Secretome analysis of patient-derived GBM tumor spheres identifies midkine as a potent therapeutic target

Glioblastoma (GBM) is the most lethal primary brain tumor with few treatment options. The survival of glioma-initiating cells (GICs) is one of the major factors contributing to treatment failure. GICs frequently produce and respond to their own growth factors that support cell proliferation and survival. In this study, we aimed to identify critical autocrine factors mediating GIC survival and to evaluate the anti-GBM effect of antagonizing these factors. Proteomic analysis was performed using conditioned media from two different patient-derived GBM tumor spheres under a growth factor-depleted status. Then, the antitumor effects of inhibiting an identified autocrine factor were evaluated by bioinformatic analysis and molecular validation. Proteins secreted by sphere-forming GICs promote cell proliferation/survival and detoxify reactive oxygen species (ROS). Among these proteins, we focused on midkine (MDK) as a clinically significant and pathologically relevant autocrine factor. Antagonizing MDK reduced the survival of GBM tumor spheres through the promotion of cell cycle arrest and the consequent apoptotic cell death caused by oxidative stress-induced DNA damage. We also identified PCBP4, a novel molecular predictor of resistance to anti-MDK treatment. Collectively, our results indicate that MDK inhibition is an important therapeutic option by suppressing GIC survival through the induction of ROS-mediated cell cycle arrest and apoptosis.

The acquired MS/MS spectra were compared against various databases using SEQUEST in Proteome Discoverer 1.4 (Thermofisher) against the human UniProtKB database (released in April 2014) supplemented with experimentally validated B27 and N2 contaminant sequences. The B27 and N2 contaminant sequences were mammal UniProt sequences identified earlier by LC-MS/MS analysis of B27 and N2 supplements in growth medium. 3 Two trypsin missed cleavages, fixed modification of carbamidomethylation at cysteine (+ 57.02 Da) and variable modification of oxidation at methionine (+ 15.99 Da) were allowed.
Mass tolerances for MS/MS and MS were set to ± 0.5 and ± 2 Da, respectively. Peptide and protein assignment and validation (FDR >1%) were performed.
The biological functions of identified proteins were analyzed using DAVID functional classification tool (https://david.ncifcrf.gov/). The Functional networks based on the biological processes were analyzed using the ClueGO Cytoscape plugin. 4
For immunofluorescence analysis, tissues or cells were fixed with 4% paraformaldehyde for 20 min. Tissues were further incubated in 30% sucrose overnight and then embedded in optimal cutting temperature (OCT) compound and crysectioning at 10 µm. After blocking and permeabilization with 0.5% triton X-100 and 5% normal goat or donkey serum in PBS, sections were stained with target antibodies in PBS with 0.1% bovine serum albumin (BSA), followed by incubation with fluorescein-conjugated secondary antibodies were used for visualization. 4′, 6′-diamidine-2′-phenylindole (DAPI) staining was used to control as an intracellular localization. Images were obtained using LSM700 Confocal Laser Scanning Microscope system (Zeiss) or Operetta High Content Imaging System (PerkinElmer).

Cell proliferation assay
For evaluating short-term proliferation, 500 single cells were plated per well in a 384-well plate. Next, cells were treated with an anti-MDK antibody (Santa Cruz Biotechnology) or recombinant MDK (Sigma-Aldrich) and further incubated at 37℃ for 4 days. Cell viability was analyzed using an adenosine triphosphate (ATP) monitoring system based on firefly luciferase (ATPLite™ 1step, PerkinElmer), and luminescence was measured using an EnVIsion Mutilabel plate reader (PerkinElmer). 5 Relative cell viability was obtained by normalization to luminescence values of vehicle or control IgG treated cells.
Anti-EdU fluorescence assay was carried out according to the manufacturer's recommendation (BCK-EdU594, Sigma-Aldrich). At first, EdU solution (50 µM) was added to each well, and cells were incubated at 37℃ for 24 h. Culture medium was removed, and cells were fixed with 4% formaldehyde in PBS. Cells were washed 5 times with 3% BSA in PBS solution and permeabilized with 0.5% triton X-100 in PBS, followed by adding reaction cocktail. The nuclei were counter-stained with Hoechst 33342 solution. Fluorescence images were captured with Operetta High-Content Imaging System (PerkinElmer).

Sphere formation analysis and Limiting dilution assay (LDA)
For sphere formation assay, single cells isolated from primary tumor spheres were seeded (100 cells per well) in a 96-well plate. After 10-14 days, the tumor spheres were captured using Operetta High Content Imaging System (PerkinElmer), and the number of spheres, roundness, sphere area were analyzed using Harmony High Content Imaging and Analysis Software (PerkinElmer). Cell aggregates of size with 25-75 percentile were selected and those of roundness less than 0.9 were excluded at the analyses. LDA was also fulfilled in 96 well plates and cells were seeded at a range of 1-200 cells per well (12 wells per group).
After 10-14 days, wells without spheres were measured and analyzed. Statistically significance was analyzed using ELDA software (Walter+Eliza Hall Bioinformatics). 6

Lentivirus production and transduction
MDK knockdown shRNA lentiviral clones (TRCN0000331210 and TRCN0000331252 for shMDK-1 and shMDK-2, respectively) and shRNA lentiviral clones for PCBP4 knockdown (TRCN0000232287 and TRCN0000232288) were purchased from Sigma-Aldrich, and a pLenti-PCBP4 expression vector was obtained from Abm. Lentiviruses were produced in 293FT cells with packaging mix (ViraPower Lentiviral Expression Systems, Thermofisher) and concentrated by ultracentrifugation. Viral titers were determined by the serial dilution method. Lentiviral particles were transduced into cells, and stable transfectants were selected by incubation with puromycin (1~2 ng/ml).

Western blot assay
GBM tumor-spheres were washed in cold PBS, followed by being harvested in passive lysis buffer (Promega) and a protease and phosphatase inhibitor cocktail added (Roche).

Orthotopic GBM xenograft models
All animal experiments were approved by the Institutional Review Board of the SMC and performed according to the guidelines of the Animal Use and Care Committees. GBM cells were dissociated, 1X10 4 cells per each mouse were resuspended in 5 μl of Hank's balanced salt solution (HBSS, Thermo Scientific), followed by stereotactically (2 mm left and 1 mm anterior of the bregma, 2 mm deep from the dura) injected into the brains of Balb/c nude mice (6~8-week-old female, Orient Bio Inc.). Kaplan meier survival (GraphPad Prism v5.03) and immunohistochemical analysis were performed under blinded inspection.

Bioinformatical analysis
Gene Set Enrichment Analysis (GSEA) was conducted by using the Java with the Molecular Signatures Database (MSigDB; version 5.1) C2 gene set collection. 10 P values about enrichment of genes were estimated by 20,000 gene set permutations. Single sample gene set enrichment analysis (ssGSEA) was used to produce gene set activation scores of each sample using Genepattern (http://software.broadinstitute.org/cancer/software/genepattern).
Method for normalizing gene expression data is rank. 11 For identifying differentially expressed genes from gene expression values, two groups that were consisted by control and neutralized samples were compared with using package DEGseq of R. P values were estimated by Student's t test for paired samples and accomplish multiple correction tests using the Benjamini-Hochberg method. 12 A radar plot representing pathway enrichment analysis (GO BP on DAVID) of differentially expressed protein (DEP) data was generated using R package 'radarchart'.
For modelling genetic association of anti-MDK antibody sensitivity, we trained the standard elastic net regression (R package: glmnet_2.0-5), combining drug sensitivities and preselected mRNA expression as the input features. 13 Given an anti-MDK antibody and N GBM tumor-spheres, we used y = (y1, y2, … yn) represents drug sensitivity of n cells, and xi = (xi1,..., xip), (i = 1,…, p) indicates p features of ith cell, the elastic net regression is to solve the following optimization problem:  controls the relative balance of the L1 and L2 penalty terms, while  controls the overall penalty level of the regularized term. Similar with ,  was optimized by 10-fold cross validations, but using all features.  was screened using 50 values with   [0,1] equally spaced. For each model fitting, we used the function cv.glmnet, with its optimized  value provided from the function.
After parameter optimization, we adopt bootstrapping strategy for 100 times to obtain a robust evaluation of the predictive power of features. During each bootstrapping, we randomly select 80% of GBM tumor-spheres with 80% of the features to fit the elastic net with above optimized . For each feature, the time of appearances (non-zero fitting coefficient) out of the 100 bootstrappings, together with the average of its non-zero weights were used as its final assessment of predictive ability.

Survival Analysis of GBM patient
We assessed the association between MDK expression and overall survival of IDH1 wild type 129 samples with data from cBioPortal (http://www.cbioportal.org). Kaplan-meier survival analysis is performed using R software version 3.4.2 (http://www.R-project.org) and statistically tested using Log-rank test. Cut-off values for optimal prediction of survival were obtained with the R software package 'maxstat' (REF : Hothorn T, Lausen B (2002) Maximally Selected Rank Statistics in R. R News 2/1: 3-5)

Protein profiling
Protein array data were obtained using Phospho Explorer Antibody Array (# PEX100, Full Moon Biosystems). The antibody array experiment was performed by Full Moon Biosystems, according to their established protocols. 14 In brief, the slides were scanned on an Axon GenePix array scanner, and the images were analysed with GenePix Pro 6.0 (Molecular Devices

Cell cycle analysis
For analyzing cell cycle, single cell dissociates from GBM tumor-spheres were fixed with 100% ethanol and incubated at 4℃ overnight. Cells were stained with propidium iodide (PI, (Sigma-Aldrich) and analyzed using Flow Cytometry (FACS Aria, BD Biosciences). The flow cytometry data were interpreted with Flow Jo software (ver. 7.6, Treestar Inc).

Apoptosis assay
For Annexin V staining, single cells were washed, followed by resuspended in a Annexin V Binding Buffer (BD Pharmingen). Cells were then stained with Annexin V-APC solution (eBioscience). After adding PI (Sigma-Aldrich), Annexin V-PI positive cells were measured using FACS Aria (BD Bioscience) and data were interpreted with Flow Jo software (ver. 7.6, Treestar Inc). Caspase 3/7 activities were measured using ApoTox-Glo Triplex assay (#G6320, Promega) or CellEvent™ Caspase-3/7 Green Detection Reagent (Thermofisher) as manufacturer's instructions. The luminescence or fluorescence intensities were obtained using EnVIsion Mutilabel plate reader (PerkinElmer). The apoptosis array analysis was performed using the Proteome Profiler Human Apoptosis Array kit (ARY009, R & D Biosystems) according to the manufacturer's instruction and array image were analyzed using the Image J software (https://imagej.nih.gov/ij/index.html).