Astrocytes donate mitochondria to increase glioblastoma tumorigenicity

Justin Lathia (  lathiaj@ccf.org ) Cleveland Clinic https://orcid.org/0000-0003-3168-7290 Dionysios Watson University Hopsitals Cleveland Medical Center Defne Bayik Cleveland Clinic https://orcid.org/0000-0002-4740-8869 Sarah Williford University of Alabama Adam Lauko Cleveland Clinic https://orcid.org/0000-0001-5654-0784 Yadi Zhou Cleveland Clinic Gauravi Deshpande Cleveland Clinic Juliana Seder Cleveland State University Jason Mears Case Western Reserve University Feixiong Cheng Cleveland Clinic https://orcid.org/0000-0002-1736-2847 Anita Hjelmeland University of Alabama at Birmingham

tumor microenvironment as a mechanism that enhances tumorigenesis in glioblastoma. We found that this transfer occurs primarily from brain-resident cells, including astrocytes, and can be appreciated in vitro and in vivo through intercellular connections between GBM cells and non-malignant host cells. The acquisition of astrocyte mitochondria drives an overall enhancement of mitochondrial membrane potential and metabolic capacity, while increasing 15 glioblastoma cell self-renewal and tumor-initiating capacity. Collectively, our findings demonstrate that astrocyte mitochondrial transfer augments the tumorigenic capacity of glioblastoma cells and reveals a previously unknown cell-cell communication mechanism that drives tumor growth. We anticipate our findings will open new research directions to decipher the molecular events linking mitochondria acquisition from non-malignant cells to increased 20 tumorigenicity of recipient GBM cells. This line of research will lead to new therapeutic opportunities targeting this understudied phenomenon and its sequelae in GBM.

Main Text
While dynamic microenvironmental interactions drive tumor growth and therapeutic resistance in GBM 2,3 , the underlying direct cell-cell communication mechanisms remain poorly 25 understood. Recent studies established that mitochondrial transfer in mouse models following ischemic stroke was important for neuronal recovery 7 . Given the crucial role of brain microenvironmental interactions in driving GBM pathobiology 8 , we hypothesized that acquisition of mitochondria from the tumor microenvironment results in metabolic changes and augmented tumorigenicity in GBM cells. 30 To assess whether non-malignant host cells also transfer mitochondria to GBM in vivo, we orthotopically implanted GFP-expressing syngeneic mouse GBM models (SB28 and GL261) into transgenic C57BL/6 mice expressing a mitochondria-localized mKate2 fluorophore by fusion to the localization peptide of cytochrome c oxidase 8 (mito::mKate2 mice 9 ), Fig. 1A). Confocal microscopy of GBM tumors from mito::mKate2 mice revealed mKate2 + puncta 35 within 15-50% of GFP + GBM cells ( Fig. 1B-C, fig. S1), demonstrating that host cell mitochondria were being transferred to GBM cells in vivo. By co-staining tissue sections with wheat germ agglutinin to highlight glycoprotein-rich structures 10 , we observed that transferred mitochondria frequently localized at host-tumor interfaces and at the termini of intercellular connections between GFP + GBM cells and GFPhost cells containing mKate2 + mitochondria 40 (Fig. 1B, fig. S2-3). Thus, cytoplasmic interconnections previously reported to exist among GBM cells 4 likely also play a role in organelle transfer between GBM and microenvironment cells.
Having observed mitochondrial transfer from the tumor microenvironment to GBM cells in mouse models, we proceeded to interrogate the identity of the host mitochondrial donor cells. 3 GBM tumors in mouse models and patients are known to have significant infiltration by both brain-resident glia and peripheral immune cells that transmigrate into the microenvironment 8 . We established orthotopic GBM tumors in wildtype C57BL/6 mice that had first received lethal irradiation with subsequent bone marrow reconstitution from mito::mKate2 mice (mito::mKate2àWT) to restrict mKate2 expression to immune cells (Fig. 1A). Analysis of 5 single-cell suspensions by flow cytometry indicated negligible host mitochondrial transfer to GFP + GBM cells in mito::mKate2àWT mice, while 20-60% of GFP + GBM cells detected were mKate2 + in mito::mKate2 mice (Fig. 1D-E). Taken together, these data suggest that brain-resident cells and not tumor-infiltrating immune cells were the major donors of mitochondria to GBM cells in vivo. 10 To further elucidate the identity of the predominant mitochondrial donor cell populations, we co-cultured prevalent tumor-infiltrating cell types with GFP + GBM cells at a 1:2 donor-torecipient ratio. After 2 hours, we assessed the percentage of mKate2 positivity as a marker of mitochondrial transfer ( Fig. 2A). Consistent with our in vivo observations, we found that brainresident glia (astrocytes and microglia) donated significantly more mitochondria on a per-cell 15 basis compared to bone marrow-derived macrophages (Fig. 2B, fig. S4A). While further polarization of macrophages to an M2-or M1-like phenotype potentially favored increased mitochondrial transfer, the degree of transfer was on average 5-10-fold less than that observed with the brain-resident glia in vitro ( fig. S4B). In addition, culturing GBM cells with donor cell-conditioned medium did not result in mKate2 signal transfer to tumor cells, further 20 suggesting that the mitochondrial transfer is primarily contact dependent (Fig. 2C, fig. S4C). Live confocal imaging of astrocyte/SB28 cell co-cultures confirmed the occurrence of mitochondrial transfer at donor-recipient contact sites (Fig. 2D, Movie S1), consistent with our in vivo observation that host-tumor intercellular connections contained host mitochondria ( fig. S2-3). Taken together, these results indicate that physical interaction of tumor cells with 25 donor cells is required for effective mitochondrial transfer.
To interrogate the functional significance of mitochondria acquisition, we sorted mKate2 + (with exogenous mitochondria) and mKate2 -(without exogenous mitochondria) GBM cells from mito::mKate2 astrocyte co-cultures after 2 days (Fig. 3A). Unsupervised clustering of RNA-seq analysis revealed that tumor cells had a distinct gene expression profile compared to 30 astrocytes, confirming that sorted mKate2 + SB28 cells were not significantly contaminated by astrocytes from the co-culture ( fig. S5A-C). Additionally, there were a limited number of differentially expressed genes between mKate2 + and mKate2cells, indicating that mitochondrial transfer in vitro does not induce broad transcriptional changes ( fig. S5D, Table  S1). However, not surprisingly, genes that were upregulated >1.5-fold in mKate2 + samples 35 were enriched within pathways related to mitochondrial biology, in particular electron transport and mitochondrial organization (Fig. 3B, fig. S6). In line with this finding, mKate2 + GBM cells exhibited increased mitochondrial membrane potential and mass compared to their mKate2counterparts, as measured by MitoStatus and MitoTracker Deep Red staining (Fig.  3C, fig. S7A, fig. S7B). When normalized to total mitochondrial content, mitochondrial 40 potential remained higher in mKate2 + cells ( fig. S7A, fig. S7C).
To further test the metabolic consequences of acquiring astrocyte mitochondria, we assessed metabolic activity via Seahorse assays, as well as expression of key metabolic proteins by adapting a previously reported metabolic flow cytometry panel 11 . Metabolic profiling of sorted cells revealed that both mKate2 + and mKate2 -GBM cells had a lower oxygen consumption-45 to-extracellular acidification ratio (OCR/ECAR) than matched astrocytes, suggesting that they 4 rely more heavily on glycolytic metabolism in the glucose-rich assay environment, in line with the observed high and invariable GLUT1 expression among these GBM cell subsets ( fig. S8A,  8E, 9). In contrast, despite similar rates of mitochondrial respiration in the presence of excess glucose, mKate2 + cells had higher levels of ATP5A (mitochondrial ATP synthase), SLC20A1 (cellular phosphate import) and ACC1 (fatty acid synthesis), pointing to potential differences 5 in the cellular energetics of mitochondria recipients. Increased ATP5A protein expression was also observed in mKate2 + GBM cells isolated from intracranial tumors implanted in mito::mKate2 mice (data not shown).
To assess whether mitochondrial acquisition impacts the tumorigenic potential of GBM cells, we utilized an in vitro limiting dilution assay and found that mKate2 + SB28 cells had a higher 10 self-renewal capacity compared to matching cells without exogenous mitochondria, with an estimated stem cell frequency that was ~2.5-fold higher (Fig. 4A, fig. S10A). We further performed an in vivo limiting dilution assay by intracranial implantation of decreasing numbers of mKate2 + and mKate2 -SB28 cells. In three independent experiments, we consistently observed that cohorts receiving mKate2 + GBM cells displayed increased penetrance and/or 15 earlier symptomatic disease ( fig. S10B), corresponding to an estimated 3-fold higher frequency of tumor-initiating cells (Fig. 4B). Finally, ~10-20 % of patient-derived xenograft (PDX) cells also demonstrated the ability to acquire mitochondria from human astrocytes in co-culture, as assessed by flow cytometry and ImageStream in two separate lines (Fig. 4C, fig. S11). Taken together, these data demonstrate increased tumorigenicity of cells that received astrocyte-20 derived mitochondria and suggest that this mode of intercellular communication could play a role in human GBM tumors by altering cellular energetics (Fig. 4D).
Organelle transfer is an increasingly recognized biological process in models of GBM 12-14 and other cancers 15-17 . However, there is a lack of understanding about its in vivo relevance, specific donor-recipient interactions and downstream effects. We found that in the context of GBM, 25 this is a frequent in vivo event, involving primarily brain-resident glial cell donors, and is associated with recipient GBM cells acquiring a more tumorigenic phenotype. Mitochondrial transfer led to changes in key metabolic proteins and conferred an advantage to GBM cells within the brain microenvironment. Additional studies into the mechanisms of transfer, as well as into the molecular changes in the recipient cells, are crucial to expand our understanding of 30 this observation and lead to the identification of a new class of therapeutic targets. This line of research could also have broad applicability to tumors outside the central nervous system.

Methods
Mouse tumor cell maintenance and transduction SB28 cells were gifted by Dr. Hideho Okada (University of California, San Francisco). GL261 cells were obtained from the Developmental Therapeutics Program, National Cancer Institute. All cell lines were treated with 1:100 MycoRemoval Agent (MP Biomedicals) upon 5 thawing and routinely tested for Mycoplasma spp. (Lonza). Cells were maintained in RPMI 1640 (Media Preparation Core, Cleveland Clinic) supplemented with 10% FBS (Thermo Fisher Scientific) and 1% penicillin/streptomycin (1% Pen/Strep, Media Preparation Core). For the generation of GFP-expressing GL261 cells, parental GL261 cells were transduced with pReceiver-Lv207 (Genecopoeia) and were selected with 300 μg/ml hygromycin B (Invitrogen). GFP 10 expression was confirmed by flow cytometry.

Mice
All animal experiments were approved by the Institutional Animal Care and Use Committee of Cleveland Clinic and performed in accordance with established guidelines. Tg(CAG-15 mKate2)1Poche/J (mito::mKate2, stock #032188) mice were purchased from The Jackson Laboratory and were housed in the Cleveland Clinic Biological Research Unit. Both sexes of mito::mKate2 mice were intracranially injected at 4-8 weeks old with 10,000-20,000 SB28 or 100,000 GL261-GFP cells in 5 μl RPMI null media into the left cerebral hemisphere 2 mm caudal to the coronal suture, 3 mm lateral to the sagittal suture at a 90° angle with the murine skull to a 20 depth of 2.5 mm, using a stereotaxis apparatus (Kopf). Age-and sex-matched wildtype littermates were used as controls. Mice were monitored daily for neurological symptoms, lethargy and hunched posture that would qualify as signs of tumor burden.

Bone marrow transplantation 25
Four-week-old male mice were treated with 11 Gy radiation in two fractions 3-4 hours apart. Reconstitution was achieved by retro-orbital injection of 2 x 10 6 bone marrow cells from mito::mKate2 mice. Drinking water was supplemented with Sulfatrim (trimethoprimsulfamethoxazole; Pharmaceutical Associates, Inc.) during the first 10 days, and mice were monitored for an additional 6 weeks for weight loss and symptoms of infection before tumor 30 inoculation. Survival analysis was performed as described above.

Assessment of mitochondrial trafficking in vivo by confocal microscopy
At experiment endpoint, mice were perfused with 4% paraformaldehyde in PBS (4% PFA, Fisher Chemical) using the Perfusion One system (Leica). Brains were dissected and stored for 48 35 h in 20 ml 4% PFA at 4°C, transferred to 20 mL PBS for 48 h at 4°C, and finally transferred to 20 mL 30% sucrose (Fisher Chemical) in PBS at 4°C until density equilibration (typically ~48 h). Samples were embedded in OCT Compound (Tissue-Tek), flash frozen and stored at -80°C. Ten micron sections were prepared using a cryostat (Leica Biosystems), immediately mounted on charged glass slides (Superfrost Plus, Fisherbrand) and stored at -20°C until staining. For staining, 40 selected slides were allowed to warm to room temperature, washed twice with 0.1% Triton-X 100 (Sigma) in PBS (PBS-T) for 10 min, and permeabilized overnight at 4°C with 5% donkey serum (Jackson ImmunoResearch), 0.3% Triton-X 100, 1 mg/mL bovine serum albumin (molecular biology grade, Sigma) in PBS. Sections were then stained with chicken anti-GFP antibody (1:1000 in permeabilization buffer; clone AB_2307313, AvesLabs) overnight at 4°C. Sections were then 45 washed 3 times with PBS-T for 10 min and stained with donkey anti-chicken Alexa fluor (AF) 2 488-conjugated secondary antibody (1:500 in permeabilization buffer, AB_2340375, Jackson ImmunoResearch) overnight at 4°C. Sections were then washed 3 times with PBS-T for 10 min, then exchanged to HBSS (without phenol red) with CaCl2, MgCl2, MgSO4 and 0.1% Triton-X 100 (HBSS-T) by washing 3 times for 10 min. Sections were immediately stained with wheat germ agglutinin conjugated to AF680 (4 μg/mL in HBSS-T; ThermoFisher) for 1 hour at room 5 temperature and washed 3 times with HBSS-T for 10 min. Nuclear staining was performed with Hoechst 33342 (3.3 μg/mL in PBS-T). Finally, stained sections were mounted and coverslipped with Vectashield Vibrance and no. 1.5 glass coverslips (Fisherbrand). Mounting medium was given at least 24 hours to cure prior to imaging.
Confocal microscopy of stained tissue sections was performed using a Leica SP8. Full-

45
A total of 20,000-40,000 astrocytes and microglia were separately cultured in a 96-well flat-bottom plate (Thermo Fisher Scientific) in NSC medium or MPM. Forty-eight hours later, supernatants were collected and centrifuged at 400 g for 5 minutes to remove residual cells. Tumor 3 cells were added at a recipient:donor ratio of 2:1. Samples were incubated for 2 hours and treated with Accutase to generate single-cell suspensions. Cells were transferred to 96-well U-bottom plates (Thermo Fisher Scientific) to stain with Live/Dead dye. Exogenous mitochondria uptake of GFP + tumor cells was assessed with a BD LSRII Fortessa. 5 Generation of mouse macrophages and in vitro mitochondria uptake assay Bone marrow from the femur and tibia of 4-8-week-old male and female mito::mKate2 mice was flushed with PBS using a 27 G needle. A total of 80,000 cells was cultured in 24-well plates (Corning) and treated with 50 ng/ml recombinant mouse M-CSF (Biolegend) in IMDM (Media Preparation Core) supplemented with 1% Pen/Strep and 20% FBS for 6 days. IFNγ or IL-10 4 (50 ng/ml; Biolegend) was added for 48 hours to further induce polarization of macrophages to M1-or M2-like macrophages. Supernatants were collected and centrifuged at 400 g for 5 minutes to remove residual cells. Tumor cells were added at 2-fold abundance in technical duplicates for a 2-hour incubation. Samples were incubated with Accutase for 5 minutes and transferred into 96well U-bottom plates for staining with the viability dye and anti-CD11b antibody as described 15 above. Exogenous mitochondria uptake was analyzed from GFP + tumor cells using a BD LSRII Fortessa.
Confocal microscopy time-lapse imaging SB28 cells (40,000) were co-cultured with 80,000 mito::mKate2 astrocytes in a glass-20 bottom 35 mm dish (Mat-tek) overnight. Growth medium was replaced with phenol-red free NSC medium. Multiple full-thickness z-stacks were obtained every 10 min using a Leica SP8 microscope in a 37°C chamber supplemented with 5% CO2, 95% humidity using a 20X/0.8NA objective lens. Time-lapse frames were subsequently analyzed by LasX software.

25
Sorting of tumor cells from co-cultures Astrocytes were harvested from flasks by Accutase treatment and stained with a 1:1000 dilution of CellTrace™ Violet Cell Proliferation Dye in PBS at 37°C for 20 minutes. Tumor cells and astrocytes were co-cultured at a 1:1 ratio for 48 hours in NSC media. Samples were sorted into RPMI with 20% FBS and cultured overnight in complete RPMI for subsequent functional assays:
In vivo limiting-dilution assay 40 Sorted mKate2 + and mKate2 -SB28 cells were harvested with Accutase after overnight post-sort culture, pelleted by centrifugation and resuspended in basal RPMI. Cells were counted with trypan blue using a TC-20 cell counter (BioRad) and volume adjusted to achieve decreasing cell concentrations (18,000-3,000) for intracranial implantation. After any volume adjustments, cells were recounted immediately prior to implantation to ensure accurate counts. Subsequently, 45 4-6-week-old C57BL/6 mice were intracranially implanted with equal numbers of either mKate2 + or mKate2 -SB28 cells as described above. The identity of the implanted cells in mice was then blinded to investigators, who monitored animals daily for neurological symptoms indicative of tumor-related morbidity endpoint.

Seahorse assay
A total of 15,000-25,000 astrocytes, mKate2 + and mKate2tumor cells were seeded in 5 Seahorse XFe24 Cell Culture Microplates (Agilent) in 250 μL complete RPMI overnight. The media was replaced with phenol red-free RPMI supplemented with 10 mM glucose, 2 mM glutamine and 1 mM pyruvate (Agilent) and incubated for 45 min in a non-CO2 incubator. The oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) were measured using the Seahorse XF24 platform (Agilent). containing 10 ng/mL EGF, 10 ng/mL FGF, 1% sodium pyruvate, 2% GEM21 (Gemini Bio), and 1% Pen/Strep was used to culture the PDX lines. A clone of HEK293T cells able to grow in serumfree media (CSC293T) was generated and cultured as previously described 4 . Normal human astrocytes (NHAs; provided by Dr. Russell Pieper at UCSF) were cultured in DMEM supplemented with 10% FBS, 1% N2 NeuroPlex Supplement (Gemini Bio), 3 ng/mL EGF, and 1% Pen/Strep. CSC293T cells were transfected with psPAX2, pCMV-VSVG, and pLYS1-Mito-GFP or pCMV-RFP using FuGENE HD transfection reagent (Promega). Viral concentration was determined using the Lenti-X qRT-PCR Titration kit (Takara Bio). NHAs were transduced with 5 mito-GFP lentivirus and selected for cells stably expressing mito-GFP using puromycin. D456 and JX22 cells were infected with RFP lentivirus and selected using Blastacidin S (Gibco) to select for stable RFP-expressing cells. Where indicated, cells were sorted for mito-GFP + or RFP + with assistance from the Flow Cytometry Core at the University of Alabama at Birmingham.

10
ImageStream RFP-expressing D456 or JX22 cells were co-cultured at a 1:1 ratio in the presence of mito-GFP + NHAs for 24hrs. Samples were imaged at 40x magnification and extended depth of field (EDF). Mito-GFP was acquired on ch02 and RFP on ch04. Ch01 and ch09 are used for brightfield imaging, and ch12 was used for side scatter. A total of 5000 events was recorded, and relevant 15 single color and unstained controls were used. Data was analyzed using IDEAS software (version 6.2; EMD Millipore). manufacturer's protocol with the following exception. Up to 1400 μL of sorted cell suspension was sequentially loaded onto each DNA binding column. When multiple columns were used for a given sample, elution of DNA was performed sequentially, using 15 μL of RNAse-free water to pool the isolated DNA in a minimal volume. RNA sequencing and analysis were performed by GENEWIZ (South Plainfield, NJ).

30
Briefly, samples were sequenced using an Illumina HiSeq, 2x150 bp configuration and ≥350 M raw paired-end reads. An average of 41.6 M paired-end reads was sequenced across 9 samples. After Illumina universal adapters were trimmed, the reads were mapped to the Mus musculus GRCm38 reference genome using the STAR aligner v.2.5.2b. Unique gene hit counts were calculated by using featureCounts from the Subread package v.1.5.2.

35
For comparison of tumor cells with astrocytes, genes with an adjusted p-value < 0.05 and absolute log2 fold change > 1 were called as differentially expressed genes using DESeq2. For assessment of differentially up-regulated pathways in mKate + versus mKate -SB28 cells, genes that were up-regulated >1.5-fold with a count number of >50 and an unadjusted p-value of <0.05 (table S1) were plugged into https://maayanlab.cloud/Enrichr/.

Protein-protein interactions and network visualization
Differential expression analysis was performed using edgeR 3.34 5 . Genes with count per million greater than 1 in at least 2 samples were used for the analysis. P values < 0.05 were considered significant. The mouse differentially expressed genes (DEGs) were mapped to the 45 human homologs using the NCBI HomoloGene database (https://www.ncbi.nlm.nih.gov/homologene). We then performed the enrichment analysis using Enrichr 6 for the entire DEGs and for the up-and down-regulated genes separately. 6 The protein-protein interactions (PPIs) among the DEGs were extracted using a human protein interactome we built previously 7 that contains 17,706 protein nodes and 351,444 PPI edges. We then visualized this protein-protein interaction network using Cytoscape 3.8 8 . Genes that localize to mitochondria are indicated by diamond node shape based on the Human MitoCarta2.0 database 9 .

Data representation and analysis
Flow cytometry data were analyzed and generated using FlowJo software (BD Biosciences, v10.7.2). Graphs were generated and statistical analysis were performed using Excel (Microsoft Office, v16.52) or Prism (GraphPad, v9.2.0) software. All measurements 10 shown represent distinct samples, unless otherwise indicated. All statistical tests are 2-tailed and corrected for multiple comparisons, unless otherwise indicated.

Supplementary Text
Figs. S1 to S11 Tables S1 Captions for Movies S1

Other Supplementary Materials for this manuscript include the following:
Movies S1   Protein-protein interaction network for the differentially expressed genes. Mouse genes were mapped to human genes according to NCBI HomoloGene. Protein-protein interactions were extracted for these genes/proteins using our human protein interactome. Diamond-shaped nodes indicate that these genes are mitochondrially localized based on the Human MitoCarta2.0 database. Node color shows the log2 fold change of the genes.   S8: GBM cells that acquire astrocyte mitochondria have higher levels of metabolic proteins associated with ATP and fatty acid synthesis. GBM cell lines were co-cultured with mito::mKate2+ astrocytes for 24 hours and stained with antibodies against key metabolic proteins, as denoted. Expression levels were assessed by flow cytometry. (A) Representative histograms depicting differential expression of critical metabolic proteins by astrocytes and mKate2 + and mKate2 -SB28 cells. Astrocytes had higher levels of acetyl-CoA carboxylase (ACC1), SLC20A1 Fig. S9: mKate2 + and mKate2 -SB28 cells sorted from astrocyte co-cultures similarly rely on glycolysis in a glucose-rich environment. GBM cells were sorted from mito::mKate2 + astrocyte co-cultures after 48 h, cultured overnight, and then subjected to Seahorse assay in standard assay media in the presence of excess glucose. Oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) were measured at baseline in n=4-8 biological replicates. * p<0.05 as determined by one-way ANOVA.

A B
Movie S1. Time-lapse confocal images of in vitro mitochondrial transfer. Astrocytes from mito::mKate2 mice were co-cultured with GFP-expressing SB28 cells for 16 hours before live imaging. Full-thickness z-stacks were obtained at 10-minute intervals. Movie depicts a mito::mKate2 + astrocyte (magenta) in contact with a GFP-expressing SB28 cell (green). An mKate2 + mitochondrion is transferred to the dividing GBM cell, which shuttles the mitochondrion back and forth along an intercellular connection linking the GBM daughter cells. At the end of the video, the mKate2 + mitochondrion is retained by the lower daughter cell. See also Fig. 2D for still images with accompanying z-reconstructions.