Reprogramming of arachidonate metabolism confers drug resistance to glioblastoma through Enhancing Mitochondrial Activity in Fatty Acid Oxidation

Sp1 is involved in the recurrence of glioblastoma (GBM) due to the acquirement of resistance to temozolomide (TMZ). Particularly, the role of Sp1 in metabolic reprogramming for drug resistance remains unknown. RNA-Seq and mass spectrometry were used to analyze gene expression and metabolites amounts in paired GBM specimens (primary vs. recurrent) and in paired GBM cells (sensitive vs. resistant). ω-3/6 fatty acid and arachidonic acid (AA) metabolism in GBM patients were analyzed by targeted metabolome. Mitochondrial functions were determined by Seahorse XF Mito Stress Test, RNA-Seq, metabolome and substrate utilization for producing ATP. Therapeutic options targeting prostaglandin (PG) E2 in TMZ-resistant GBM were validated in vitro and in vivo.


Background
Sp1 is involved in the recurrence of glioblastoma (GBM) due to the acquirement of resistance to temozolomide (TMZ). Particularly, the role of Sp1 in metabolic reprogramming for drug resistance remains unknown.
Methods RNA-Seq and mass spectrometry were used to analyze gene expression and metabolites amounts in paired GBM specimens (primary vs. recurrent) and in paired GBM cells (sensitive vs. resistant). ω-3/6 fatty acid and arachidonic acid (AA) metabolism in GBM patients were analyzed by targeted metabolome. Mitochondrial functions were determined by Seahorse XF Mito Stress Test, RNA-Seq, metabolome and substrate utilization for producing ATP. Therapeutic options targeting prostaglandin (PG) E2 in TMZ-resistant GBM were validated in vitro and in vivo.

Results
Among the metabolic pathways, Sp1 increased the prostaglandin-endoperoxide synthase 2 expression and PGE2 production in TMZ-resistant GBM. Mitochondrial genes and metabolites were obviously increased by PGE2, and these characteristics were required for developing resistance in GBM cells. For inducing TMZ resistance, PGE2 activated mitochondrial functions, including fatty acid β-oxidation (FAO) and tricarboxylic acid (TCA) cycle progression, through PGE2 receptors, E-type prostanoid (EP)1 and EP3.

Conclusion
Sp1-regulated PGE2 production activates FAO and TCA cycle in mitochondria, through EP1 and EP3 receptors, resulting in TMZ resistance in GBM. These results will provide us a new strategy to attenuate drug resistance or to re-sensitize recurred GBM.
Background Glioblastoma (GBM), which is classi ed as a grade IV glioma, is the most aggressive and malignant brain tumor. [1] Its current standard treatment is tumour resection followed by chemotherapy and radiotherapy, and in this regard, temozolomide (TMZ), an oral alkylating agent, is the most widely used chemotherapy agent. However, with this current standard treatment, GBM still remains incurable and is characterized by a high prevalence of recurrence owing to its acquisition TMZ resistance. [1] Previously, we reported that speci city protein 1 (Sp1) promotes the development of malignant characteristics in GBM, including drug resistance and cancer stem cell enrichment. [2][3][4][5][6][7][8] Particularly, it enhances resistance to TMZ by upregulating the expression of cytochrome p450 (CYP) 17A1, which catalyses the metabolism of cholesterol to neurosteroids. [4,5,9,10] Therefore, these ndings suggest that Sp1 plays an important role in the acquisition of drug resistance in cancers by regulating the reprogramming of arachidonate metabolism. Therefore, in this study, our objective was to clarify the role of Sp1 in regulating arachidonate metabolism, leading to GBM drug resistance.
With respect to phospholipid-prostaglandin metabolism, prostaglandin E2 (PGE2) is upregulated and is frequently observed in various types of cancer. [11][12][13] Speci cally, in GBM, the enzyme that is primarily responsible for PGE2 synthesis, cyclooxygenase-2 (COX2)/prostaglandin-endoperoxide synthase 2 (PTGS2), reportedly promotes cancer cell proliferation and migration, [14,15] suggesting that COX2 enhances arachidonic acid (AA) metabolism, which is necessary for the survival of GBM cells, to synthesize PGE2. However, the correlation between AA-derived metabolites and drug resistance in GBM remains unclear.
To sustain rapid proliferation, cancer cells take advantage of the reprograming of metabolic networks to generate su cient bioenergy to support their cellular functions. Even though it is well known that the Warburg effect is the metabolic reprogramming that enhances aerobic glycolysis for cancer proliferation, [16] the importance of mitochondria-mediated metabolism has been raised recently to challenge the Warburg effect in GBM. Evidence suggests that GBM cells rely on oxidative phosphorylation in mitochondria to generate bioenergy. [17,18] Additionally, fatty acid β-oxidation (FAO), which is the main step in lipid metabolism for the generation of bioenergy inside the mitochondria, has also been considered to play an important role in tumour development in nutrient-deprived environments[18] and cancer metastasis. [19,20] However, its role in cancer drug resistance is still unknown.
In this study, we observed that Sp1 enhances the synthesis of PGE2 from AA in patients with recurrent GBM, and that PGE2 induces TMZ resistance by enhancing mitochondrial activity. Our ndings provide a novel mechanism for the reprograming of AA metabolism to the end of combating drug resistance. They also suggest the possibility of realizing GBM treatment based on the combination of the EP1 antagonist, ONO-8713, and TMZ to reverse drug resistance.

Materials And Methods
Human samples GBM samples were obtained from patients with GBM admitted to the Keelung Chang Gung Memorial Hospital, Linkou Chang Gung Memorial Hospital, and Taipei Medical University-Shuang-Ho Hospital. The characteristics of these patients are summarized in Supplementary Tables S1 and S2 and Supplementary  Fig. 1 The pathologies of the human brain tumour samples were determined according to WHO classi cation. [21,22] Isocitrate dehydrogenase 1 (IDH-1), glial brillary acidic protein (GFAP), and O 6 -Tissue samples (about 50 mg) and 10 µL of BHT (butylated hydroxytoluene)/MeOH solution (W/V, 4.8 g/100 mL) were subjected to protein precipitation by adding 100 µL of MeOH containing deuteriumlabeled IS, at a nal concentration of 50 ng/mL each of PGE2-d4, 6-keto PGF1α-d4, 5-HETE-d8 and 100 ng/mL of 9-HODE-d4. Samples were extracted with 500 µL of MeOH using a tissuelyzer at 50 Hz for 30 sec (3 times) followed with 5 ultrasonication cycles (1 min treatment and 1min break). The supernatant was transferred to new tubes after centrifuged at 12000 rpm for 10 min at 4°C, and followed by diluting with pure water to 15% MeOH concentration, followed by solid phase extraction (SPE) pretreated with MeOH and equilibrated with H2O. The extract was dried and then re-dissolved in 100 µL of MeOH, followed by ltering the solution with a 0.22 µm membrane lter before UPLC-MS/MS analysis.
Liquid chromatography and mass spectrometry UPLC-MS/MS analyses were conducted on an Agilent UPLC-MS/MS system consisting of 1290 UPLCsystem coupled with an Agilent 6470 triple-quadrupole mass spectrometer (Agilent Technologies). For analysis, 3 µL of the extract were injected. Chromatographic separation was achieved on an Agilent ZORBAX RRHD Eclipse XDB C18 column (2.1×100 mm, 1.8 µm particles) using a ow rate of 0.659 mL/min at 45°C during a 13 min gradient (0-12 min from 68 % A to 20 % A, 12-13 min 5 %A), while using the solvents A, water containing 0.005% formic acid, and B, acetonitrile containing 0.005% formic acid.
Electrospray ionization was performed in the negative ion mode using N2 at a pressure of 30 psi for the nebulizer with a ow of 10 L/min and a temperature of 300°C, respectively. The sheath gas temperature was 350°C with a ow rate of 11 L/min. The capillary was set at 3500 V and the nozzle voltage was 500 V. Multiple reaction monitoring (MRM) has been used for quanti cation of screening fragment ions.
Data preprocessing: Peak determination and peak area integration were performed with MassHunter Workstation software (Version B.08.00, Agilent Technologies) while auto-integration was manually inspected and corrected if necessary. The obtained peak areas of targets were corrected by appropriate IS and calculated response ratios were used throughout the analysis.

Untargeted metabolomics
The ultra-high performance liquid chromatography-quadrupole time-of-ight mass spectrometry (UHPLC-Q-TOF-MS) analysis was performed using an UHPLC system (1290, Agilent Technologies, Santa Clara, CA, USA) with a UPLC BEH Amide column (1.7 µm 2.1*100mm, Waters, Milford, MA, USA) coupled to TripleTOF 6600 (Q-TOF, AB Sciex, Framingham, MA, USA). The mobile phase consisted of 25 mM NH4OAc and 25 mM NH4OH in water(pH=9.75)(A) and acetonitrile (B) were carried with elution gradient as follows: 0 min, 95% B; 7 min, 65% B; 9 min, 40% B; 9.1 min, 95% B; 12 min, 95% B, which was delivered at 0.5 ml per min. The injection volume was 2 µL. The Triple TOF mass spectrometer was used for its ability to acquire MS/MS spectra on an information dependent basis (IDA) during an LC/MS experiment. In this mode, the acquisition software (Analyst TF 1.7, AB Sciex) continuously evaluates the full scan survey MS data as it collects and triggers the acquisition of MS/MS spectra depending on preselected criteria. In each cycle, 12 precursor ions whose intensity greater than 100 were chosen for fragmentation at collision energy (CE) of 30 V (15 MS/MS events with product ion accumulation time of 50 msec each).
ESI source conditions were set as following: Ion source gas 1 as 60 Psi, Ion source gas 2 as 60 Psi, Curtain gas as 35 Psi, source temperature 650℃, Ion Spray Voltage Floating (ISVF) 5000 V or -4000 V in positive or negative modes, respectively.
Enzyme-linked immunosorbent assay (ELISA) AA and PGE2 in culture media of wild type and TMZ-resistant U87MG cells were determined using Arachidonic Acid ELISA Kit (E4602, Biovision, Milpitas, CA, USA) and Prostaglandin E2 ELISA Kit (514010, Cayman, Ann Arbor, MI, USA), respectively, according to the manufacturers' instruction.
Promoter reporter assay Plasmids containing each promoter region were transfected into targeted cell lines. Cells were harvested with diluted Cell Culture lysis 5X Reagent (Promega, San Luis Obispo, CA, USA). 10 µL of sample were then mixed with 10 µL of luciferin (Promega). Luminometer (HIDEX, Tampa, FL, USA) was used to measure the promoter activity of the mixture.

Transmission electron microscopy (TEM)
To prepare the sample for TEM, cells were washed by PBS and xed in 2% paraformaldehyde, 2.5% glutaraldehyde, and 0.1 M cacodylate (MERCK Millipore, Billerica, MA, USA) at room temperature. The samples were then sent to Taipei Medical University Core Facility (Taipei, Taiwan) for further preparation and imaging.

RNA-Seq and bioinformatics
After total RNA extraction, samples were subjected to genomic sequencing. The gene expression in uenced by Sp1 for 1.5 folds was sorted, and the functional grouping was performed using the Ingenuity Pathway Analysis (IPA) system (https://www.qiagenbioinformatics.com/products/ingenuitypathway-analysis). Metabolism genes were sorted using Cancer Cell Metabolism Genes [25]. Heatmaps were prepared based on the level of expression using ToppCluster (https://toppcluster.cchmc.org/). Chromatin immunoprecipitation coupled with sequencing (ChIP-Seq) U87MG cells were xed with 1% formaldehyde for preserving the protein-DNA interactions, and DNAprotein complexes were harvested using the Simple ChIP enzymatic chromatin IP kit (#9003, Cell Signaling Technology, Danvers, MA, USA) followed by NextSeq 500 high throughput sequencing system (Illumina, San Diego, CA, USA) as described previously [7,26]. XFe24 Seahorse Mitochondrial Respiration Mito Stress Test Cells were treated with different conditions of TMZ, PGE2 or EP1-EP4 antagonists for 4 days. After the treatment, cells were trypsinized and 2x10 4 cells/well were seeded into the XFe24 Cell Culture Microplates (Agilent Technologies) and incubated for a day. Meanwhile, a sensor cartridge (detecting probes, Agilent Technologies) in Seahorse XF Calibrant at 37°C was hydrated in a non-CO2 incubator overnight for the following experiments. In the assay day, the cell-cultured medium was replaced with assay medium (DMEM without sodium bicarbonate, supplemented with 2% FBS and Penicillin/Streptomycin, pH: 7.4) and incubated for 1 h. Cells were then incubated at 37°C in a non-CO2 incubator that ready for experiments. Oligomycin (10 µM), FCCP (2 µM), and rotenone/antimycin A (5 µM) were prepared and placed into the sensor cartridge for the injection in the running procedure. The procedure of the assay was performed according to the guidelines from the XFe24 Seahorse Mitochondrial Respiration Mito Stress Test (Agilent technologies) [27]. For the evaluations of FAO percentage, etomoxir (40 µM) were added into the medium, 90 min before running the assay. FAO-dependent oxygen consumption was calculated as [OCR from groups without etomoxir -OCR from groups treated with etomoxir].
Plasmids and transfection GFP-Sp1 and PTGES2 plasmids (HG19428-ACG, Sino Biological, Wayne, PA, USA) were transfected into cells with Poly-Jet TM Reagent (SignaGen Laboratories, Rockvillie, MD, USA) for overexpression. PTGS2 siRNA (Dharmacon, Lafayette, CO, USA) was transfected into cells with Lipofectamine RNAiMAX Reagent (Thermo Fisher Scienti c) for knockdown. PLA2G5, ABHD8, and PTGS2 promoter were designed as a 1000 bp sequence before the coding region. The information of the constructs was listed in Supplementary Table 4.

Western blotting
Protein samples were separated on SDS-PAGE and transferred onto the PVDF membrane (Bio-Rad). The PVDF membrane was blocked in 5% nonfat milk in TBST buffer at room temperature for an hour, and then incubated with speci c primary antibodies (Supplementary Table 5) at 4°C overnight. After washing with TBST buffer, the membranes were incubated with the appropriate secondary antibodies for another one hour. Finally, the membranes were washed, and then developed by using T-Pro LumiLong Plus Chemiluminescent detection kit (T-Pro Biotechnology, New Taipei City, Taiwan).

Real-time PCR
The RNA sample was extracted by TRIzol (Thermo Fisher Scienti c), and 1 µg of total RNA was subjected to real-time PCR reagent using Prime Script TM RT Reagent kit (Takara Bio. Inc, Shiga, Japan). The expression of each mRNA was determined using 2 × SYBR real time master mix (AB Sciex) and the speci c primers (Supplementary Table 6). GAPDH was used as the internal control. SYBR green uorescence was then monitored using an ABI 7000 Sequence Detection System (AB Sciex).
MitoPlates analysis for estimating the consumption of NADH/FADH-producing substrate

Statistical analysis
The data obtained were represented as means ± S.E.M. Two-tailed unpaired Student's t-test or two-way ANOVA (animal experiments) were used to analyse the differences between the control and experimental groups. *P < 0.05, **P < 0.01, and ***P < 0.001 were considered signi cant in all comparisons.

Results
Importance of Sp1 in phospholipid metabolism to generate AA in TMZ-resistant GBM Metabolic reprogramming is critical for cancer progression.
Importance of Sp1-regulated AA metabolism in TMZresistance acquisition in GBM To elucidate the role of AA metabolism in TMZ-resistance acquisition, the effects of inhibitors that target PLA2, PTGS2/COX2, ALOX5, and ALOX12 on TMZ-resistant U87MG-R cells were evaluated. As shown in Fig. 2A, pyrrophenone (PLA2 inhibitor) and celecoxib (PTGS2 inhibitor) synergistically enhanced TMZinduced cell death, whereas neither zileuton (ALOX5 inhibitor) nor ML-355 (ALOX12 inhibitor) exhibited any therapeutic effects (Fig. 2B). These results support the idea that the regulation of AA metabolism by Sp1, leading to the synthesis of prostaglandin and not lipoxygenase, is critical for the acquisition of TMZ resistance in GBM.
Moreover, to investigate the clinical relevance of Sp1-regulated prostaglandin synthesis, we collected 14 pairs of GBM specimens (primary vs. recurrent) for analysis (Supplementary Table 1). RNA-Seq revealed that the enzymes related to AA generation and prostaglandin metabolism, including PLA2G5, ABHD8, PTGS2, PTGES2, PTGDS, and AKR1C3, were obviously upregulated in the recurrent GBM specimens (Fig. 2C). Additionally, chromatin immunoprecipitation (ChIP)-Seq and reporter assay analyses revealed that Sp1 signi cantly enhanced the transcriptional activities of PLA2G5, ABHD8, PTGS2, PTGES2, PTGDS, and AKR1C3 with respect to prostaglandin synthesis in TMZ-resistant GBM cells by binding to their promoter regions (Fig. 2D-E). Overall, our results indicated the Sp1-regulated metabolic pathway from AA to prostaglandins is required for TMZ resistance acquisition in GBM.
Enhancement of PGE2 synthesis in recurrent/resistant GBM Further, targeted metabolome analysis focusing on ω-3/6 fatty acid metabolism was performed to determine the differences in the amounts of AA-related metabolites between paired (primary vs. recurrent) tissues from patients with GBM (Supplementary Table 2). As shown in Fig. 3A and Supplementary  Fig. 1A, four out of a total ve recurrent specimens exhibited signi cantly higher PGE2 and PGD2 levels. Additionally, we also estimated the amounts of AA-related metabolites via targeted arachidonate metabolome analysis. Thus, it was observed that the levels of PGD2 and PGE2 were also obviously higher in the recurrent tissue specimens than in normal brain tissue and primary specimens, suggesting that these two prostanoids are involved in TMZ resistance and tumour recurrence ( Supplementary  Fig. 1B). Further, as shown in Fig. 3B and Supplementary Fig. 1C, PGE2 signi cantly attenuated TMZinduced cytotoxicity in patient-derived GBM cells, Pt#3, and U87MG. However, PGD2 did not affect cellular response to TMZ.
Besides, compared with the wild type glioblastoma cell line, an increase in PGE2 secretion was observed in TMZ-resistant glioblastoma cell lines (Pt#3-R, P1S, and U87MG-R cells), while a decrease was observed in U87MG-Sp1 cKO cells (Fig. 3C-D). Speci cally, P1S cells, which exhibit TMZ resistance, secreted more PGE2 than patient-derived glioblastoma cells Pt#3 (Fig. 3C, right). Further, long term TMZ treatment resulted in the enhancement of PTGS2 expression and PGE2 secretion (Fig. 3E and F and Supplementary Fig. 2A and 2B). It was also observed that GBM cells, Pt#3 and U87MG, gradually showed increased tolerance to TMZ in the long-term treatment, and this increased tolerance was compromised by the PTGS2 inhibitor, celecoxib, resulting in a decrease in the number of adhered cells, an increase in the number of round-up cells, and the destruction of cell morphology ( Supplementary Fig. 2C and 2D). These results supported the idea that the synthesis and secretion of PGE2 are signi cantly increased and required for the establishment of the recurrent and TMZ resistant GBM.
Induction of mitochondrial fusion to rescue of TMZimpaired respiration by PGE2 via EP1/EP3 cascades Given that mitochondrial dynamics act as regulators in cancer processes, [29][30][31] and that mitochondria play a critical role in the generation of bioenergy for GBM, [17,18] we evaluated whether mitochondrial ssion and fusion are involved in drug resistance in GBM. Thus, RNA-Seq showed that mitochondrial fusion-related proteins, such as MFN1, MFN2, and OPA1, were signi cantly upregulated in resistant Pt#3-R and recurrent GBM specimens (Fig. 4A), while mitochondrial ssion-related proteins showed a different pattern ( Supplementary Fig. 3A). Further, TEM also showed condensed mitochondria, which favours mitochondrial fusion, in TMZ-resistant GBM cell lines (Supplementary Fig. 3B). Moreover, consistent with PGE2 upregulation in long term TMZ-treated GBM cells, MFN1 and OPA1 levels showed a gradual increase (Fig. 4B). These results suggested that mitochondrial fusion, which enhances ATP production and respiration, [29][30][31]
PGE2-based promotion of FAO and TCA cycle progression contribute to increased ATP production In this study, we also investigated the contribution of the key proteins and metabolites that are regulated by Sp1 and PGE2 to the upregulation of ATP production after mitochondrial fusion. [29][30][31] In this regard, RNA-Seq revealed that Sp1 knockdown downregulated the expression of several mitochondrial ATP production-related genes, which were obviously upregulated in recurrent specimens (Table 1, Fig. 5A, and Supplementary Fig. 4A). Among these genes, CPT1A and ACAA2, which play predominant roles in the regulation of FAO pathways, were consistently and signi cantly upregulated in the recurrent GBM and TMZ-resistant GBM specimens ( Fig. 5A and B, and Supplementary Fig. 4B). These results are consistent with our hypothesis that Sp1 enhances lipid metabolism in TMZ-resistant GBM, with CPT1A as the ratelimiting step of FAO, [32] facilitating the transfer of fatty acids into the mitochondria for bioenergy generation. Therefore, we con rmed that PGE2 signi cantly enhances the protein expression of CPT1A without affecting the levels of glycolysis-related proteins, LDHA, and NQO1 (Fig. 5C, and Supplementary   Fig. 4C). Moreover, PGE2 signi cantly increased the percentage contribution of FAO to mitochondrial respiration. This is an important metabolic characteristic of TMZ-resistant GBM (Fig. 5D and Supplementary Fig. 4D). These results indicated that Sp1-regulated PGE2 enhances FAO by inducing CPT1A expression to acquire TMZ resistance.
Supporting the importance of FAO in drug resistance, untargeted metabolome analysis showed a signi cant increase in the levels of FAO-related metabolites, including acyl-carnitine, trans-2-Enoyl-CoA, 3hydroxyacyl-CoA, and acetyl-CoA, in TMZ-resistant GBM cells (Fig. 5E). Consistent with this observation, for TCA cycle progression, PGE2 signi cantly increased the levels of citrate, cis-aconitate, α-ketoglutarate, succinate, and malate (Fig. 5F). Furthermore, to generate bioenergy, PGE2 enhanced the consumption of fatty acid-and TCA cycle-related substrates, such as short chain (Acetyl-L-carnitine) and long chain (Pamitoyl-L-carnitine) fatty acids, citric acid, cis-aconitic acid, fumaric acid, and L-malic acid, (Fig. 5G), implying that in the presence of PGE2, GBM cells show preference for the consumption of fatty acids as a major source of bioenergy. Therefore, Sp1-regulated PGE2 increases the number of metabolites in FAO and drives TCA cycle progression after mitochondrial fusion to rescue TMZ-induced nutrient deprivation and mitochondria damage.

Discussion
Given that GBM utilizes mitochondrial oxidation to generate bioenergy, [17,18] we investigated which part of the ATP-generating system is upregulated in TMZ-resistant GBM. Our study indicated that FAO and TCA cycles are upregulated in TMZ-resistant GBM owing to the action of PGE2. It has also been reported that FAO plays an important role in tumour development in nutrient-deprived environment[18] and cancer metastasis. [19,20] Particularly, increases fatty acid uptake and the alteration of lipid metabolism are associated with drug resistance in cancers and provide the essential bioenergy for cancer survival based on the enhancement of FAO e cacy. [33] For example, the resistance of cancers to antiangiogenic drugs is due to the alteration of FAO and lipid metabolism in the tumour environment. [34] Therefore, therapeutic strategies that target the alteration of lipid metabolism, such as blocking cancer-associated adipocytemediated fatty acid production followed by the inhibition of free fatty acid uptake by cancer cells, have been designed to combat drug resistance. [34] However, strategies that target FAO for the treatment of drug resistant cancer remain limited. Even though etomoxir reportedly blocks FAO by inhibiting CPT1 and suppressing GBM cell growth,[18] our experimental results revealed that etomoxir is not effective in inhibiting drug resistance in GBM (data not shown). Hence, the PGE2/EP1 cascade, which has a signi cant regulatory effect on CPT1-dependent FAO, is expected to be a potential target for overcoming drug resistance.
Celecoxib is a well-known COX2/PTGS2 inhibitor that attenuates the synthesis of PGE2, and reportedly, kills GBM cells. [35] However, it failed a phase II clinical trial owing to its poor permeability with respect to crossing the blood-brain barrier.[36, 37] Therefore, a treatment strategy that targets the PGE2-induced functional pathway, such as PGE2 receptor (EP1-EP4) blocking, instead of COX2/PTGS2-mediated PGE2 synthesis suppression, is an alternative medical option. Reportedly, PGE2 receptors are associated with cancer malignancy, while their antagonists have also shown tumorigenesis inhibition effects.
[38] Among EP1-EP4 receptors, EP1 is involved in the activation of cancer cell migration and invasion, [39][40][41] supports tumour adaptation to hypoxia, [42] and enhances cancer initiation. [43,44] EP2 is known to be involved in the induction of angiogenesis [38] and in the suppression of antitumor immune response. [45] Further, its inhibition in glioma results in tumour growth suppression.
[46] Conversely, the role of EP3 in tumorigenesis is still controversial.
[38] The results of our previous study showed that the EP3 antagonist, ONO-AE3-240, also exerts cell cytotoxicity on TMZ-resistant GBM cells. Additionally, it has been demonstrated that EP4 is involved in cancer migration, metastasis, and aberrant DNA methylation. [47][48][49] The results of this study indicated that the EP1 antagonist, ONO-8713, is capable of blocking PGE2induced mitochondrial activation and suppressing the survival of TMZ-resistant GBM cells. Therefore, EP antagonists, instead of celecoxib, have potential to attenuate drug resistance by blocking PGE2-mediated signalling in GBM. Overall, PGE2 induces TMZ resistance in GBM via mitochondria-mediated FAO activation under the control of EP1 and EP3. Further combining TMZ with an EP1 antagonist present as a potential combination therapeutic strategy for TMZ-resistant GBM. Therefore, the development of EP1 antagonists that can cross the blood-brain barrier during the treatment of TMZ-resistant GBM have great application prospects.

Conclusion
Sp1 increases PGE2 synthesis through enhancing gene expression involved in AA metabolism to PGE2 in recurrent GBM, leading to TMZ resistance. Further, PGE2 increases mitochondrial fusion, resulted in the enhancement of FAO and TCA cycle to increase the ATP production, through the EP1 and EP3 receptors.
Moreover, EP1 antagonist ONO-8713 exhibits a potentially therapeutic effect on TMZ-resistant GBM in vivo and in vitro (Figure 6f). Availability of data and materials

Abbreviations
The author declared that all and the other data supporting the ndings of this study are available within the paper. The raw data that support the ndings of this study are available from the corresponding author upon reasonable request.

Competing interests
The authors declare no competing interests.   Table   S5.

Figure 2
The Sp1-regulated COX2/PTGS pathway in recurrent glioblastoma. Effects of multiple inhibitors, including (A) Pyrrophenone (left) and celecoxib (right), (B) Zileuton (left) and ML-355 (right) on the viability of U87MG-R cells. The cells were treated with TMZ in the presence of the indicated inhibitors for specimens. Samples were collected from 14 patients and subjected to RNA-Seq. The red arrow-marked genes play important roles in the metabolic pathway involving the synthesis of prostaglandins from AA (the COX pathway) as shown on the right panel. (D) De ned promoter regions of PLA2G5, ABHD8, and PTGS2. The binding regions of Sp1 were determined via ChIP-Seq. (E) Sp1-induced activities of pGL2conjugated promoter constructs, including PLA2G5, ABHD8, and PTGS2. Luciferase reporter assay was employed to analyse the promoter activities in wild type and TMZ-resistant glioblastoma cells. Data were analysed by performing two-tailed unpaired Student's t test.   after treatment with PGE2 for four days based on targeted metabolomics analysis. The alteration of the levels of the different metabolites is illustrated using bar plots. Data were analysed by performing twotailed unpaired Student's t test. (G) A172 and A172R cells after treatment with PGE2 for four days based on MitoPlate assay for 2 h. Data were analysed by performing two-tailed unpaired Student's t test. Figure 6