Retrospective study on patients
The peri-operative cohort (Fig. 1) was a local prospective cohort composed by 38 adults (≥ 18 years) IDHwt GBM patients included at initial diagnosis between June 2016 and October 2017 at La Timone Hospital (Assistance Publique - Hôpitaux de Marseille, AP-HM, France). For these patients, the blood samples were collected before and 48h after surgical resection. A concomitant MRI was performed at the time of diagnosis including T1, T1 with gadolinium, T2 and FLAIR sequences at least, before and within 48 h after surgical resection.
The following data were recorded: age, gender, type of surgery (gross total or partial resection), Karnofsky Performance Status (KPS), oncological treatment, clinical symptom, steroid dose, and MRI characteristics.
All samples were stored in the AP-HM Biological Resource Center (CRB BB-0033-00097). All patient data were obtained according to a protocol approved by the local institutional review board and ethical committee (PADS 20–343). The present study was conducted in accordance with the declaration of Helsinki.
Characterization of the post-surgical microenvironment in a mouse model
The in vivo experiments reported in this work have been approved by the institution’s Animal Care and Use Committee (CE71, Aix-Marseille Université, reference n° 22185) and performed following the French national regulation guidelines in accordance with EU Directive 2010/63/EU. Mice were housed in enriched cages placed in a temperature- and hygrometry-controlled room, had free access to water and food and were monitored daily. The mice were immunocompetent seven-week-old female C57BL/6JOlaHsd mice (for nuclear imaging, immunophenotyping, brain clearing and efficacy studies; Envigo, Gannat, France) or transgenic LysM-EGFP//CD11c-EYFP reporter mice (for biphotonic imaging and immunohistochemistry) 25. The animals were sacrificed at the end of the study (e.g. long-term survivors) or when they reached the end-points (≥ 20% body weight loss or 10% body weight loss plus clinical signs of distress e.g., paralysis, arched back, or lack of movement for glioma-bearing animals).
Orthotopic glioma mouse model
For tumor grafting, animals were anesthetized by ketamine/xylazine (100 and 10 mg/kg intraperitoneal injection, respectively) and fixed in a stereotactic frame. The skin surface on the head was disinfected by application of an antiseptic solution (Vétédine® solution, Vetoquinol, Lure, France) and the hair was removed. Lidocaine (10 mg/mL; Aguettant, Lyon, France) was injected subcutaneously at the site of incision, and the eyes were protected with an ophthalmic gel (Ocry-gel, TVM lab, Lempdes, France). To hydrate the animal, 1000 µL of physiological solution (0.9% Sodium Chloride; Aguettant, France) were injected subcutaneously in the flank. An incision was made along the midline and a burr hole was drilled into the skull at the parietal lobe (2 mm posterior and lateral to the bregma) using a high-speed drill (Tack life Tools, New York, USA; 0.8 mm diameter round end engraving burrs: Dremel, Breda, The Netherlands). A 10 µL 26 s gauge syringe with cemented 51 mm needle (Hamilton, Rungis, France) was used to inject 1×105 GL261-DsRed glioma cells - cultured as previously reported25 -- in the cortex (1 mm deep from the outer border of the brain), using an automatic pump device at a speed of 0.2 µL/min. The wound was then closed using a tissue adhesive glue (3M Vetbond®, Sergy-Pontoise, France) and the animals recovered under an infrared heating lamp.
Tumor resection and cranial window implantation
Fourteen days post-grafting, animals were anesthetized with ketamine/xylazine (100 and 10 mg/kg intraperitoneal injection, respectively), fixed again in a stereotactic frame and prepared for surgery as previously described. With scissors, the skin in the midline along the previous surgical scar to expose the skull was cut. The periosteum was removed with fine tweezers to reveal the bregma and ensure adherence of the cranial window. Then, a 5 × 5 mm circular cranial window was defined in the left parietal bone around the previous injection hole to expose the brain. The skull was gently lifted and removed passing the tweezers between the brain and skull to avoid removal of the tumor attached to the skull. The presence of the tumor was confirmed by fluorescent microscopy.
To resect the tumor, the biopsy punch technique previously developed and validated by the authors was used 32. Briefly, a biopsy punch (2 mm Ø, Kai Medical, Germany) was placed around the injection hole and inserted slowly 2 mm deep, followed by gentle twist to cut the brain/tumor tissue. A Pasteur pipette connected to a vacuum pump was used to remove the explant and build up blood. Hemostasis was achieved inserting an absorbable hemostatic triangle (Sugi® Sponge Points Kettenbach, Questalpha GmbH, Germany) in the formed cavity. The dural cranial window was covered using a round glass coverslip (5 mm Ø, 0.15 ± 0.02 mm thickness; Warner Instruments, USA) glued using histo-compatible acrylic glue (Cyanolit, AC Marca, France) and then sealed on the adjacent bone and fixed to the skull by dental cement (Unifast Trad™, GC America, USA). Control animals either received the cranial window implantation at the day of cells grafting (cranial window control) or at day 14 post-grafting (unresected control) but did not receive the biopsy punch tumor debulking. The presence or absence of tumor post-cranial window implantation and/or post-surgery was observed by fluorescent microscopy.
Blood immunophenotyping by flow cytometry
To evaluate the systemic immune response following surgery, complex cellular phenotyping was performed on blood cells the day before surgery and cranial implantation and 3, 7 and 10 days later. Animals were anesthetized with 1.5% vol% isoflurane (IsoVet®, Laboratoire Osalia, Paris, France) and approximately 120 µl of blood was collected by retrorbitary sampling using K2 EDTA pre-coated capillaries for microhematocrit (Vitrex®, Vitrex Medical, VWR, France) and stored at room temperature in Eppendorf tubes.
For erythrocytes lysis, lysis solution (eBioscience™ 1X RBC Lysis Buffer, Invitrogen, USA) was added on each blood sample, mixed and stored at room temperature for 5 min. Then, the reaction was stopped by adding a buffer solution constisting in Dulbecco’s Phosphate buffered saline (DPBS 1X no calcium no magnesium, GibcoTM, Thermo Fisher Scientific, France) with 0.5% Bovine Serum Albumin (BSA Fraction V, Eurobio scientific, France) and 2 mM EDTA (Invitrogen, Life Technologies, USA) in each tube and centrifuged (1200 rpm, 4°C for 10 min). For antibody staining, supernatant was removed and an appropriate volume of BD Horizon™ Brilliant Stain Buffer (BD Biosciences, France) were added on the pellet. Appropriate dilutions of the antibodies (Table S1) were prepared in buffer solution and the antibody mix was added on the cell suspension and incubated in the dark at 4°C for 20 minutes. Then, the cells were washed, pelleted by centrifugation and appropriately suspended in buffer solution for flow cytometry acquisition (Cytoflex LX instrument, Beckman Coulter, France) at the AMUTICYT core facility of Aix-Marseille University. Data were acquired using Cytexpert 2.2 software. Dimensionality reduction was performed using opt-SNE in OMIQ after data cleanup on manually gated viable singlets leukocytes CD45+. The 10 channels were selected for running the algorithm: CD45, Ly6G, CD161, Ly6C, CD357, MHCII, CD11b, F Siglec, CD8 and CD11c. The advances settings were setup using default mode of the OMIQ software (Max iterations 1000, Perplexity 30). Dimensional reduction was completed after 560 iterations. Opt-SNE representation show manually selected cell population using the gating strategy illustrated in Fig. 2C. Thus, to validate the manual selection of immune cell cluster, FlowSOM algorithm using Elbow metaclustering was done on the 10 channels and optsne_1 and optsne_2.
Biphotonic imaging studies
Images acquisition was conducted at day 7 and 14 post-surgery and/or cranial window establishment as previously described 25. Prior to each imaging session, mice were anesthetized by ketamine/xylazine (100 and 10 mg/kg intraperitoneal injection, respectively) and injected intravenously with 100 µL of a quantum dots (QDots) solution (Qtracker™ 705 Vascular Labels, ThermoFisher, 6 µg/100 µL in phosphate saline buffer (PBS, Sigma Aldrich) and positioned on a stereotaxic frame allowing movements in the three-directions.
For the acquisitions, it was used a Zeiss LSM 780-MP 2-photon microscope home modified to allow animal positioning below the 20× water immersion objective (1.0 NA) and coupled to a femtosecond pulsed infrared tunable laser (Chameleon Ultra 2, Coherent). Images were acquired using an excitation wavelength tuned at 920 nm to excite all fluorophores simultaneously. Signals were epicollected and separated by dichroic mirrors and filters on five independent non-descanned detectors. Gains and offsets were identical for all the detectors, except for the red channels whose gain was reduced by 30% to compensate for the strong expression of DsRed in tumor cells. Images were acquired below the dura matter over a depth of 500 µm using 10 µm steps. Laser power was linearly increased with depth. Z-stack images were acquired as mosaics (stitching mode) to cover the whole tumor surface.
For the biphoton images analysis, only horizontal plans were considered. Spectral unmixing was first applied to raw 2-photon images (Zen software) and the mean cell density (number of cells/mm3) was calculated over tumoral volume using Imaris microscopy image analysis software v10.0 (Bitplane, Oxford Instruments). We exploited the different Imaris imaging analysis options to perform all the further analyses. The volume of each tumor was defined by creating a 3D mask using the surface tool. Only cells present in the tumor mask were segmented according to the local intensity contrast. Cells found in blood vessels were easily excluded thanks to a 3D mask associated to the vasculature. Hence, to access the different cell densities, the reported number of LysM-EGFP+, CD11c-EYFP+ and LysM-EGFP+/CD11c-EYFP+ cells was reported over the resection cavity or the tumor volume. Colocalization analyses were performed to identify LysM-EGFP+/CD11c-EYFP+ double-labeled cells. Morphology and area of individual cells were analyzed with the sphericity score. To measure distance from vessels and from the resection cavity borders, each cell was dotted with spot tool according to the local intensity contrast and shortest distance from vessels or resection cavity borders was analyzed.
Nuclear imaging studies
Single Photon Emission Computed Tomography (SPECT) Imaging
The BBB opening was evaluated by SPECT/CT imaging using [99mTc]Tc-DTPA as radiotracer at day 13 post-grafting (day prior to surgery, controls) and 1, 2, 3, 7, and 10 days post-surgery. The controls were all tumor bearing animals without or with cranial window implantation on the day of cells grafting.
[99mTc]Tc-DTPA was administered via an intravenous injection of 50 µL of radioactive tracer (30 MBq) in isotonic and pyrogen-free solution using an insulin syringe (27G)34. Thirty minutes after injection, the SPECT/CT acquisition was done for 20 min under anesthesia with 1.5% vol% isoflurane (IsoVet®, Laboratoire Osalia, Paris, France) using a NanoSPECT/CTplus® camera and the Nucline® 1.02 acquisition software (Mediso Medical Imaging System Ltd., Budapest, Hungary). SPECT and CT DICOM files were fused for reconstruction and image processing was carried out with VivoQuant® 3.5 and InvivoScope® 2.00 reconstruction software (InviCRO, Boston, MA, USA) to assess tracer uptake in the brain.
Characterization of the microenvironment of primary vs recurrent tumors
3D analysis of the brain and tumor microenvironment
Fourteen days post-surgery and/or cranial window implantation, animals were deeply anesthetized by xylazine/ketamine (100 and 10 mg/kg intraperitoneal injection, respectively) and the presence of the tumors was evaluated by fluorescence microscopy. Then, they were perfused by cardiac injection of 10 mL of DPBS 1X and then 10 mL of Paraformaldehyde (PFA 4% aqueous solution, methanol free; Electron Microscopy Sciences, UK). Brains were extracted and placed in PFA 4% for 24h at 4°C, and then stored in PBS containing 0.02% sodium azide (Sigma Aldrich) at 4°C until further use.
Hemi brains were processed following the iDISCO+ protocol as previously described35 and routinely performed in our lab36. Shortly, brains were dehydrated in successive methanol baths, put in a solution of 66% dichloromethane (Sigma Aldrich) and bleached with 5% H2O2 (Sigma Aldrich). Then samples were re-hydrated in sequential baths of methanol and permeabilized (PBS/0.2% TritonX-100 (Euromedex)/20% DMSO (Sigma Aldrich)/0.3M glycine (Euromedex)) at 37°C for 2.5 days. Samples were then incubated in a blocking solution (PBS/0.2% TritonX-100/10% DMSO/3% Donkey Serum (Jackson Immunoresearch)) at 37°C for 4.5 days and incubated for 5 days in primary antibodies solution (Table S2; PBS-Tween 0.2% with heparin 10 mg/mL (Sigma Aldrich)/5% DMSO/3% donkey serum) at 37°C for 5 days. After washing, samples were incubated with secondary antibodies (Table S2) in PBS-Tween 0.2% with heparin 10 mg/mL/3% donkey serum at 37°C for 5 days. Samples were finally dehydrated in successive baths of methanol and cleared in a BABB solution (33% Benzyl alcohol, Sigma Aldrich, France; 66% Benzyl benzoate, Fischer Scientific, France) for 24 h and then with BABB alone until light sheet acquisition.
Samples were imaged in MACS Imaging solution (Miltenyi Biotec) in sagittal orientation with the Ultramicroscope BlazeTM (Miltenyi Biotec) equipped with a 4.2 Megapixel sCMOS camera and 1.1×/4×/12× objectives. A numerical aperture of 0.1/0.35/0.53 were used with a fixed 4-sources illumination. The microscope is equipped with LED lasers (488 nm, 561 nm and 639 nm). Emission filters used were 525/50, 595/40, 680/30. The samples were scanned with sheet of 5 µm of thickness (for 1× and 4× objectives). A horizontal dynamic correction was applied when mosaics were performed with the 4× objective.
Images were analyzed as 3D projections, performed using Imaris software v10.0 (Bitplane) as previously described 36. The volume of each tumor was defined by creating a 3D mask using the surface tool. Only cells present in the tumor mask were segmented according to the local intensity contrast. For brains analysis, to access cell density, the number of CD45+ and TMEM119+ cells were reported to the tumor volume. To measure distance from tumor center and vessels, each cell was dotted with the spot tool according to the local intensity contrast and shortest distance from tumor center or vessels, reconstructed in 3D with the surface tool, was analyzed.
Tumor immunophenotyping by flow cytometry
To evaluate the local brain immune landscape of recurrent tumors, complex cellular phenotyping was performed on tumors extracted at day 10–14 post-surgery and/or cranial window implantation in wild-type mice bearing tumors. Animals were deeply anesthetized by xylazine/ketamine (100 and 10 mg/kg intraperitoneal injection, respectively) and were perfused by cardiac injection of 20 mL of PBS 1X. Brains were extracted and the cerebral hemisphere containing the tumor was excised using fine tweezers. The tumor and surrounding healthy cerebral cortex were dissociated in gentleMACS™ Tubes (Miltenyi Biotec, Germany) on a GentleMACS Octo Dissociator (Miltenyi Biotec) using a Brain Tumor Dissociation Kit (Miltenyi Biotec). The suspension was then filtered using 70 µm MACS® Smart Strainers (Miltenyi Biotec). For erythrocytes lysis, lysis solution was added on each sample and stored at room temperature for 3 min. Then, the reaction was stopped by dilution in buffer solution (DPBS with 0.5% BSA and 2 mM EDTA). The cell pellet was then suspended in buffer solution containing FcR blocking reagent mouse (Miltenyi Biotec, Germany) and anti-CD45 microbeads mouse (Miltenyi Biotec, Germany) and incubated 15 min at 4°C. For magnetic cell sorting, LS columns (Miltenyi Biotec, Germany) were used to separate CD45+ from total suspension. For antibody staining, an appropriate volume of BD Horizon™ Brilliant Stain Buffer (BD Biosciences, France) was added on the pellet of CD45+ cells. Appropriate dilutions of the antibodies (Table S3) were prepared in buffer solution and the antibody mix was added on the cell suspension and incubated in the dark at 4°C for 20 minutes. Then, the cells were rinsed, pelleted and appropriately suspended in buffer solution for flow cytometry acquisition with Cytoflex LX instrument. Data was extracted using Cytexpert 2.2 software and analyzed using Kaluza analysis software. The gating strategy used to extract and separate immune cells is illustrated in Figure S4 indicating expression of fluorophores among cell subsets.
Exploiting the post-surgical microenvironment as a therapeutic target
Anti-tumor efficacy studies
To evaluate the anti-cancer efficacy of our proposed combination regimens in clinically relevant conditions, we used a preclinical model previously developed and validated in our lab to resect GBM tumors 32. In this model, animals were grafted with 1×105 GL261-DsRed cells as described above but the injection coordinates were 2 mm posterior and lateral from bregma, 2.2 mm deep from the outer border of the brain. These coordinates allowed to perform a deeper resection cavity, able to host 5 µl of therapeutic hydrogel.
At day 14 post-tumor inoculation, all mice received tumor resection and were randomly assigned to one of the following groups: 1) untreated (n = 9); 2) GemC12-LNC (n = 8); 3) GDC-0152 (n = 9); 4) GemC12-LNC and GDC-0152 (n = 8). N-dodecanoyl gemcitabine (GemC12, Atlanchim Pharma, France) was formulated as a nanomedicine hydrogel as previously reported 37. The GemC12-LNC treatments were injected into the tumor resection cavity at the time of surgery (GemC12 dose: 4 mg/kg, 5µl of GemC12-LNC hydrogel) 10,11. Monovalent SMAC mimetic GDC-0152 (Selleckchem, USA) was dissolved in DMSO (Sigma-Aldrich, France) and administered by retroorbital injection at a dose of 20 mg/kg (100 µl) 36,38. The GDC-0152 treatment was performed post-surgery and every week (8 administrations in total).
For the tumor surgery, tumor surgery was performed as previously described but this time the biopsy punch was inserted at a depth of 2 mm to allow the local injection of therapeutic concentrations of hydrogel. For groups 1 and 3, no treatment was administered in the tumor cavity. For groups 2 and 4, GemC12-LNC hydrogel was injected into the resection cavity using a 0.3 mL insulin syringe. The dural window was repaired by covering it with a 4 × 4 mm square piece of Neuro-Patch® (Aesculap, Germany) impregnated with fibrin sealant (Tisseel Prima; Baxter, France). The wound was then closed using tissue adhesive glue. Mice were then monitored daily, and their body weight was measured 2–3 times per week. The mice were sacrificed when they reached the endpoints.
Immunohistochemistry
Brains were perfused and stored in PBS as previously described until further use. For immunohistochemistry analysis, brains were then embedded in 4% agar (Sigma-Aldrich, France) and 50 µm slices were obtained using a vibratome (Leica VT1200 S). To label microglia, antigen-presenting cells (APCs) and anti-inflammatory cells, we used anti-TMEM119, anti-MHCII and anti-Arginase 1 (ARG-1) antibodies respectively.
For the double TMEM119/MHCII labeling, brain sections were incubated for 2h at room temperature in blocking buffer (5% donkey serum, 5% BSA and 0.03% Triton 100X in 1X PBS). Sections from 4 different mice per condition were incubated with monoclonal anti-TMEM119 primary antibody (Abcam, 1:100; ab209064, clone 28 − 3) overnight at 4°C. Sections were then rinsed with 5% PBS/BSA (3 × 10 min) and monoclonal anti-MHCII primary antibody (Ebiosciences, 1:50; 14-5321-85) was incubated for 1h at room temperature. For ARG-1 labeling, brain sections are permeabilized with denaturation buffer (10% PBS 10X, 16% HCL 37%, 0.5% Triton 100X in distilled H2O) at 37° for 20 min, then incubated with neutralization buffer (3.8% sodium tetraborate in distilled H2O, pH adjusted to 8.5) for 10 min at room temperature. Slices are then rinsed with 5% PBS/BSA (3 times 10 min at room temperature) and incubated in blocking buffer. Anti-Arginase 1 polyclonal primary antibody (Novus Bio, 1:100; NB100-59740) was incubated at 4°C overnight. Following incubation with primary antibodies, the sections were washed with PBS/BSA 5% (3 × 10 min) and then anti-Rabbit Alexa Fluor 488 (Invitrogen, 1/600; A11055), anti-Goat Alexa Fluor 647 (Invitrogen, 1/600; A21447), anti-Rat Alexa Fluor 647 (Jackson ImmunoResearch, 1:600; 712-605-150) and Hoechst (Sigma-Aldrich, 1:1000) secondary antibodies were incubated for 1h at room temperature. After 3 × 10 min washes in 1X PBS, sections were incubated in TrueBlack (Biotium, 1:20; 23007) in 70% ethanol for 30 sec. After a final wash (3 × 10 min), slides were mounted using a mounting medium (Mowiol-488 /Glycerol/DABCO) and stored at 4°C. Confocal microscopy acquisitions were carried out on a Zeiss LSM 700 microscope with a 20× objective at 0.8 numerical aperture on 2 different tumor areas per slide (center and periphery). Images were then processed with ZEN Blue Edition 3.0 version and analyzed with Imaris software.
Statistical analysis
In the manuscript, “n” corresponds to the number of independent experiments performed (single animals for animal experiments) and data quantification is presented as average ± SEM. Statistical analyses were performed using GraphPad Prism (GraphPad Software, USA). All reported p-values are two-sided and p-values < 0.05 were considered statistically significant (#p = 0.05, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001).
For the human studies, categorical variables were presented as frequencies and percentages, continuous variables as median, mean, range and standard error. Paired T-tests were used to compare quantitative values of blood cells before and after resection.
For the blood immunophenotyping analysis, multiple comparisons were performed using Kruskal-Wallis nonparametric analysis with uncorrected Dunn's test to compare the control and resected groups. For the brain immunophenotyping, data obtained from unresected and resected tumors were compared using a nonparametric Mann-Whitney test.
For the nuclear imaging quantification analysis, statistical analysis was performed using Unpaired t test with Welch's correction at each time point vs. the control value.
For the biphotonic imaging quantification analysis, quantification of images obtained from at least three animals were analysed using Imaris software and compared using unpaired nonparametric Mann-Whitney test between the groups.
For the in vivo efficacy studies, the statistical analysis was estimated from a comparison of Kaplan–Meier survival curves using the log-rank test (Mantel-Cox test). To compare the quantification of the histological stainings, we used unpaired t test. The Chi² test of independence was used to compare the values of the repartition of CD45+ and TMEM119+ cells within the tumors and their distance to blood vessels.