Neuro-Oncology Advances
micromanipulator (Tokyo, Japan). The cells were systematically injected in parallel into both the brainstem and cortex of P2-4 mouse brains, two injections per region. The injections of each region were performed consecutively, alternating the order of injections between brainstem and cortex. After injections, the brain regions were gently separated and tissue pieces were incubated on top of 40-µm cell filters. Each well was filled with medium (Neurobasal, 15 mg/ml glucose, AmphoB 25 µg/ml, Pen-Strep, and B27 [without vitA]) up to the level of the filter to allow diffusion. Incubation was for 24-72 h in 5% CO 2 , 37° C.

Ex vivo multiphoton imaging and quantification
Each injected cortex and brainstem pair was incubated for 24 h, then sealed with a cover slip (moisturized with phosphate-buffered saline (PBS)) and imaged live, using the large image option of Nis-Elements, and the maximal width in the Z stack option so as to capture all of the injected GFP + cells. The analysis is described in Supplementary methods.

Ex vivo GFP quantification analysis.-After 24 h, slices
were dissociated and analyzed in a Fortessa LSR cell analyzer (BD Biosciences, CA, USA). We made sure to record all of the events from each sample (about 3-8 million events per sample), so that quantification represents the absolute GFP + cell count per brain region injected. For each cell line, at least 9 brains were injected, in at least 3 injection sessions.
Ex vivo CellTrace proliferation analysis.-Glioma cells were labeled with CellTrace reagent and injected into mouse brains as previously described (CellTrace™ Violet Cell Proliferation Kit, for flow cytometry, Thermofisher, MA, USA). A T0 measurement (maximal CellTrace intensity) was obtained per trial and used to determine the percentage of dividing cells after 24-or 72-h incubation.
Ex vivo cell death analysis.-Following slice culture incubation of either 6 or 24 h, injected tissue samples were dissociated and stained with 100 µg/ml Propidium Iodide (PI) (Thermofisher, MA, USA). Flow cytometry was used to quantify the percentage of PI-positive cells out of GFPpositive cells.

Microglia and astrocytes cell culture
Preparation of mixed glia primary cell cultures and astrocytes from 2-to 3-day-old mice was previously described. 16 Microglia were prepared as described here. 17 Briefly, to produce an astrocyte cell culture, the cells are plated in very low density, 20 ng/ml EGF (PeproTech, Rehovot, Israel) is applied in the first few days, and when confluent the culture gets shaken overnight, treated with AraC (0.1 mM, Sigma Aldrich) and passaged twice in order to ensure purity. For microglia isolation, mixed glial cells were plated densely and treated with 0.5 ng/ml gmCSF (PeproTech, Rehovot, Israel), until the cells are easily isolated from the culture and plated for experiments. An extended explanation is found in Supplementary methods. qPCR DNA quantification.-Testing of cell number with DNA-based qPCR and primers was carried out based on a previous paper with some modifications. 18 Briefly, DNA from a controlled number of glioma and mouse (astrocytes and microglia) cells was acquired, then a qPCR reaction (iTaq Universal SYBR Green Supermix, BioRad, Rishon Le Zion, Israel) using either a human-specific primer or a primer designed to detect both human and mouse cells was performed on the control DNA samples. (Oligonucleotide sequences can be found in Supplementary methods.) The results were used to create transformation formulas from cycle threshold (CT) to cell number/DNA amount with a high prediction power (R 2 = 0.99 for both primers). To test the strength of these transformation formulas, sets of quantified human and mouse DNA samples (from 100% human DNA, 90% human with 10% mouse, etc., all the way to 100% mouse DNA) were tested. Using our transformation formula, we were able to accurately estimate the ratio of mouse and human cells in a mixed culture, with a strong predictive value (R 2 = 0.98).

Co-culture proliferation experiments.-Eight thoushand
human HGGs were placed on top of mouse glial cell cultures (12 well)-astrocytes and microglia, originating either in the brainstem or the cortex of the same mice. For each condition, within each trial, T0 wells were acquired about 2 h after the trial began. The co-culture was then incubated for 72 h in glioma medium, after which DNA was acquired using the Monarch DNA cleanup kit (NEB, Ipswich, MA, USA), and DNA-based qPCR quantification was done, as described above.
Microglial conditioned medium preparation. For each trial, a confluent microglial cell culture was prepared, taking care to maintain similar conditions for the brainstem and cortical derived microglia. Once cell cultures reached a similar confluence, the cells were thoroughly washed to remove any residual serum, and TSM without factors was added. For each 4-cm 2 well, 1.5 ml of medium was added. After 72 h, the medium was collected and filtered.
Luciferase-based viability assay. Luciferase-labeled DIPG cells were plated on top of purified microglia or astrocytes (670 cells/well), in 0.3-cm 2 wells. Following a 72-h incubation, luciferase activity was measured using LUMIstar® Omega (BMG LABTECH, Ortenberg, Germany) microplate luminescence reader. Each experiment consisted of a minimum of 4 wells per condition, and each experiment was repeated at least 6 times per cell line.
Three-dimensional growth experiments. Recombinant human IL33 (50 ng/ml, PeproTech, Rehovot, Israel) was added to a TSM and Matrigel mixture (1:1 ratio) with DIPG cells highly concentrated to create very dense cell-matrigel droplets. These marked a specific starting point. Control droplets were also added. After the initial solidifying of the gel, the radius of the droplet was expanded by adding more Matrigel-TSM mix. After final solidification, TSM-based medium was added on top of the Matrigel drops to cover them. The diameter of each DIPG drop was measured after 2 h and 7 days. ImageJ was used to measure the change in diameter. In each trial, at least 7 droplets for each condition were used.
Proliferation experiments. CellTiter-Glo® Luminescent Cell Viability Assay (Promega, Madison, WI, USA) was used as described in Supplementary methods in the "Recombinant IL33 concentration calibration. " Colony count and average colony size experiments. Four hundred DIPG cells were plated in 0.3-cm 2 wells. This low-density plating allowed clear segregation of colonies. For each condition, at least 7 wells were plated per repeat. After 5 days incubation, the cultures were analyzed using Incucyte® S3 (Sartorius, Göttingen, Germany) Live-Cell Analysis system (whole well image X4). mIL33 neutralization trial in ex vivo slice cultures DIPG cells, labeled with CellTrace Dye, were injected with 1 μg/ml of mIL33 antibody or isotype control antibody into each side of the cortex and of the brainstem, then slice cultures were prepared. After 3 days, the proliferation of DIPGs within each condition was measured using flow cytometry analysis. An additional cortical injection was used to exclude any brains in which no brainstem preference was detected from the statistical analysis, excluded brains: DIPGXIII 3/13, DIPGXXV 1/12, BT869 3/14.

RNA-sequencing analysis
shRNA IL1RL1 knockdown.-This was performed essentially as previously described. 21 For IL1RL1 knockdown experiments, we used Mission® shRNA PLKO lentiviral vectors which include puromycin selection marker: TRCN0000358832/sh-IL1R-1 and TRCN0000058513/ sh-IL1R-2. For a control, we used an empty vector (pLKO.1puro); all vectors purchased from Sigma Aldrich (St. Louis, MO, USA). Details are described in Supplementary section. qPCR Analysis.-This was performed essentially as previously described. 21 Differential expression was determined using the delta CT method. Each primer set was tested on human and mouse samples to make sure each primer is species specific. Further details are in Supplementary methods. Statistical analysis.-Generally, non-parametric, coupled (Wilcoxon) or un-coupled (Mann-Whitney), one-tailed statistical tests were applied. Details are described in Supplementary methods.

Preference of pediatric HGGs of brainstem origin toward the brainstem
To test the intrinsic compatibility of DIPGs to their anatomical origin, we adopted an ex vivo approach, which allowed us to test the initial response of DIPGs to their microenvironment. We injected equal numbers of GFP-labeled, patient-derived DIPG cells into the brainstem and cortex of the same brains followed by organotypic slice cultures. Using two-photon microscopy (Figure 1a-c), we detected higher numbers and a broader distribution of DIPG cells in the brainstem slices in comparison to the cortex.
To quantify the number of tumor cells in each region, we used flow cytometry analysis. In all DIPGs tested, more cells were present in the brainstem. In contrast, the pediatric HGG of cortical origin, KNS-42, showed a strong preference toward the cortex, while another cortical line, SU-pcGBM2, showed no preference to either site (Figures 1d and S1).
To test proliferation, we used CellTrace™ dye at 24 and 72 h (Figure 1e). At 24 h, there was very little DIPG division overall, suggesting improved survival as opposed to proliferation in the brainstem at this time point. To test this directly, tumor cell death in the slice cultures was analyzed. After 6 h (Figure 1f), cells from all DIPG lines died less in the brainstem while SU-pcGBM2 showed an opposite response. At 24 h ( Figure S2), only BT869 and pcGBM maintained their original preference while the other lines showed no significant difference ( Figure S2). Surprisingly, KNS-42 died less in the brainstem. We then tested proliferation after 72 h, when initial difference in proliferation was detected. We found significantly more DIPG division in the brainstem. In contrast, SU-pCGBM2 and KNS-42 showed a trend toward the cortex although this was not statistically significant (Figures 1g and S3).

Brainstem microglia increase the accumulation of HGGs of brainstem origin
To assess the glial cells' contribution, we tested the proliferation of SU-DIPGXIII on mixed glial cultures from the brainstem and cortex using a DNA-based qPCR approach. We calculated the estimated proportion of DIPG cells according to cycle threshold, based on calibration curves prepared with known numbers of human and mouse cell mixes ( Figure S4). Interestingly, DIPG cells proliferated significantly more when incubated with brainstem glia (Figure 2a). Since microglia are known to infiltrate the tumor microenvironment and contribute considerably to glioma development, [22][23][24][25] we decided to test their role. DIPGs were plated with purified microglia from each brain region. Tumor cell number was estimated using the DNAbased qPCR quantification described above. All HGGs of brainstem origin accumulated significantly more on brainstem microglia, while SU-pCGBM2 did not (Figures 2b and  S5a). In contrast, DIPGs showed no anatomical preference on astrocytes (Figures 2c and S5b), while SU-pcGBM2 proliferated more on cortical astrocytes. Although microglia and astrocytes are altered and activated in culture, we assumed these alterations occur in cells from both regions, allowing an informative comparison.

Neuro-Oncology Advances
To validate the DNA qPCR results, two DIPG lines labeled with luciferase were used to estimate DIPG accumulation on microglia or astrocytes, with similar results (Figures 2d,e and S5c). As a final validation step for the effects of microglial cells on DIPGs, we used anti-human-specific    . (e) Representative immunofluorescence for IL33 and CD11b on brainstem or cortical microglia (scale bar 10 μm). Analysis was performed with Nis-elements by measuring the ratio of antibody to identify and count SU-DIPGXIII cells with a comparable outcome ( Figure S6). Altogether, a clear intrinsic preference of DIPGs to brainstem microglia was documented.

Neuro-Oncology Advances
To test the involvement of secreted factors, a conditioned medium from brainstem and cortical microglial cell cultures (incubated without tumor cells) was used in a colony formation assay. In all DIPGs tested, the number of colonies was significantly higher with brainstem conditioned media (Figure 2f). Thus, at least part of the propagative influence of brainstem microglia is mediated by secreted factors.

Differential gene expression between brainstem and cortical microglia
Based on the observed functional difference between microglia of brainstem and cortical origin, we analyzed the differences in gene expression between them. Since HGGs and microglial cells might impact each other's gene expression, microglia were incubated with SU-DIPGXIII. For gene expression analysis, the microglia-DIPGXIII sample data were aligned to a combined human and mouse reference genome. Though differences were detected, no statistically significant differences could be found in gene expression of the human genes (SU-DIPGXIII) by applying a uniform statistical cutoff. In contrast, mouse microglia revealed a significant number of differentially expressed genes (Figure 3a; Table S1).
To select relevant genes for analysis, we used the Uniprot database, and identified 92 protein-coding genes of secreted or membrane-associated proteins (Table S1). We then focused on potentially novel candidates with a high fold change and known links to cancer, particularly glioma. Based on these criteria, we selected 7 genes. To validate their differential expression, RNA from microglia-SU-DIPGXIII co-cultures was tested using mouse-specific qPCR analysis (Figure 3b).
Next, we tested the ability of proteins coded by these genes to induce the proliferation of SU-DIPGXIII ( Figure  3c). Out of the 7 candidates, only IL33 significantly enhanced HGG proliferation. To test whether IL33's differential expression is intrinsic, we performed qPCR ( Figure 3d) and immunostaining (Figure 3e) on purified microglia. Results indicated that this difference is indeed intrinsic. Consistent with this, the expression level of IL33 mRNA transcripts and protein is higher in the brainstem of mice at a young age (Figure 3f,g), with this higher expression being at least partly due to the expression in brainstem microglia (Figures 3h and S7). Furthermore, according to qPCR analysis, this anatomically differential expression significantly lessens at the age of 2 mo and is most significant from birth up to 8 weeks ( Figure S8) Interestingly, IL33 expression in DIPG patients is significantly higher than in patients with cortical low-grade glioma or medulloblastoma (Figure 3i). The higher expression levels of IL33 in human DIPGs and pediatric cortical HGG samples are consistent with the IL33 data in adult HGGs. [26][27][28] Since cellular resolution is lacking, the source of IL33 in DIPGs may be the microenvironment, as our data suggests, while in the cortical HGGs, the source could be the glioma cells, as has been shown for adult HGGs. 26,27 IL33 is able to support DIPG tumor behavior in vitro and in slice cultures To find the working concentration of IL33, we carried out a dose-response proliferation assay using DIPGXIII ( Figure  S9), and selected a concentration of 50 ng/ml, which is consistent with previous studies. [29][30][31] We then tested the effect of recombinant IL33 on DIPG proliferation ( Figure  4a), 3D growth (Figures 4b and S10), and colony formation (Figures 4c and S11). In all DIPG lines tested, IL33 significantly enhanced tumorigenesis.
To test the role of IL33 in proliferation in the brainstem, we used CellTrace™-labeled DIPG cells injected into slice cultures, together with either IL33 blocking antibody or isotype control antibody (Figure 4d). All three DIPGs tested proliferated significantly less in the presence of the IL33 antibody ( Figure 4d). The relatively modest effect might indicate limited accessibility of the antibody or perhaps the contribution of other factors present in the brainstem.

Expression of IL33 receptors in DIPGs
We explored the PedcBioPortal and found that expression levels of IL1RL1, IL33's binding receptor, are higher in DIPG and cortical HGG biopsies compared to expression levels in cortical low-grade cortical gliomas ( Figure  5a

Neuro-Oncology Advances
in which IL33 has been suggested to play a role. [26][27][28]31,32 Consistent with the biopsy data, IL1RL1 was expressed at much higher levels in SU-DIPGXIII cells injected into the brainstem (organotypic slice culture) as compared to cultured cells, while IL1-RAcP (IL33's co-receptor) was also affected, but more moderately (Figure 5b). Furthermore, both receptors were expressed at higher levels in cells injected into organotypic slice cultures of the brainstem in comparison to the cortex (Figure 5c,d), and on brainstem microglia (Figure 5e). To test the role of IL1RL1 signaling in DIPG proliferation, a knockdown approach was utilized. SU-DIPGXIII was treated with shRNAs targeting IL1RL1, and the effect was validated by qPCR ( Figure S12). The knocked-down cells were then tested in an organotypic proliferation assay (Figure 5f). Consistent with the IL33 blocking experiments, targeting IL33's receptor reduced proliferation in the brainstem which as expected had no effect on proliferation in the cortex.

Discussion
In this study, we sought to begin exploring the mechanism underlying the observed spatiotemporal pattern of pediatric HGGs. Two explanations have been put forward in the past. One explanation suggests that the locations of pHGGs coincide with waves of myelination in the childhood and adolescent brains and so reflect tumor intrinsic factors, while a second possibility suggests that this pattern is due to the microenvironment. These two explanations are not necessarily mutually exclusive.
Here we tested the potential of the brainstem surroundings to differentially impact DIPGs' survival and proliferation. As compared to the cortex, the brainstem shows a potential to better support the progression of HGGs of brainstem origin. This was demonstrated by the increased survival and proliferation of DIPG tumors in the brainstem. , and then comparing to the final area after 7 days. Images were obtained and the area measured, using ImageJ (Fisher's Combined Probability Test; SU-DIPGXIII n = 6, SU-DIPGXXV n = 6, BT869 n = 5. Within each trial, a Mann-Whitney U test was performed to obtain the p-values, n = 7 droplets. One tailed). (c) DIPG colony count per well, after a treatment period of 5-7 days with IL33 in comparison to control. Quantification was performed with Incucyte® S3 Live-Cell Analysis system (Fisher's Combined Probability Test; SU-DIPGXIII n = 11, SU-DIPGXXV n = 8, BT869 n = 6. Within each trial, a Mann-Whitney U test was performed to obtain the p-values, n = 6-8 wells. One tailed). (d) mIL33 neutralization trial in ex vivo slice cultures. Flow cytometry analysis of CellTrace dye staining of DIPG cells, injected with 1 μg/ml of mIL33 antibody or isotype control antibody. The cells were injected into two sides of the brainstem from a single P2 mouse brain, then slice cultures were prepared. After 72 h, the proliferation of DIPGs was measured (Wilcoxon's signed-rank test, one-tailed; SU-DIPGXIII n = 10, SU-DIPGXXV n = 10, BT869 n = 11, when n represents a single brain).
Our findings support the idea that at a young age, the microenvironment of the brainstem is better able to foster the progression of DIPG tumor cells. Interestingly, these differential effects were in most cases specific to DIPGs as compared to pediatric cortical HGG lines, suggesting a cell-intrinsic component of the tumor cells.
The pro-tumor activity of microglia and macrophages is well documented in adult gliomas. 23 As in other gliomas, microglia/macrophages account for a significant portion of the cellular mass of DIPGs. 33 Although the involvement of microglia in promoting DIPG propagation may not be surprising in itself, the differential ability of brainstem microglia as compared to cortical microglia to promote DIPG growth is striking. In recent years, there has been accumulating evidence indicating differences in microglial populations throughout different brain regions including anatomical cell number distribution, cell density and morphology, molecular signature, and functional differences. 6,[34][35][36][37][38] Here we show that microglia differentially support cancer cells based on their anatomical location, partly via IL33. Flow cytometry analysis of CellTrace dye staining of GFP + SU-DIPGXIII cells which underwent IL1RL1 knockdown using two separate shRNAs, 72-h postinjection into the cortex and brainstem of the same brain. For each brain, a parallel injection of control (plko) shRNA-treated cells was performed on the other side. Each point on the graph represents the percentage of GFP + cells which underwent division (at least 1), out of the entire GFP + cell population ± SEM (Wilcoxon's signed-rank test, one tailed; sh-IL1R-1, sh-IL1R-2 n = 8).