We present here a model for the H3K27M-altered DIPG TME that includes tumor tissue analogs (TTA) formed through 3D co-culture, with monitored size, cell number, and cellular composition. These TTAs are capable of self-assembling into tissue-like microstructures. The miniaturized 3D disease model for DIPG allows controlled manipulation of the microenvironment and therapeutic intervention in real-time as monitored by standardized imaging. Exploiting the 3D TTA to elucidate the mechanistic basis of microenvironment-induced changes in tumor cells by integrating proteomic and transcriptomic analysis provides insights into chemotherapy resistance and the molecular basis of stroma-induced synthetic lethality of immunotherapeutic approaches in DIPG. Such progress made toward an improved therapeutic strategy for DIPG can lead to neuro-oncologic advances of high importance and may provide therapeutic targets for validation in clinical trials.

All cell lines were maintained at 37°C and 5% CO2 as 2D or 3D cultures. Patient-derived non-adherent DIPG cell lines, SU-DIPG-6 (H3.3-K27M, female), SU-DIPG-13 (H3.3-K27M, female), SU-DIPG-17 (H3.3-K27M, male), SU-DIPG-4 (H3.1-K27M, female), were gifted by Dr. Michelle Monje's laboratory at Stanford University [32]. Cells were cultured in Tumor Stem Media (TSM) consisting of a 1:1 mixture of DMEM/F12 (Invitrogen, Carlsbad, CA, USA) and Neurobasal (-A) medium (Invitrogen) with 0.1 mg/ml Primocin (InvivoGen, #ant-pm-1, San Diego, CA, USA), and supplemented with B27(-A) (Invitrogen), human-βFGF (20 ng/mL), human-EGF (20 ng/mL), human PDGF-AA (20 ng/mL), human PDGF-BB (20 ng/mL) all from Shenandoah Biotechnology, Warminster, PA, USA and heparin (10 ng/mL) (StemCell Technologies, Vancouver, Canada). Patient-derived SF8628 cell line (H3.3-K27M, female) was purchased from EMD Millipore (SCC127) and maintained in DMEM-High Glucose (Sigma, D6546, St. Louis, MO, USA), 10% Fetal Bovine Serum (FBS, ATCC® 302020™, Manassas, VA, USA), 2 mM L-Glutamine (EMD Millipore Cat. No. TMS-002-C, Darmstadt, Germany) and 0.1 mg/ml Primocin.

Other pediatric brain tumor cell lines (Neuroblastoma, Medulloblastoma and Glioblastoma) were used to assess relative expression of the tumor specific antigen, GD2 in DIPG cells and were provided by the Asgharzadeh laboratory (CHLA). Patient-derived MYCN non-amplified human neuroblastoma cell lines CHLA-15 (female) and L-AN-6 (male) were established in the Seeger laboratory (CHLA) [33]. CHLA-15 cells were grown in IMDM (Gibco Cat. No. 12440053, Grand Island, New York, USA) with 20% FBS, 4mM L-glutamine, and 1X ITS (5 µg/mL insulin, 5 µg/mL transferrin, 5 ng/mL selenous acid, Cat. No. 41-400-045, Gibco). L-AN-6 cells were grown in RPMI-1640 plus 10% FBS, and 2mM L-glutamine. MYCN-driven 9464D neuroblastoma cell line derived from transgenic mice [34] was cultured in DMEM supplemented with 10% FBS, 1 mM sodium pyruvate, 2 µM L-glutamine, 1X MEM nonessential amino acids, and 0.1 mg/ml Primocin. D54 (adult glioblastoma; female) and UW-228-2 (SHH-activated pediatric medulloblastoma; female) cell lines were maintained in DMEM with 10% FBS and 0.1 mg/ml Primocin. SV40-immortalized human microglial cells HMC-3 (ATCC®CRL-3304) were cultured in EMEM (ATCC® 30-2003™) supplemented with 10% FBS and 0.1 mg/ml Primocin. SV40-immortalized human cerebral microvascular endothelial cells, HBEC-5i (ATCC®CRL-3245), were grown in DMEM:F12 (ATCC® 30-2006™) with 10% FBS and 40 µg/mL endothelial cell growth supplement (ECGS, EMD Millipore #02-102). All cell lines were authenticated by short tandem repeat DNA profiling (University of Arizona Genetic Core, Tucson, AZ) performed every two years, and tested for mycoplasma contamination three days before an experiment (Mycoalert Detection kit, Lonza, LT07-705, Rockville, MD, USA).

SPY650-DNA (Spirochrome AG, SC-501, Stein am Rhein, Switzerland) is a far-red (λ 652/654 nm), nontoxic, cell-permeable, and highly specific live cell DNA probe for fluorescence imaging of nucleus and DNA. A 1000x SPY650-DNA stock solution was reconstituted in 50 μL of anhydrous DMSO. Staining at this concentration allowed persistence of the dye in long term cultures, despite necessary media change and ongoing cell division. HBEC-5i microvascular endothelial cells at a 60-70% confluency were incubated overnight with staining solution of SPY650-DNA in a 1:3000 dilution of fresh cell culture medium as described. The single-cell suspension of labeled endothelial cells was then used for multicellular 3D co-cultures.

Lentivirus-based labeling reagents enabling co-expression of a nuclear-restricted green (tagGFP2) or orange (TagRFP) fluorescent protein, as well as puromycin driven by an EF-1α promoter were purchased from Sartorius (Cat. No. 4624 & 4771, Göttingen, Germany). Polyclonal DIPG cell lines expressing tagGFP2 (GFP), and HMC3 microglial cell line expressing tagRFP (RFP) were generated in-house as per the manufacturer's instructions. Briefly, cells were seeded at 30% confluency and were transduced 18-24 h later with lentivirus particles at an MOI (multiplicity of infection) of 3 using standard fresh medium supplemented with Polybrene® (Millipore Sigma, TR-1003-G, Burlington, MA, USA) at concentration of 8 μg/mL. The cultures were replaced with fresh medium and microscopically evaluated for expression of fluorescent proteins 48-72 h later. Fluorescent cells were enriched by puromycin (Gibco, A11138-03) selection (0.5 μg/mL) after 48-72 h, later followed by flow sorting. Each fluorescent cell line was checked for mycoplasma contamination and identity by STR analysis.

We describe a technique for self-assembly of monodispersed cells to generate tumor tissue analogs (TTA) of multiple cell types that can be precisely size-controlled and can influence the DIPG microenvironment. Multicellular 3D aggregates were cultured from a combination of (i) SU-DIPG-6 and SU-DIPG-13 cell lines, stably transduced to express nuclear restricted green (tagGFP2) fluorescent protein, (ii) human brain microglial cells (HMC-3), transduced to express red (tagRFP) fluorescent protein, and (iii) untransduced human cerebral microvascular endothelial cells (HBEC-5i), visualized as needed via staining with a fluorescent dye (SPY650) as described below. The 3D multicellular tissue-like aggregates are formed and maintained in the same serum-free medium supplemented with brain specific growth factors [35]. Clear flat or round bottom, low attachment 96-well plates (Corning, NY) were used in experimental studies with 15,000 to 45,000 cells/well for 96-well flat- and U-bottom plates or 60,000 cells/well for 8-well micro chamber slides (ibidi, Gräfelfing, Germany). Cell counts and viability were determined using Countess 3 automated Cell Counter (Invitrogen) before plating. The 3D TTA were imaged using a STELLARIS 5 confocal microscope (Leica Microsystems, Deerfield, IL, USA). Confocal z-stacks were acquired, maximum intensity projections were generated, and spatial measurements were performed using Leica LAS X software. Violin plots were generated with GraphPad Prism.

Time-lapse fluorescence and phase-contrast microscopy were performed using a Zeiss Axio Observer 7 live cell imaging microscope (Carl Zeiss Microimaging, Thornwood, NY, USA) with 5x or 10x objectives and a custom-stage incubator to house 3D culture samples in an 8-well micro chamber slide (ibidi, 80826 Fitchburg, WI, USA). Quantitative data were obtained from an average of 108 images per experimental condition at each timepoint (Nine fields of view were examined per condition, performed separately in four dishes, and with each experiment repeated a total of three times). Images were acquired every 30 min over a course of 48 h and each image sequence contained approximately 108 frames of 0.65 μm/pixel.

Images acquired with time lapse microscopy were analyzed using the Imaris 9.9.1 (Bitplane AG, Zurich, Switzerland) software with spot detection function to track migration of nuclear restricted GFP-expressing SU-DIPG cells alone and in co-culture with stromal cells. Spot size was set to 10 µm as an upper threshold for background signal filtration. Autoregressive motion was used for tracking and "maxDistance" was designated as 30 µm with a "Maximum Gap Size" of 3 frames between each track. The integrated "Edit Tracks" function in the Imaris software was used to correct for potentially fragmented and/or falsely connected tracks. The integrated "statistics" function was additionally used to extract data for track length, average speed, average displacement, and straightness of tumor cell trajectories. Cell tracks were graphically represented using MATLAB (Portola Valley, CA, USA).

IHC was performed on Formalin-fixed Paraffin-embedded 3D Tumor cell only cultures (T) and Tumor Tissue Analogs (TEM) using the automated Leica BOND RX platform. Briefly, the slides were heated at 60°C for 10 min and deparaffinized using Dewax (Leica Biosystems, Buffalo Grove, IL, USA) at 72°C for 30 s; Sequentially rinsed with three changes of 100% reagent alcohol followed by rinse in Bond Wash (3 min), and DI water (4 min). Antigen retrieval was done with ER1 buffer at 100°C for 20 min. Samples were then incubated in hydrogen peroxide for 10 min, and rinsed three times with Bond Wash (2 min each) again and incubated for 1 h with either of the following primary antibodies from Abcam : (a) Olig2 (1:125, ab109186), CD133 (1:1000, ab216323), GFAP (1:2000, ab7260), Ki-67 (1:800, ab15580), CD31 (1:100, ab9498), or IBA-1 (1:100, PA527436, Fisher Scientific); rinsed three times in Bond Wash, for 2 min each. The BOND Refine (DS9800; Leica BioSystems) detection system was used for visualization. Slides were then dehydrated, cleared, and coverslipped. Images were viewed using Aperio ImageScope software (Version v12.4.6.5003, Leica Biosystems, Milton Keynes, UK).

Histones were extracted by standard extraction techniques using the Active Motif Histone Extraction Minikit (40026). Histone ELISAs were conducted using the trimethyl K27 Elisa Kit (Active Motif, 53106) normalized to a H3K27me3 standard curve and total H3 protein. 4μg of total histone protein was used per antibody binding reaction in the ELISA. The experiment was repeated 3 times with three replicates per sample. The absorbance was read at 450 nm and 655 nm that served as the reference wavelength for background subtraction. The absorbance values of the unknown samples were then used to interpolate their protein concentrations from the standard curve generated on graphpad prism.

DIPG cells only in 3D cultures and DIPG cells in co-culture with endothelial cells and microglia were collected at day 5, 10 and 15 for proteomic analysis. The cultures were maintained for 5 days in the same medium, after which the medium in each sample was replaced every 2 days. Total protein from each sample (supernatant and 3D cell cultures) was reduced, alkylated, and purified by chloroform/methanol extraction prior to digestion with sequencing grade modified porcine trypsin (Promega, Madison, WI, USA). Tryptic peptides were then separated by reverse phase XSelect CSH C18 2.5 um resin (Waters, Milford, MA, USA) on an in-line 150 × 0.075 mm column using an UltiMate 3000 RSLCnano system (Thermo Fisher Scientific, Waltham, MA USA). Peptides were eluted using a 90 min gradient from 98:2 to 65:35 buffer A:B ratio (Buffer A = 0.1% formic acid, 0.5% acetonitrile; Buffer B = 0.1% formic acid, 99.9% acetonitrile). Eluted peptides were ionized by electrospray (2.4 kV) followed by mass spectrometric analysis on an Orbitrap Eclipse Tribrid mass spectrometer (Thermo Fisher Scientific). Mass spectrometry (MS) data were acquired using a Fourier transform mass spectrometry analyzer in profile mode of full width at half maximum (FWHM) resolution of 120,000 over a range of 375 to 1400 m/z with advanced peak determination. Following HCD (high energy collisional dissociation) activation, tandem mass spectrometry (MS/MS) data were acquired using the ion trap analyzer in centroid mode and normal mass range with a normalized collision energy of 30%.

Proteins were identified by query of the Uniprot Homo Sapiens Database (June 2021) using MaxQuant (version 2.0.3.0, Max Planck Institute, Martinsried, Germany) with a parent ion tolerance of 3 ppm and a fragment ion tolerance of 0.5 Da. Scaffold Q + S (Proteome Software, Portland, OR, USA) was used to verify MS/MS based peptide and protein identifications. Protein identifications were accepted if they could be recognized with less than 1.0% false discovery and contained at least 2 identified peptides. Protein probabilities were assigned by the edge features. Protein MS1 iBAQ intensity values were assessed for quality using ProteiNorm [36]. The data were normalized using variance stabilizing normalization (VSN) and statistical analysis was performed using Linear Models for Microarray Data (limma) with empirical Bayes (eBayes) smoothing of the standard errors [37]. Proteins with an FDR adjusted p-value < 0.05 and a fold change > 2 were considered significant.

Cells were detached using trypsin (0.05%) with EDTA, washed in complete media and pelleted by centrifugation. They were counted and 5 ×105 cells were washed once in cold FACS buffer (1X PBS supplemented with 2% heat-inactivated FBS and 2 mM EDTA and the endocytosis inhibitor NaN) followed by incubation with anti-GD2-APC antibody (Clone 14G2a, Isotype mouse IgG2a, BD Biosciences, Franklin Lakes, NJ, USA) or isotype-matched control mAb of irrelevant specificity (murine IgG2a-APC; BD Biosciences) for 30 min in a 4 °C ice water bath in the dark. Cells were washed twice in FACS buffer and transferred to filter top tubes with addition of DAPI live/dead stain (final concentration 1 μg/ ml; Sigma-Aldrich). Data were acquired using a BD FACS Aria I flow cytometer, acquiring 10,000 events from the live (DAPI negative) singlet gate for each cell line. Exported Flow Cytometry Standard (fcs) files were then analyzed using FlowJo_v10 software (Ashland, OR, USA) and geometric mean fluorescence intensity determined for every sample. The stain index, which is the mean fluorescence intensity (MFI) of the positive and negative populations divided by two times the standard deviation (SD) of the negative population was calculated for each cell line and graphically represented. Stain Index (∆) = MFI (positive) - MFI (negative) / 2 X SD (negative)

Human whole blood was collected from four IRB-consented healthy donors (CHLA-21-00319) by venipuncture in EDTA-sprayed lavender top BD Vacutainer® collection tubes. Samples were obtained after informed consent, and in accordance with an institutional review board approved protocol. Cells were separated from plasma by centrifugation for 15 min at 1500 x g using a refrigerated centrifuge. The resulting supernatant, designated as plasma, was aliquoted and stored at -80 °C.

Live cell imaging was performed using IncuCyte S3 (Sartorious). SU-DIPG-6 and/or SU-DIPG-13 cells were seeded alone and in co-culture with stromal cells and scanned on 96-well flat bottom

(CytoOne, CC7682-7596, USA Scientific, Ocala, FL, USA) or U-bottom (Corning, 7007, Corning, NY, USA) plates in the cell culture incubator over time. Each well was scanned using a 10X objective lens in 4 randomly selected positions or with a 4X objective at a single position, repeated at indicated intervals in the experiment. Imaging included high-definition phase contrast and epifluorescence microscopy with 483/506 and 555/584 nm filter sets to detect green (tagGFP2, DIPG) and orange (tagRFP, microglia) fluorescence, respectively. Image processing and cell counting were performed using IncuCyte software (Sartorius). There were 3 to 6 replicates for each sample set per experiment (r = 3-6) and each experiment was repeated 3 times (n = 3).

Suspension of nuclear restricted GFP-expressing SU-DIPG-6 cells at a cell density of 15,000/well and 45,000/well and suspension of SU-DIPG-6 cells in equal proportion with stromal endothelial cells and nuclear restricted tagRFP-expressing microglial cells at a cell density of 45,000 cells/ well were plated in a 96-well flat bottom plate (CytoOne, US Scientific) overnight. Cell counts and viability were determined using Countess 3 automated Cell Counter (Invitrogen) before plating. Targeting of GD2-expressing DIPG cells alone and in co-culture was assessed with Incucyte S3 live-image fluorescent microscopy system (Sartorius) housed in a cell culture incubator. Cytotoxicity assays were performed utilizing dinutuximab (Unituxin™, United Therapeutics Oncology, Silver Spring, MD, USA) and Plasma collected from human donors. Dinutuximab stocks and human plasma from different donors were diluted in TSM medium with growth factors before addition to wells. Final concentrations in a final well volume of 200 μl were in the following range: dinutuximab (0.5 μg/ml); Human Plasma (10%). Cultures with and without dinutuximab and Heat inactivated human plasma served as control. Cell survival for all groups was normalized to cell survival in cultures that were untreated.

GFP-expressing SU-DIPG-6 and SU-DIPG-13 cells were flow sorted from 3D cultures of tumor cells only and TTA comprised of tumor cells with microglia and endothelial cells on day 5. The SMART-Seq V4 Ultra Low RNA Input Kit for Sequencing (Takara Bio, Kusatsu, Shiga, Japan) was used for reverse transcription and generation of double stranded cDNA for subsequent library preparation using the Nextera XT Library Preparation kit (Illumina, San Diego, CA, USA). An input of 10 ng RNA was used for oligo(dT)-primed reverse transcription, followed by cDNA amplification and cleanup. Quantification of cDNA was performed using Qubit 4 fluorometer (Thermo Fisher Scientific). cDNA normalized to 80 pg/μl was fragmented and sequencing primers added simultaneously. A limiting-cycle PCR added Index 1 (i7) adapters, Index 2 (i5) adapters, and sequences required for cluster formation on the sequencing flow cell. Indexed libraries were pooled. Enriched libraries were verified using 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA) and quantified via Qubit 4 fluorometer (Thermo Fisher Scientific). Libraries were sequenced on a NovaSeq 6000 (Illumina) using 1×75 bp read length and coverage of over 30 M reads/sample. Raw sequence data was deposited to NCBI GEO (GSE222891).

Sequences were aligned to Ensembl genes version 107 corresponding to the Genome Research Consortium Homo sapiens build number 38 (GRCh38.p13) using the STAR aligner with 'GeneCounts' output [38]. Reads per kilobase of the transcript, per million mapped reads (RPKM) values were determined with the R/Bioconductor software package 'edgeR' [39, 40] and differential gene expression was determined using the software, 'limma-voom' [41]. Principle component plots and dendrograms of hierarchical clustering were generated from RPKM using the R software 'ggdendro' [42]. Volcano plots and heatmaps were generated using RPKM values and the R software 'gplots' [43] of differentially expressed genes. Gene set enrichment analysis was performed on differentially expressed genes with fold change >1.5 or <-1.5 using the Gene Ontology resource:'GOstats' [44], and target gene/protein networks were constructed using open source String database [45] and Cytoscape [46] softwares. Connectivity map (CMap) analysis was performed for high scoring genes (https://clue.io) to search for potential small molecule compounds. Genes found to be highly significant in the unsupervised analysis were input into The Broad Institute CMap (clue.io) for gene set testing against the latest version of the L1000 dataset (CMAP LINCS 2020) containing over 3 million gene expression profiles. The output is a list of "perturbagens" (chemicals, and biologics) that closely match the input (+ raw_cs) and the inverse of the input (- raw_cs).

Parametric t-test at significance level (α = 0.05) was used to assess significant differential change on log-transformed values of the tracking parameters including mean speed, track length, displacement, and straightness between tumor-cell only culture (T) and co-culture (TEM) conditions with the R Statistical Package (Version 4.3.2). Box-whisker plots of the log-transformed values representing the median, 25% percentile, and 75% percentile were presented for visualization of the differential change.