LDHA-regulated tumor-macrophage symbiosis promotes glioblastoma progression

Abundant macrophage infiltration and altered tumor metabolism are two key hallmarks of glioblastoma. By screening a cluster of metabolic small-molecule compounds, we show that inhibiting glioblastoma cell glycolysis impairs macrophage migration and lactate dehydrogenase (LDH) inhibitor stiripentol (an FDA-approved anti-seizure drug for Dravet Syndrome) emerges as the top hit. Combined profiling and functional studies demonstrate that LDHA-directed ERK pathway activates YAP1/STAT3 transcriptional co-activators in glioblastoma cells to upregulate CCL2 and CCL7, which recruit macrophages into the tumor microenvironment. Reciprocally, infiltrating macrophages produce LDHA-containing extracellular vesicles to promote glioblastoma cell glycolysis, proliferation, and survival. Genetic and pharmacological inhibition of LDHA-mediated tumor-macrophage symbiosis markedly suppresses tumor progression and macrophage infiltration in glioblastoma mouse models. Analysis of tumor and plasma samples of glioblastoma patients confirms that LDHA and its downstream signals are potential biomarkers correlating positively with macrophage density. Thus, LDHA-mediated tumor-macrophage symbiosis provides therapeutic targets for glioblastoma.


INTRODUCTION
Glioblastoma is a devastating brain tumor in human adults with a median survival averaging 15-20 months following initial diagnosis 1,2 .Unfortunately, current therapies have failed to improve the survival of glioblastoma patients meaningfully over the last four decades 3,4,5,6,7 .Due to glioblastoma cell heterogeneity and genetic instability, clinical trials for targeted therapies (e.g., therapies targeting receptor tyrosine kinase signaling) have also failed to improve glioblastoma patient outcomes 8,9 .There is an increasing recognition that the signaling from glioblastoma cells not only impacts cancer cell biology, but also regulates the biology (e.g., recruitment and activation) of immune cells in the tumor microenvironment (TME), thus inducing a tumor-immune cell symbiotic interaction 5,6 .Among the TME, tumor-associated macrophages and microglia (TAMs) are the largest and most prominent population of immune cells, which account for up to 50% of total live cells in glioblastoma tumor mass 10,11 .Our recent studies have demonstrated that PTEN-yes-associated protein 1 (YAP1)-lysyl oxidase (LOX) in glioblastoma cells, and CLOCK-olfactomedin like 3-legumain and tissue factor pathway inhibitor 2 signaling in glioblastoma stem cells (GSCs) are the key drivers for the in ltration of macrophages and microglia, respectively, which, in turn, promote tumor growth and immunosuppression in glioblastoma 3,12,13,14 .Such studies highlight the opportunity of identifying the key signals that establish symbiotic interactions between cancer cells and the TME, thus inducing a pro-tumor and immunosuppressive environment for glioblastoma tumorigenesis.
Metabolic reprogramming enables cancer cell growth and proliferation, which is recognized as a prominent hallmark of cancer 15 .Interestingly, recent studies have shown that metabolic reprogramming (such as the regulation of glucose, lipid, tryptophan, and NAD + metabolism) in cancer cells evades antitumor immunity by suppressing lymphocytes 16,17,18 and recruiting immunosuppressive myeloid cells, including macrophages 19,20,21 .These ndings gain added signi cance as myeloid cells (e.g., macrophages), lymphocytes, and glioblastoma cells, as well as their symbiotic interactions, are critical for affecting tumor growth and immunotherapy resistance 5,6,8,22 .Encouraged by their functional signi cance 6 , a large body of pharmacological tools have been proposed to target these symbiotic interactions in glioblastoma mouse models 5 .However, certain challenges remain, such as the bloodbrain barrier (BBB) that can limit drug delivery into the glioblastoma TME.This creates di culties in regard to translating the preclinical ndings into the clinic 5 .Together, these insights prompted us to conduct a screen of metabolic and brain-penetrant small-molecule compounds that may inhibit glioblastoma cell-induced macrophage in ltration.In this screen, lactate dehydrogenase (LDH) inhibitor stiripentol emerged as the top hit.
LDH is a key player in glucose metabolism that regulates the conversion between pyruvate and lactate.LDH is comprised of two major subunits (e.g., LDHA and LDHB) with LDHA converting pyruvate to lactate in anaerobic conditions and LDHB favoring lactate to pyruvate in the presence of oxygen 23 .However, most cancer cells use aerobic glycolysis (also known as "Warburg effect") to maintain their tumor potential even in the presence of oxygen and produce high levels of lactate 24,25 .Increasing evidence has shown that LDHA-mediated glycolysis promotes glioblastoma cell proliferation and survival and induces resistance to radiotherapy and chemotherapy 26,27,28,29 .However, the potential link between immune cells and LDHA-mediated tumor glycolysis in glioblastoma has not been established.Here, we elucidate a novel function and molecular mechanism of glioblastoma cell LDHA in promoting macrophage in ltration into the TME and reveal the co-dependencies for macrophage-derived extracellular vesicles (EVs) in supporting glioblastoma cell glycolysis, growth, and survival.Preclinical trials in glioblastoma mouse models, followed by clinical-pathological validations using patient tumor and plasma samples, point to LDHA and its downstream signals as promising therapeutic targets for glioblastoma.

RESULTS
Glioblastoma cell glycolysis promotes macrophage migration.The metabolism signature and immune score in The Cancer Genome Atlas (TCGA) glioblastoma tumors have been de ned based on gene expression data to infer the levels of tumor metabolism 30 and immune cell populations 31 , respectively.To identify the potential connection between tumor metabolism and immunity that might in uence glioblastoma tumor biology, the correlation analyses among the two signatures and patient survival were performed.We found that high tumor metabolism signature was correlated with poor outcomes (Fig. 1a) and correlated positively with immune score (Fig. 1b).Moreover, the metabolism signature was enriched in mesenchymal and IDH1-WT glioblastoma, but not related to the status of tumor recurrence, gender and MGMT promotor methylation (Extended Data Fig. 1a-e).These ndings aligned with the Gene Set Enrichment Analysis (GSEA) on hallmark pathways, Gene Ontology Enrichment Analysis (GOEA) on the sub-ontologies of Biological Process, and KEGG Enrichment Analysis showing prominent representations of cytokine and chemokine signatures, immune response networks, and leukocyte and myeloid cell signatures in metabolism-high TCGA glioblastoma patient tumors compared to metabolism-low patient tumors (Table S1).To identify speci c immune cells linked to tumor metabolism in glioblastoma, we audited the TCGA glioblastoma patient tumors for 18 types of immune cells using validated gene set signatures 3,12,13,32,33 .As a result, macrophage and monocyte were identi ed as the top immune cell types correlating positively with the metabolism signature (Fig. 1c).Conversely, CD8 + activated T cells showed a negative correlation with the metabolism signature (Fig. 1c).Together, these ndings suggest a connection between tumor metabolism and macrophage/monocyte in ltration in glioblastoma patient tumors.
Given the importance of macrophages in glioblastoma progression 8 , we hypothesized that pharmacological inhibition of tumor metabolism-induced macrophage in ltration is a promising therapeutic strategy 5 .We selected a cluster of 55 brain-penetrant small-molecule compounds with metabolic reprogramming functions (Table S2) and performed a screen focusing on macrophage migration using conditioned media (CM) from CT2A cells treated with or without these compounds at 10 µM.This screen resulted in identi cation of 24 compounds that signi cantly inhibited CT2A CM-induced macrophage migration (Fig. 1d and Extended Data Fig. 1f).Next, we performed a second round of screen with these 24 compounds at a lower concentration (5 µM) and found that 7 (stiripentol, lopinavir, o oxacin, vorasidenib, IDH889, progesterone, and IOX4) of them impaired CT2A CM-induced macrophage migration (Fig. 1e and Extended Data Fig. 1g).Consistently, the LDH inhibitor stiripentol showed the strongest effect in these two rounds of screens, which led us hypothesize that glioblastoma cell glycolysis is essential for macrophage in ltration.To con rm it, we analyzed the single-cell RNA sequencing (scRNA-seq) data from 44 fragments of tumor tissues of 18 glioma patients, including 2 lowgrade gliomas (LGG), 11 newly diagnosed glioblastoma (ndGBM), and 5 recurrent glioblastoma (rGBM) 34 .Speci cally, glioblastoma tumors containing tumor cells (Fig. 1f) and tumor-in ltrating myeloid cells (Fig. 1g) were analyzed.The glycolysis hallmark signature 35 was highly expressed in glioblastoma cells (Fig. 1h), which correlated positively with the abundance of macrophages and monocytes (Fig. 1i) and negatively with the level of microglia, but did not show a signi cant correlation with dendritic cells (DCs) (Extended Data Fig. 2a).Similarly, bioinformatics analyses in TCGA glioblastoma tumors demonstrated that high glycolysis signature correlated with increased immune score (Extended Data Fig. 2b); prominent representations of immune response networks, cytokine and chemokine signatures, as well as leukocyte and myeloid cell migration signatures (Table S3); and increased macrophages, monocytes, and to a lesser extent, DCs, and decreased microglia (Extended Data Fig. 2c).Finally, glycolysis signature, but not other metabolism-related hallmark signatures, was enriched in glioblastoma patient tumors compared to normal brain tissues (Extended Data Fig. 2d, e), and was increased in glioma cells of glioblastoma compared to LGG (Extended Data Fig. 2f).
To biologically validate the role of glycolysis in triggering macrophage in ltration, we treated mouse glioblastoma cells (e.g., CT2A and GL261) and glioblastoma patient-derived GSC272 with a glycolysis inhibitor 2-deoxy-D-glucose (2-DG) 21 , which can inhibit glycolysis as shown by reduced extracellular acidi cation rate (ECAR) and lactate production (Extended Data Fig. 2g-j).As expected, CM from 2-DGtreated glioblastoma cells and GSCs reduced the migration of macrophages, including mouse Raw264.7 macrophages (Fig. 1j-l and Extended Data Fig. 2k), primary mouse bone marrow-derived macrophages (BMDMs) (Fig. 1m and Extended Data Fig. 2l), human THP-1 macrophages (Fig. 1n and Extended Data Fig. 2m), and primary human BMDMs (Fig. 1o and Extended Data Fig. 2n), relative to the CM from control cells.Together, these ndings suggest a critical role of glioblastoma cell glycolysis in regulating macrophage in ltration.
Glioblastoma cell LDHA promotes macrophage in ltration.To determine the molecular basis of tumor glycolysis in support of macrophage in ltration, we examined the connection between the expression of key glycolysis and tricarboxylic acid (TCA) cycle enzymes (e.g., HK1, HK2, HK3, PGM1, PGM2, LDHA, LDHB, MDH1, MDH2, FH, SDHA, SUCLA2, OGDH, IDH3A, IDH3B, IDH3G, CS, and ACO1) with patient survival, immune score, and macrophage signature in TCGA glioblastoma patient tumors.Following these analyses, LDHA was identi ed as the only gene that correlated negatively with patient survival and positively with immune score and macrophage signature (Extended Data Fig. 3a).Analysis of the scRNAseq data 34 also identi ed LDHA as the only gene that was increased in glioma cells of glioblastoma (including ndGBM and rGBM) compared to LGG (Fig. 2a) and correlated positively with the abundance of macrophages and monocytes in glioblastoma patient tumors (Fig. 2b).Next, GSEA on hallmark pathways with the RNA-Seq pro ling data from CT2A cells treated with or without a LDHA speci c inhibitor FX11 demonstrated that FX11-treated cells displayed an impaired representation of immune response networks including interferon alpha and gamma responses, and in ammatory response (Fig. 2c).Similarly, LDHAhigh glioblastoma patient tumors showed prominent representations of leukocyte and myeloid cell migration signatures, immune response networks, and cytokine and chemokine signatures (Table S4).Finally, GSEA on distinct immune cell signatures con rmed that macrophage and monocyte were the top immune cell types enriched in LDHA-high TCGA glioblastoma patient tumors (Extended Data Fig. 3b).
Together, these ndings support a pivotal role of glioblastoma cell/GSC LDHA in triggering macrophage in ltration into the glioblastoma TME.
Glioblastoma cell LDHA promotes macrophage in ltration via upregulating CCL2 and CCL7.GSEA on KEGG pathways of CT2A cells with FX11 treatment versus control exhibited a prominent reduction of signatures related to chemokine and cytokine-cytokine receptor interaction (Fig. 3a), suggesting that LDHA in glioblastoma cells may regulate the expression of chemokines and cytokines.To elucidate such chemokines and/or cytokines governing macrophage recruitment in LDHA-high glioblastoma cells, we examined putative factors exhibiting a ≥ 2.0-fold change in CT2A cells (FX11 treatment versus control) and TCGA glioblastoma patient tumors (LDHA-high versus -low) using a human secreted protein dataset 38 .This analysis led to identi cation of eleven genes (e.g., CCL2, CCL7, IL1B, IL1RAP, IL1RN, MMP9, NPY, PLAU, PROS1, S100A8 and SLPI) encoding secreted proteins that were upregulated in LDHA-high patient tumors compared to LDHA-low tumors and downregulated by LDHA inhibitor FX11 treatment in CT2A cells (Fig. 3b, c).To reveal the importance of these genes in glioblastoma tumor biology, we conducted bioinformatics analyses in TCGA glioblastoma tumors showing that the expression of most of these genes (except for NPY) correlated positively with macrophage signature, but only CCL2, CCL7, IL1RAP, PLAU, and S100A8 correlated negatively with patient survival (Extended Data Fig. 5a).RT-qPCR demonstrated a decreased expression of Ccl2, Ccl7, Plau, and S100a8, but not Il1rap, in CT2A and GL261 cells upon the treatment with LDHA inhibitor FX11 (Fig. 3d and Extended Data Fig. 5b).Reduced expression of Ccl2, Ccl7, Plau, and S100a8, was further con rmed by additional pharmacological (using LDHA inhibitor stiripentol) and genetic (using shLdha) strategies in CT2A cells (Fig. 3e, f) and GL261 cells (Extended Data Fig. 5c, d).To validate the capacity of CCL2, CCL7, PLAU, and S100A8 functioning as macrophage chemoattractants, we performed transwell migration assay showing that recombinant CCL2 and CCL7, but not PLAU and S100A8, protein-supplemented media increased the migration of Raw264.7 macrophages (Fig. 3g, h).Similar experiments in human GSC272 demonstrated that LDHA inhibitor stiripentol treatment reduced CCL2 and CCL7 expression (Fig. 3i, j) and secretion (Fig. 3k, l).Conversely, recombinant LDHA protein treatment increased the expression of CCL2 and CCL7 in both mouse CT2A cells and human GSC272 and rescued the impaired CCL2 and CCL7 levels in shLdha CT2A cells (Extended Data Fig. 5e-h).Consistent with the data from mouse Raw264.7 macrophages, recombinant CCL2 and CCL7 protein-supplemented media increased the migration of human THP-1 macrophages (Fig. 3m and Extended Data Fig. 5i).More importantly, the impaired macrophage migration induced by CM from shLdha CT2A cells was prevented by the treatment with recombinant CCL2 and CCL7 proteins (Extended Data Fig. 5j-m).The GOEA on the sub-ontologies of Biological Process and Molecular Function, and KEGG Enrichment Analysis in TCGA glioblastoma patient tumors also demonstrated that the migration of leukocytes and/or myeloid cells and the activity of chemokines and cytokines were the top CCL2-and CCL7-regulated processes (Tables S5 and S6).
To further con rm the role of glioblastoma cell CCL2 and CCL7 in macrophage in ltration, we rst employed shRNAs to deplete CCL2 and CCL7 in CT2A and GL261 cells.As expected, CM from CT2A and GL261 cells expressing shCcl2 (Fig. 3n, o and Extended Data Fig. 6a-d) and shCcl7 (Fig. 3p, q and Extended Data Fig. 6e-h) induced signi cantly less macrophage migration than CM from shRNA control (shC) cells.Next, we depleted CCL2 and CCL7 in GSC272 and con rmed that CM from GSC272 expressing shCCL2 (Fig. 3r, s and Extended Data Fig. 6i, j) and shCCL7 (Fig. 3t, u and Extended Data Fig. 6k, l) inhibited the migration of THP-1 macrophages and primary human BMDMs compared to CM from shC cells.In summary, these results reinforce that the expression of CCL2 and CCL7 in glioblastoma cells/GSCs is regulated by LDHA and that glioblastoma cell CCL2 and CCL7 function as potent macrophage chemoattractants.
To investigate the potential functional relevance of HOXA9, YAP1, and STAT3 in regulating CCL2 and CCL7 expression and macrophage in ltration, bioinformatics analyses in TCGA glioblastoma patient tumors were performed.As a result, we found that the expression of YAP1 and STAT3, but not HOXA9 and TP53, correlated positively with CCL2, CCL7, and macrophage signature (Extended Data Fig. 7l).Then, CT2A cells, GL261 cells, and GSC272 were treated with YAP1-TEAD interaction inhibitor vertepor n 39 and STAT3 inhibitor WP1066.The results of these experiments demonstrated that vertepor n treatment reduced P-STAT3, and, reciprocally, WP1066 treatment impaired YAP1 expression at both mRNA and protein levels (Fig. 4j, k).Moreover, the nuclear localization of STAT3 was reduced when CT2A cells harboring shLdha or treated with stiripentol and vertepor n (Extended Data Fig. 7m, n).Similarly, depletion of LDHA or treatment with stiripentol and WP1066 in CT2A cells reduced the nuclear localization of YAP1 (Extended Data Fig. 7o, p).These ndings suggest that LDHA-regulated YAP1 and STAT3 are transcriptional co-activators 40 , prompting us to investigate the role of YAP1 and STAT3 in transcriptional regulation of CCL2 and CCL7 in glioblastoma cells.Correspondingly, we observed speci c YAP1 and STAT3 binding to the Ccl2 and Ccl7 promoters in CT2A cells, which was reduced upon LDHA depletion (Fig. 4l, m).Moreover, pharmacological treatment with vertepor n and WP1066 in CT2A and GL261 cells repressed Ccl2 and Ccl7 expression (Fig. 4n, o).To further investigate whether LDHAregulated lactate contributes to this process, we treated LDHA-depleted glioblastoma cells with lactate and found that this treatment rescued the impaired signaling of P-ERK, YAP1, P-STAT3, CCL2, and CCL7 in shLdha CT2A cells (Extended Data Fig. 7q-s).Together, these ndings suggest that YAP1 and STAT3 transcriptional co-activators contribute to LDHA/lactate-ERK axis-dependent CCL2 and CCL7 expression in glioblastoma cells.
Macrophage-derived LDHA-containing EVs promote tumor growth and activate the ERK-YAP1/STAT3-CCL2/CCL7 axis in glioblastoma cells.Once in ltrating into the glioblastoma TME, macrophages are educated to promote glioblastoma progression by secreting distinct factors and EVs 8 .To mimic this process, we rst utilized glioblastoma cell CM to educate macrophages (hereafter such educated macrophages are referred to as EMφ), and then examined the role of CM from EMφ on glioblastoma cells.As a result, we found that EMφ CM promoted LDHA expression in CT2A and GL261 cells (Fig. 5a, b), prompting a speculation that TAMs may support glioblastoma cell growth and survival via upregulating LDHA.To con rm the role of LDHA in glioblastoma cell biology, we performed cell cycle, apoptosis, and proliferation analyses in glioblastoma cells with or without LDHA inhibition.We found that CT2A cells expressing shLdha or treated with LDHA inhibitors (e.g., isosafrole, FX11 or stiripentol) displayed decreased G1 and upregulated G2-M fractions (Extended Data Fig. 8a-d), upregulated apoptosis (Extended Data Fig. 8e-h), and reduced proliferation (Extended Data Fig. 8i-l).
To reveal how TAMs upregulate LDHA in glioblastoma cells, we depleted LDHA using shRNAs (Extended Data Fig. 9a) and inhibited LDHA using FX-11 in EMφ.Surprisingly, we noticed that LDHA inhibition in macrophages abolished EMφ CM-induced LDHA upregulation in glioblastoma cells (Fig. 5a, b), suggesting a potential for LDHA delivery from EMφ to glioblastoma cells.scRNA-seq data analysis on tumors from a cohort of four glioblastoma patients 41 demonstrated that LDHA was highly expressed in both glioblastoma cells and CD68 + CX3CR1 − macrophages, but not in CD68 + CX3CR1 + microglia (Extended Data Fig. 9b-e).As expected, genetic and pharmacological inhibition of LDHA in macrophages reduced glycolysis as shown by the impaired lactate production (Extended Data Fig. 9f, g) and ECAR (Extended Data Fig. 9h, i).To investigate whether LDHA could be delivered from macrophages into glioblastoma cells via EVs, we treated EMφ with GW4869 (an EV biogenesis and release inhibitor) and found that this treatment abolished EMφ CM-induced LDHA upregulation in both CT2A and GL261 cells (Fig. 5c, d).Next, we puri ed EVs from CM of EMφ using ultracentrifugation and con rmed their identity nanoparticle tracking analysis (Extended Data Fig. 10a) and Western blotting for EV markers (e.g., CD63 and ALIX) and calnexin that is absent from EVs (Fig. 5e).Notably, CT2A and GL261 CM treatment did not affect the size and distribution of macrophage-derived EVs (Extended Data Fig. 10a), but increased LDHA levels in EMφ EVs, an effect that was abolished by shRNA-mediated Ldha depletion in macrophages (Fig. 5e).Moreover, EMφ EVs were labeled with the uorescent dye 1,1′-dioctadecyl-3,3,3′,3′tetramethylindocarbocyanine perchlorate (DiD) and then incubated with glioblastoma cells.Recipient glioblastoma cells exhibited equal uptake e ciency for EVs from control macrophages, as well as CT2A EMφ and GL261 EMφ expressing shC and shLdha (Extended Data Fig. 10b-e).However, LDHA in glioblastoma cells was upregulated upon uptake of EVs from shC EMφ, but not from LDHA-depleted EMφ (Fig. 5f, g and Extended Data Fig. 10f, g).These ndings reinforce the role of EVs in delivery of LDHA from EMφ to glioblastoma cells.
Next, we aimed to investigate the role of EMφ-derived EVs in regulating glioblastoma cell growth and survival.Among a series of cellular analyses in EMφ EV-treated glioblastoma cells, we found that EMφ EV treatment abolished Ldha knockdown-induced cell cycle transition from G1 to G2/M and apoptosis in CT2A (Extended Data Fig. 11a-d) and GL261 (Extended Data Fig. 11e-h) cells.Similarly, LDHA inhibitor stiripentol treatment-induced cell cycle transition from G1 to G2/M and apoptosis in CT2A and GL261 cells were rescued by the treatment with EVs from EMφ (Fig. 5h, i and Extended Data Fig. 12a-f).However, such effects were abolished or inhibited by depletion of LDHA in macrophages (Fig. 5h, i and Extended Data Fig. 12a-f).
Given the importance of glycolysis in tumor-macrophage symbiosis in glioblastoma, we further investigated the potential role of EMφ EVs in this process.Notably, we found that the impaired glycolytic activity as shown by reduced ECAR in LDHA-depleted glioblastoma cells was negated by the treatment with EVs from EMφ, but not LDHA-depleted EMφ (Fig. 5j).Moreover, treatment with EMφ EVs upregulated the levels of P-ERK, YAP1, P-STAT3, CCL2, and CCL7 in glioblastoma cells (Extended Data Fig. 12g, h).The decreased P-ERK, YAP1, P-STAT3, CCL2, and CCL7 in LDHA-depleted glioblastoma cells was rescued by the treatment with EMφ EVs, but not LDHA-depleted EMφ EVs (Fig. 5k-m).Together, these results demonstrate that TAM-derived LDHA-containing EVs can promote tumor growth by triggering a positive feedback loop between glioblastoma cell glycolysis and macrophage in ltration.
Inhibition of LDHA-regulated tumor-macrophage symbiosis extends survival in glioblastoma mouse models.To further investigate the role of LDHA-mediated tumor-macrophage interplay in glioblastoma tumor biology, we utilized shRNA knockdown system to deplete LDHA in CT2A and GL261 tumors implanted into C57BL/6 mice.We found that LDHA depletion signi cantly inhibited tumor growth and extended survival in both glioblastoma mouse models (Fig. 6a, b and Extended Data Fig. 13a, b).Given the brain-penetrating ability of LDHA inhibitor stiripentol and isosafrole 42 , we developed preclinical trials evaluating the anti-tumor effect of pharmacological inhibition of LDHA in glioblastoma mouse models.We found that stiripentol and isosafrole treatment impaired tumor growth and extended the survival of C57BL/6 mice implanted with CT2A cells, GL261 cells, and 005 GSCs (Fig. 6c-e and Extended Data Fig. 13c-e).Moreover, we developed a patient-derived xenograft (PDX) model in nude mice by intracranial implantation of GSC272 and found that stiripentol treatment also extended survival (Fig. 6f).To con rm that macrophages were the critical target of stiripentol in impairing tumor growth and progression, we compared the anti-tumor effect of stiripentol and BLZ945 (an CSF-1R inhibitor that can impair macrophage role in mice) in GL261-bearing mice.Each agent extended survival; however, their combination treatment did not exhibit additional anti-tumor effects (Extended Data Fig. 13f).On the histological level, immuno uorescence for Ki67 and cleaved caspase 3 (CC3) demonstrated that glioblastoma cell proliferation was dramatically reduced, whereas apoptosis was increased upon Ldha depletion (Extended Data Fig. 14a-d) and treatment with stiripentol and isosafrole (Extended Data Fig. 14e-h).Flow cytometry demonstrated that macrophages were profoundly reduced in LDHA-depleted CT2A tumors (Extended Data Fig. 14i-k) and LDHA inhibitor-treated GL261 (Fig. 6g, h), CT2A (Fig. 6i and Extended Data Fig. 14l-n) and 005 GSC tumors (Extended Data Fig. 14o, p).Similarly, immuno uorescence for F4/80 con rmed that in ltrating macrophages were profoundly reduced in CT2A tumors by inhibition of LDHA genetically (Extended Data Fig. 14q, r) and pharmacologically (Extended Data Fig. 14s, t).However, LDHA inhibition with stiripentol did not change macrophage apoptosis in CT2A tumors (Extended Data Fig. 14u, v).
To con rm the role of LDHA in regulation of the YAP1/STAT3-CCL2/CCL7 signaling axis in vivo, we performed immuno uorescence for YAP1 and STAT3 in tumors and ELISA for CCL2 and CCL7 in plasma from control and LDHA-inhibited glioblastoma tumor-bearing mice.We found that the nuclear level of YAP1 and STAT3 in CT2A tumors (Fig. 6j-m) and plasma level of CCL2 and CCL7 from glioblastoma tumor-bearing mice (Fig. 6n, o) were signi cantly reduced upon stiripentol treatment.Similarly, blockade of the YAP1/STAT3 signaling using STAT3 inhibitor WP1066 reduced plasma level of CCL2 and CCL7 and intratumoral macrophages in CT2A-bearing mice (Fig. 6p-r and Extended Data Fig. 14w).Next, we investigated the in vivo role of CCL2 and CCL7 by implantation of shC, shCcl2, and shCcl7 CT2A cells into the brains of C57BL/6 mice and found that depletion of CCL2 and CCL7 signi cantly extended survival (Fig. 6s) and reduced intratumoral macrophages (Fig. 6t and Extended Data Fig. 14x).
Finally, we aimed to investigate the role of TAM-derived LDHA-containing EVs in glioblastoma progression and treated shLdha CT2A-bearing mice with stiripentol and EMφ EVs.As expected, stiripentol treatment did not exhibit additional anti-tumor effects in LDHA-depleted tumors (Fig. 6u), supporting our above in vivo ndings that LDHA is the key target of stiripentol.However, EMφ EVs treatment rescued the impaired tumor growth in LDHA-depleted CT2A tumors (Fig. 6u).To con rm the role of macrophage LDHA in this process, we generated macrophage-speci c LDHA null (LDHA-mKO) mice by crossing LDHA ox mice with Lysozyme-Cre (LyzCre) mice.Orthotopic transplantation of CT2A cells into the brains of LDHA-mKO and WT mice showed signi cant survival extension in LDHA-mKO mice compared to WT mice (Fig. 6v).However, stiripentol treatment showed similar anti-tumor effects in both WT and LDHA-mKO mice (Fig. 6v).Together, these results validate the importance of LDHA-regulated tumormacrophage symbiosis in promoting glioblastoma progression and support a therapeutic potential of targeting this co-dependency in glioblastoma.
The LDHA-YAP1/STAT3-CCL2/CCL7 axis tracks with macrophages in glioblastoma patient tumors and is increased in glioblastoma patient plasma and EVs.The clinical relevance of above experimental ndings was supported by bioinformatics using scRNA-seq data from 16 glioblastoma patients 34 showing that glioblastoma cell LDHA, YAP1, STAT3, and CCL2 correlated positively with macrophage abundance (Fig. 7a).Moreover, bioinformatics analyses in TCGA glioblastoma dataset con rmed that LDHA, YAP1, STAT3, CCL2, and CCL7 positively correlated with each other and with macrophage signature in patient tumors (Fig. 7b).Next, we performed immuno uorescence for LDHA and Mac-2 (a macrophage marker) in tumors from a cohort of 30 glioblastoma patients and found that LDHA signaling showed a positive correlation with the density of intratumoral macrophages (Fig. 7c, d).Since LDHA, CCL2 and CCL7 are genes encoding secreted proteins, we compared their protein levels in patient plasma showing that all of them were higher in glioblastoma patients than that in healthy controls, but such levels were not changed in meningioma patients (Fig. 7e-g).Moreover, plasma LDHA correlated positively with plasma CCL2 and CCL7 (Fig. 7h, i) and intratumoral macrophages in glioblastoma (Fig. 7j).Moreover, the median survival time of glioblastoma patients with high plasma LDHA (389 days) was lower than the patients with low plasma LDHA (675 days, Fig. 7k).However, it should be noted that this survival analysis did not reach statistical signi cance due to limited patient numbers (Fig. 7k).Moreover, plasma LDHA, CCL2, and CCL7 levels were not related to the status of recurrence, gender, age, and MGMT methylation in glioblastoma patients (Extended Data Fig. 16).Finally, we examined the levels of LDHA in plasma EVs from a cohort of healthy controls and glioblastoma patients and con rmed that LDHA in glioblastoma patient plasma EVs was signi cantly higher than that from healthy controls (Fig. 7l, m).
Together, these correlative glioblastoma patient's ndings are consistent with the hypothesis that LDHA-YAP1/STAT3-CCL2/CCL7 axis drives macrophage in ltration, and suggest that LDHA, CCL2 and CCL7 might function as biomarkers for glioblastoma patients, although these data are still relatively preliminary.

DISCUSSION
Glioblastoma cells can reprogram metabolic pathways to maintain their tumor potential.Aerobic glycolysis is used in tumors across cancer types (including glioblastoma) and is considered a hallmark of cancer 15 .However, whether and how aerobic glycolysis affects the biology of immune cells, such as macrophages, and, in turn, modulates tumor immunity and progression are not determined in glioblastoma.In this study, we screened a panel of metabolic small-molecule compounds and demonstrated that glioblastoma cell glycolysis is essential for macrophage in ltration.Mechanistically, LDHA-lactate-directed ERK pathway activates YAP1 and STAT3 transcriptional co-activators to upregulate CCL2 and CCL7 in glioblastoma cells, which promote macrophage in ltration into the TME.In addition to functioning as immunosuppressive cells inhibiting anti-tumor immunity 6,44 , TAMs are known to promote glioblastoma cell proliferation and survival 3,6,8 .We provided further evidence showing that these in ltrating macrophages promote glioblastoma cell glycolysis, proliferation, survival, and tumor growth through secretion of LDHA-containing EVs.Clinical validations demonstrated that the intratumoral LDHA-YAP1/STAT3-CCL2/CCL7 signaling axis and plasma LDHA track with macrophage density and may function as potential biomarkers for glioblastoma patients.Therefore, our current work reveals the molecular mechanisms underlying tumor-macrophage symbiosis and supports the hypothesis that targeting this LDHA-mediated symbiosis could provide clinical bene ts for glioblastoma patients (Fig. 8).
Emerging evidence has shown that tumor-macrophage symbiotic interactions are critical for tumor progression 6, 8, 45 .Cancer cell metabolism not only provides su cient energy for maintaining tumor growth but also affects the biology of myeloid cells (e.g., macrophages 19,20,21 ) across cancer types, including glioblastoma 46,47 .LDHA is an aerobic glycolysis-related key enzyme contributing to lactate production in cancer cells 24,25 .Upon secretion, lactate plays an important role in regulating macrophage immunosuppressive polarization across cancer types, including breast cancer 48,49 , lung cancer 50, 51, 52 , melanoma 52 , cervical cancer 53 , and colon cancer 52 .In our study, we established that LDHA-mediated glioblastoma cell glycolysis promotes the in ltration of macrophages into to the TME, which, in turn, supports tumor progression in glioblastoma mouse models.These results are consistent with the ndings observed in multiple sclerosis, where enhanced glycolytic metabolism triggers the in ltration of macrophages 54 .Together, our work reinforces the importance of glioblastoma cell glycolysis in modulating the TME, particularly the in ltration of macrophages.
In exploring the connection between LDHA and macrophage biology, we demonstrated that glioblastoma cell LDHA upregulates multiple downstream chemokines, most prominently CCL2 and CCL7, to trigger macrophage in ltration, consistent with previous work 44,55 .Mechanistically, our study demonstrated that these two chemokines are regulated by LDHA/lactate-induced activation of YAP1 and STAT3 in glioblastoma cells.Moreover, we discovered that the ERK pathway is required for LDHA-induced YAP1 and STAT3 activation, which is consistent with previous work showing that ERK is a downstream of LDHA in the heart 56 and breast cancer cells 57 .YAP1 is a transcription coregulator that plays a vital role in tumor progression 58 .In the context of glioblastoma, we have shown that YAP1 is essential for PTEN de ciency-induced transcriptional upregulation of LOX, which, in turn, triggers macrophage in ltration into the TME 3 .Here, we further identi ed that YAP1 activation promotes macrophage in ltration via direct transcriptional regulation of CCL2 and CCL7 in glioblastoma cells, consistent with the ndings observed in liver cancer 59,60 .These distinct YAP1-driven mechanisms underlying macrophage recruitment highlight a context-dependent tumor-macrophage symbiosis and the need for developing personalized medicine to target this symbiosis.STAT3 is a transcription factor that plays a critical role in regulating macrophage immunosuppressive polarization 61,62  After dissecting the molecular mechanisms underlying tumor-macrophage symbiosis, we investigated the biological and clinical impact of targeting this symbiosis in glioblastoma.We have shown that genetic depletion of LDHA in glioblastoma cells or macrophages extends survival, reduces macrophage in ltration and glioblastoma cell proliferation, and promotes glioblastoma cell survival in mouse models.
In line with these ndings from mouse models, analysis of tumor and plasma samples from glioblastoma patients demonstrates that the LDHA-YAP1/STAT3-CCL2/CCL7 signaling axis tracks with macrophages.Together, the identi cation of tumor-macrophage symbiosis, coupled with the anti-tumor effect of LDHA inhibition in glioblastoma mouse models and clinical validations, encourages the development of therapeutic strategies targeting this symbiosis in glioblastoma patients.Emerging evidence highlights that pharmacological targeting of tumor-macrophage symbiosis is a promising strategy for glioblastoma treatment, and multiple approaches, including CSF-1R inhibition, have been proposed 5 .Previous studies have shown that CSF-1R inhibitors can impair tumor progression and decrease immunosuppressive macrophages in glioblastoma mouse models 71,72,73 .However, these treatments result in therapy resistance due to enhanced PI3K activity in glioblastoma cells driven by macrophage-derived insulin-like growth factor-1 (IGF-1) 73 .Correspondingly, clinical trials with CSF-1R inhibition failed in patients with glioblastoma 74 and resulted in serious side effects since CSF-1R is also expressed on monocytes and other stromal cells 75 .In this study, we developed preclinical trials in glioblastoma mouse and PDX models with LDHA inhibitors stiripentol and isosafrole 42 and found that these treatments extend the survival of tumor-bearing mice via blockade of tumor-macrophage symbiosis.Stiripentol is an FDA-approved antiepileptic drug for Dravet syndrome, a severe genetic brain disorder 76,77 .Isosafrole is a stiripentol analog that signi cantly inhibits the pyruvate-to-lactate conversion and suppresses seizures in a mouse model of epilepsy 42 .Based on the nature (e.g., welltolerated in patients and BBB penetrating ability) of the two compounds, coupled with their anti-tumor effect in glioblastoma mouse and PDX models, we anticipate a tremendous translational potential of LDHA inhibition to improve patient outcomes. In

DECLARATION OF INTERESTS
No potential con icts of interest were disclosed by the authors.

Isolation and culture of primary BMDMs
Primary mouse BMDMs were isolated from C57BL/6 mice and cultured as we described previously 3,49 .
Conditioned media (CM) from glioblastoma cells and GSCs or normal medium with indicated factors were added to the remaining receiver wells.After 8 hrs (Raw264.7 macrophages and mouse primary BMDMs) or 16 hrs (THP-1 macrophages and human primary BMDMs), the migrated macrophages were xed and stained with crystal violet (0.05%, Sigma, #C-3886), and then cells per eld of view were counted under the microscope.Moreover, we performed the scratch would healing assay on macrophages treated with or without CM from control and LDHA-depleted/inhibited glioblastoma cells using a protocol as we reported previously 78 .

Metabolic compound screen
For the initial screening, CT2A cells were seeded in 6-well plated and treated with 55 compounds with metabolic reprogramming function from the CNS-Penetrant Compound Library (MCE MedChemExpress, #HY-L028) at 10 mM for 24 hrs.After the treatment with the compounds for 24 hrs, CT2A cells were then cultured with FBS-and compound-free culture medium for additional 24 hrs.The conditioned media (CM) from number-matched control and compound-treated CT2A cells were collected and used for Raw264.7 macrophage transwell migration assay.The compounds with a signi cant effect on inhibiting CT2A cell CM-induced macrophage migration were selected for a second round of screen at a lower concentration (5 mM).

Colony formation assay
Colony formation assay was used to examine glioblastoma cell proliferation in vitro.In brief, 1500 glioblastoma cells were seeded and cultured for about 8 days in each well of 6-well plates.Finally, cells were xed and stained with 0.5% crystal violet for 1 hr.These experiments were performed in triplicate.

Cell cycle and apoptosis analysis
Cells were cultured in 6-well plates for 24 hrs, and xed in ice-cold 70% ethyl alcohol for 30 min at 4 °C.For cell cycle analysis, cells were incubated with RNase A solution (Promega, #A797C; 100 µg/ml) for 5 min at room temperature and then stained with propidium iodide (PI) labeled with RedX (Biolegend, #421301, 50 mg/ml) for 10 min at 4 °C.PI incorporation was analyzed by ow cytometry.For apoptosis analysis, cells were incubated with FITC-conjugated annexin V (BioLegend, #640906) and PI labeled with RedX (1 mg/ml) for 15 min at room temperature and analyzed using a ow cytometer.

Metabolic assays
Lactate levels were measured using a glycolysis assay kit (Sigma-Aldrich, #MAK439) according to the instruction.Brie y, control and glycoysis/LDHA-inhibited cells were seeded in 96-well plate at a density of 3 × 10 5 cells and cultured in 1% FBS culture media containing with or without glucose (55 mM).
Following the collection of CM, glycolytic activity (Lactate level) was measured at different time intervals for 1.5 hrs at 565 nm wavelength.On the other hand, glucose metabolism of indicated control and/or treated/modi ed glioblastoma cells and macrophages was measured using the Seahorse XF Cell Mito Stress Test Kit (Agilent Technologies, #103015-100) in a Seahorse XFe96 analyzer on Seahorse XFe96/XF Pro FluxPak Mini plates (Agilent Technologies, #103793-100) as instructed by the manufacturer's protocol.

Extracellular vesicle isolation
For EV isolation from cells, indicated cells were grown with vesicle-depleted FBS for 24 hrs, and conditioned media were collected and centrifuged at 300 x g for 10 min, 2000 x g for 10 min, and 10000 x g for 30 min to remove cell debris.The supernatant was ltered through a 0.2 mm lter and centrifuged at 100,000 x g 4 ℃ for 70 min.The pellets were resuspended in cold-cold PBS and applied for the second round of ultracentrifugation.Finally, the pellets containing EVs were resuspended in 100 ml ice-cold PBS for further experiments.For EV isolation from human plasma, the SmartSEC Single EV Isolation System (System Biosciences, #SSEC200A-1) was used according to the manufacturer's instructions.Brie y, plasma samples with an additional column buffer of up to 4 ml were placed directly into the pre-washed column, incubated for 30 min at room temperature, and centrifuged at 500 xg for 1 min to elute the EVs in the ow through.The EV's supernatant was used for further ow cytometry analysis.

Nanoparticle tracking analysis
Concentrated EVs were diluted using freshly ltered PBS and analyzed using a NanoSight NS3000 device (Nanosight, Malvern).A monochromatic laser beam at 405 nm was set to analyze the nanoparticles, and a video with a 30-second duration was taken at a rate of 30 frames per second.Approximately 30-100 particles were analyzed in each eld of view, and then particle brown-movement was assessed using the nanoparticle tracking analysis (NTA) software (version 2.3, Nanosight).NTA post-acquisition settings were optimized, and recorded video was analyzed to measure particle sizes and concentrations.

Quantitative real-time PCR (RT-qPCR)
Cells were detached with trypsin (Gibco, #25300-054) and pelleted.RNA was isolated using the RNeasy Mini Kit (Qiagen, #74106), and then reverse-transcribed into cDNA with the All-In-One 5X RT MasterMix (Applied Biological Materials, #G592).PCR was performed using the SYBR Green PCR Master Mix (Bio-Rad, #1725275).Approximately 10 ng of template was used per PCR reaction.The expression of each gene was quanti ed using the ΔΔCt method and normalized to the housekeeping gene (e.g., ACTB or GAPDH).PCR was run using the CFX Connect Real-Time PCR Detection System (Bio-Rad, #1855201).
Primers are listed in Table S7.

Hematoxylin and Eosin (H&E) staining
Staining was performed using the H&E staining kit (Abcam, #ab245880) according to a standard protocol.Brie y, tumor sections were incubated with hematoxylin, Mayer's (Lillie's Modi cation) for 5 min after washing two times in distilled water, and then incubated with the Bluing Reagent and Eosin Y Solution (Modi ed Alcoholic) for15 sec and 3 min, respectively.The images of tumor sections were captured using TissueFAXS in the Center for Advanced Microscopy (CAM) at Northwestern University.

ChIP-PCR
ChIP-PCR was performed using the commercial Pierce TM Magnetic CHIP kit (ThermoFisher, #26157) according to the manufacturer's instructions.Brie y, control and shLdha CT2A cells were cross-linked using 1% PFA (10 min), and then reactions were quenched with glycine (5 min) at room temperature.Cells were lysed with ChIP lysis buffer for 30 min on ice.Chromatin fragmentation was performed using a sonicator.Solubilized chromatin was then incubated with a mixture of YAP1 antibodies (CST, #14074S) or STAT3 (CST, #9139S) antibodies and Dynabeads (Life Technologies) overnight.Immune complexes were washed with RIPA buffer three times, once with RIPA-500, and once with LiCl wash buffer.Elution and reverse-crosslinking were performed in direct elution buffer containing proteinase K (20 mg/ml) at 65 °C overnight.Eluted DNA was used to perform qPCR.The primers were designed according to the E-box of mouse Ccl2 and Ccl7 genes.Primers are listed in Table S7.

Microarray and RNA-Seq analysis
The gene expression in human glioblastoma was analyzed using gene-pro ling data from the microarray TCGA datasets.For RNA-seq analysis, the total RNA of control and FX11-treated CT2A cells was extracted using RNeasy Kit (Qiagen, #74034).RNA-seq was performed by the Genomics Facility at the University of Chicago.Oligo-dT based library was prepared and samples were sequenced by novaseq 6000 sequencer.
Raw data were mapped to the mouse genome.The transcriptome of each gene in control and FX11treated groups was further quanti ed.GSEA was used for pathway analyses based on differentially expressed genes of these two groups.

scRNA-seq data analysis
For the analysis of scRNA-seq data from glioblastoma patient tumors, low-quality cells with detected genes < 500, and mitochondrial genes > 20% were removed.Batch effected was removed by CCA-based integration method in Seurat 80 .Both canonical genes and cluster differential genes were used to identify the cell types.scRNA-seq of GEO, GSE84465 41 , was used to perform unsupervised sub-clustering for macrophages and microglia [CD68 and CX3CR1 were selected as the positive control for TAM (macrophage + microglia) and microglia clustering, respectively].The expression of LDHA in macrophages, microglia, and other tumor cells was investigated.Next, the scRNA-seq data of GEO, GSE131928 34 , were used to analyze the connection among glioblastoma cell glycolysis (including glycolysis signature and key enzymes) and myeloid cells (including macrophage, monocyte, microglia, and DC) in patient tumors.The average expression of each gene and gene signature was represented by color (low to high was shown as blue to red).

Computational analysis of human glioblastoma datasets
For analysis of human glioblastoma data, we downloaded the microarray gene expression and survival data of TCGA dataset or other available datasets from GlioVis: http://gliovis.bioinfo.cnio.es/.The gene expression, signature expression, correlation, survival analyses, and GSEA of interesting gene signatures in IDH-WT glioblastoma patients were performed as we reported previously 3,12,13 .

Patient samples
Peripheral blood plasma from meningioma (n = 15) and glioblastoma (n = 54) patients, and tumor samples (n = 30) from surgically resected IDH-WT glioblastomas were collected at the Northwestern Central Nervous System Tissue Bank (NSTB).All patients were diagnosed according to the WHO diagnostic criteria by neuropathologist Dr. Craig Horbinski.Detailed patient information is provided in Table S8.For control plasma (n = 15), we used commercially available anonymized and de-identi ed, which were isolated from healthy human blood (Solomonpark, #4345).According to The George Washington University Institutional Review Board and based on the guidelines from the O ce of Human Research Protection, the conducted research meets the criteria for exemption #4 (45 CFR 46.101(b) Categories of Exempt Human Subjects Research) and does not constitute human research.

Statistical analysis
Statistical analyses were performed with Student t-tests for comparisons between two groups or one-way ANOVA tests for comparisons among groups.Data was represented as mean ± SD or SEM as indicated.The survival and correlation analyses in brain cancer datasets (including TCGA dataset) and animal models were performed using the Log-rank (Mantel-Cox) test and the Pearson's correlation test, respectively (GraphPad Prism 9).P < 0.05 was considered signi cant.(i, j) Quanti cation of relative migration of primary mouse bone-marrow-derived macrophages (BMDMs, i) and Raw264.7 macrophages (j) from a transwell analysis following stimulation with CM from CT2A and GL261 cells, respectively, treated with or without stiripentol (10 mM).n = 5 biological replicates.

Figures
(k, l) Quanti cation of relative migration of THP-1 macrophages (k) and primary human BMDMs (l) from a transwell analysis following stimulation with CM from GSC272 treated with or without stiripentol (10 mM).n = 5 biological replicates.(r, Quanti cation of relative migration of THP-1 macrophages (r) and primary human BMDMs (s) following stimulation with conditioned media from GSC272 expressing shC and shCCL2.n = 5 biological replicates.
(t, u) Quanti cation of relative migration of THP-1 macrophages (t) and primary human BMDMs (u) following stimulation with conditioned media from GSC272 expressing shC and shCCL7.n = 5 biological replicates.
Data presented as mean ± SEM and analysed by Student's t-test (d, e, i, k, l) and one-way ANOVA test (f, h, m, n, o, p, q, r, s, t, u).(p, q) The plasma level of CCL2 (p) and CCL7 (q) in CT2A tumor-bearing mice treated with or without STAT6 inhibitor WP1066 (60 mg/kg, oral gavage, every other day for 6 doses).n = 3 biological replicates.
(r) Quanti cation of ow cytometry analysis for the percentage of CD68 + macrophages out of CD45 high CD11b + cells in CT2A tumors treated with or without WP1066.n = 3 biological replicates.(a) The correlation of glioblastoma cell LDHA, YAP1, STAT3, CCL2, and CCL7 with macrophage abundance in glioblastoma patient tumors based on single-cell RNA sequencing data 34 .R and P values

Figure 2
Figure 2 (m, n) Quanti cation of relative migration of Raw264.7 macrophages from a transwell analysis following stimulation with CM from CT2A (m) or GL261 (n) cells treated with or without FX11 (8 mM).n=5 biological replicates.(o,p) cation of relative migration of THP-1 macrophages (o) and primary human BMDMs (p) from a transwell analysis following stimulation with CM from GSC272 treated with or without FX11 (8 mM).n=5 biological replicates.Data presented as mean ± SEM.Statistical analyses were determined by Pearson's correlation test (b) and one-way ANOVA test (f-p).

4 LDHA
-induced CCL2 and CCL7 expression is regulated by YAP1 and STAT3 transcriptional co-activators.(a)Identi cation of oncogenic pathways (using GSEA) that are downregulated by FX11 treatment in CT2A cells and enriched in LDHA-high glioblastoma patient tumors.Based on these two parameters,35

Figure 7 LDHA
Figure 7 22It is interesting to highlight the previous work showing that STAT3 transcriptionally upregulates LDHA in thyroid and bladder cancer cells63, 64.Together with our ndings, these work supports a reciprocal regulatory mechanism between LDHA and STAT3, which may induce a potent feedback loop to promote macrophage in ltration through CCL2 and CCL7 production.Our results of CCL2 and CCL7 as downstream signals of STAT3 are consistent with previous work focusing on broblasts in breast cancer65and on muscle satellite cells in injured muscles66.Consistent with previous report40, our work highlights that STAT3 and YAP1 are transcriptional coactivators that coordinately upregulate CCL2 and CCL7 in glioblastoma cells, thus stimulating macrophage in ltration into the TME.Macrophages are the most prominent immune cells in the glioblastoma TME.As a result of in ltration, they promote tumor growth and progression by secreting distinct soluble factors, including various growth factors, cytokines, and EVs 8,22.EVs can transfer proteins, RNA, microRNAs, DNA, and metabolites from parent cells to recipient cells, thus promoting tumor progression67.In our study, analysis of scRNAseq data from glioblastoma patient tumors demonstrated that LDHA is highly expressed by both glioblastoma cells and macrophages.Functional studies demonstrated that EMφ CM treatment upregulates LDHA levels in glioblastoma cells, and this effect is abolished when EMφ were pretreated with EV biogenesis inhibitor, LDHA inhibitor, or harboring LDHA knockdown/KO, suggesting that LDHA can be transferred from EMφ to glioblastoma cells.In addition to supporting previous studies focusing on a cell-autonomous role of LDHA in cancer cells 68 , including glioblastoma cells 69, 70 , our work reinforces the view that LDHA is a key molecule controlling the symbiotic interactions between glioblastoma cells and macrophages, and highlights the critical role of this symbiosis in promoting glioblastoma cell proliferation and survival.
PCR assay.K.F. and R.G.W isolated and cultured primary human BMDMs.L.K.B. and J.M. helped with the seahorse assay.K.M. and C.M.H. provided help with human patient samples.D.A.W. and M.S.L. contributed to comment this research and manuscript.P.C. designed the project, analyzed data, and wrote the manuscript.All authors participated in editing the manuscript.