Aurora kinase A inhibition reverses the Warburg effect and elicits unique metabolic vulnerabilities in glioblastoma

Abstract

Aurora kinase A (AURKA) has emerged as a drug target for glioblastoma (GBM). However, resistance to therapy remains a critical issue. By integration of transcriptome, chromatin immunoprecipitation sequencing (CHIP-seq), Assay for Transposase-Accessible Chromatin sequencing (ATAC-seq), proteomic and metabolite screening followed by carbon tracing and extracellular flux analyses we show that genetic and pharmacological AURKA inhibition elicits metabolic reprogramming mediated by inhibition of MYC targets and concomitant activation of Peroxisome Proliferator Activated Receptor Alpha (PPARA) signaling. While glycolysis is suppressed by AURKA inhibition, we note an increase in the oxygen consumption rate fueled by enhanced fatty acid oxidation (FAO), which was accompanied by an increase of Peroxisome proliferator-activated receptor gamma coactivator 1-alpha (PGC1α). Combining AURKA inhibitors with inhibitors of FAO extends overall survival in orthotopic GBM PDX models. Taken together, these data suggest that simultaneous targeting of oxidative metabolism and AURKAi might be a potential novel therapy against recalcitrant malignancies.

Fig. 1. Blocking AURKA stalls the proliferation of GBM by reducing glycolysis in a c-Myc dependent manner.

a Volcano plot of reverse-phase-protein-array (RPPA) data shows a reduced expression of c-Myc (red dot) in SF188 GBM cells treated with 100 nM alisertib for 24 h (n = 3 independent samples). FC: fold change. b Protein capillary electrophoresis of SF188 and GBM22 cells treated with the indicated concentrations of alisertib for 24 h. Vinculin is used as a loading control. c SF188 was transfected with non-targeting (siNT) or AURKA specific siRNA (siAURKA) or transduced with shARKA (shRNA) or sgAURKA (sgRNA) and the whole-cell protein lysates were subjected to protein capillary electrophoresis. d SF188 and GBM22 cells were transfected with non-targeting (siNT) or two specific siRNAs targeting Myc, treated with increasing concentrations of alisertib for 72 h, and cellular viability was analyzed (n = 4 independent samples). IC₅₀ in µM range. e SF188 and GBM22 cells were infected with empty vector or c-Myc adenovirus and treated with increasing concentration of alisertib for 72 h, and cellular viability was analyzed (n = 4 independent samples). IC₅₀ in µM range. f SF188 and GBM22 cell lysates were immunoprecipitated with IgG or c-Myc antibody and analyzed by standard western blot for the indicated antibodies. Input: cell lysate loading control. IgG: negative control. g SF188 cells were treated with DMSO or alisertib in the presence or absence of 5 µM MG132 and the whole-cell lysates were subjected to protein capillary electrophoresis for the indicated proteins. For h, i SF188 cells were treated with DMSO or alisertib in the presence or absence of 10 µg/mL cycloheximide (CHX) and the whole-cell lysates were subjected to protein capillary electrophoresis. Quantification of c-Myc protein level is shown in (i) (n = 3 independent samples). For j, k protein capillary electrophoresis for the indicated proteins of SF188 and GBM22 cells treated with DMSO or alisertib for different time points. Quantification of protein level is shown in (k) (n = 2 independent samples). l SF188 cells were transfected with c-Myc-WT or c-Myc mutant (T58A), treated with the indicated concentrations of alisertib for 24 h, and analyzed by protein capillary electrophoresis for the indicated proteins. m GBM22-Myc-WT or GBM22-T58A-Myc cells were implanted in the right striatum of nude mice. Two groups were randomly assigned: vehicle and alisertib after seven days of the implantation. Mice were treated three times per week and animal survival is provided (Kaplan−Meier-curve): GBM22-T58A-Myc-vehicle: 34d, GBM22-T58A-Myc-alisertib: 27d; GBM22-Myc-OE-vehicle: 38d, GBM22-Myc-OE-alisertib: 50d. The log-rank test was used to assess statistical significance (n = 5 independent samples) (*p = 0.0127, n.s not significant). n SF188 and GBM22 cells were transfected with non-targeting siNT or siAURKA in the presence or absence of 2 µM CHIR-908014 (CHIR) for 24 h and were subjected to protein capillary electrophoresis for the indicated proteins. o SF188 and GBM22 cells were transfected with HA-EV, HA-Aurora A-WT, HA-Aurora A-D274N and were subjected to protein capillary electrophoresis for the indicated proteins. Statistical significance was assessed by two-tailed student’s t-test in (m). Data are shown as mean ± SD in (d, e, i, k). Source data are provided as a Source Data file.

Fig. 2. Blocking AURKA stalls GBM growth by reducing glycolysis.

a SF188 and GBM22 cells were cultured in galactose or glucose media for a week, treated with increasing concentrations of alisertib for 72 h, and cellular viability was analyzed (n = 4 independent samples). b SF188 GBM cells either transduced with a non-targeting (sgNT) or two different AURKA (sgAURKA) sgRNAs were cultured in galactose or glucose media for a week and treated with increasing concentration of alisertib. Cellular growth over time was determined (n = 4 independent samples). For c, d SF188 and GBM22 cells were treated with DMSO or alisertib and analyzed in the context of a glycolysis stress assay on a Seahorse XFe24 extracellular flux analyzer. Extracellular acidification rate (ECAR) is recorded at baseline, after injection of Glucose (G), Oligomycin (OM), and 2-DG. Quantification of glycolysis is shown in (d) (n = 5 independent samples). e Real-time PCR analysis of the mRNA level of genes related to glycolysis in SF188 and GBM22 cells treated with DMSO or alisertib for 24 h (n = 8 in SF188 + DMSO and GBM22 + DMSO, n = 4 in SF188 + Ali 0.1 μM and GBM22 + Ali 1 μM, independent samples). (HK2: SF188 ***p = 0.0002, GBM22: ***p = 0.0003; **p = 0.0082, ****p < 0.0001). 18S: internal control. FC: fold change. f SF188 and GBM22 cells were treated with DMSO or alisertib for 24 h and the whole-cell protein lysates were subjected to protein capillary electrophoresis. Vinculin is used as a loading control. g SF188 cells were transfected with non-targeting siRNA or specific AURKA siRNAs (single or pool) and the whole-cell lysates were analyzed by protein capillary electrophoresis for the indicated proteins. h Real-time PCR analysis of the mRNA level of genes related to glycolysis in SF188 transfected with non-targeting siRNA or specific siRNA targeting AURKA (n = 4 independent samples) (*p = 0.0179, ***p = 0.0001). i Shown are the ATP levels measured by polar LC/MS of SF188 cells treated with DMSO or 100 nM alisertib (n = 5 independent samples) (*p = 0.0221). For j, k SF188 and GBM22 cells were transfected with non-targeting or specific siRNA targeting AURKA and were analyzed in the context of a glycolysis stress assay on a Seahorse XFe24 extracellular flux analyzer. The graphs show glycolysis level in (j) (SF188: n = 2 in siNT, n = 3 in siAURKA; GBM22: n = 3 independent samples) (**p = 0.0032) or OCR/ECAR levels in (k) (SF188: n = 3; GBM22: n = 3 in siNT, n = 2 in siAURKA independent samples) (***p = 0.0001). l SF188 cells were transfected with c-Myc-WT and c-Myc mutant (T58A), treated with increasing concentrations of alisertib for 24 h, and the whole-cell lysates were analyzed by protein capillary electrophoresis. m SF188 cells were transfected with c-Myc-WT and c-Myc mutant (T58A), treated with 50 nM alisertib for 24 h, and analyzed on a Seahorse XFe24 extracellular flux analyzer. Shown is the quantification of ECAR level (n = 5 independent samples) (**p = 0.0073, n.s not significant). n SF188 and GBM22 cells were treated with alisertib for 24 h and were subjected to CHIP with an IgG as a negative control or a c-Myc specific antibody. The HK2 region was amplified by PCR (n = 3 independent samples). Statistical significance was assessed by a two-tailed student’s t-test. Data are shown as mean ± SD in (a−e, h−k, m, n). Source data are provided as a Source Data file.

Fig. 3. AURKA inhibition facilitates oxidative energy metabolism.

a GBM22 cells were treated with DMSO or alisertib and analyzed for oxygen consumption rate (OCR) on a Seahorse XFe24 device. OM: oligomycin, R/A rotenone/antimycin. The graph (right panel) shows the OCR levels (n = 5 in DMSO, Ali 1 μM, Ali 5 μM; n = 4 in Ali 10 μM independent samples) (Ali 1 μM: *p = 0.0303, Ali 5 μM: **p = 0.0074, Ali 10 μM: *p = 0.0254). For b, c Seahorse mitochondrial stress assay of parental or chronically alisertib treated GBM22 and SF188 cells. The graph (right panel) shows the mitochondrial OCR levels (n = 3 in SF188 vs SF188AR, n = 5 in GBM22 vs GBM22AR independent samples) (**p = 0.0014, ***p = 0.0001). d SF188 cells were transduced with sgRNA against AURKA and were analyzed for the oxygen consumption rate (OCR) on a Seahorse XFe24 device. The graph (right panel) shows the mitochondrial OCR level (n = 5 independent samples) (**p = 0.0016). e Electron microscopy of parental or chronically alisertib treated U87 cells. Scale bar: 2 µm. Higher magnification images are shown in the lower panel. Scale bar: 500 nm. The white dotted square shows the enlarged region of mitochondria in the lower panel. The red arrow highlights the mitochondrial morphology. f GBM22 cells were treated with DMSO or alisertib for 24 h, labeled with either CellRox or Mitotracker dye, and analyzed by flow cytometry (n = 2 independent samples). g GBM22 cells were treated with 1 µM alisertib, 2 µM GTPP, or the combination of both for 24 h and analyzed for oxygen consumption rate (OCR) on a Seahorse XFe24 device. The graph (below panel) shows the mitochondrial OCR level (n = 3 in DMSO and Combination, n = 4 in Ali 1 μM and GTPP 2 μM independent samples) (***p = 0.0001, ****p < 0.0001). For h, i SF188 and GBM22 cells were treated with alisertib, GTPP, or the combination of both for 72 h. Thereafter, cellular viability was measured, and statistical analysis was performed. Isobolograms are shown. The graphs in (i) show the quantification (n = 4 independent samples) (****p < 0.0001). For j, k SF188 and GBM22 cells were treated with alisertib, oligomycin, or the combination of both for 72 h. Thereafter, cellular viability was determined and statistical analysis was performed. Isobolograms are shown. The quantification is shown in (k) (n = 4 independent samples) (**p = 0.0015, ****p < 0.0001). Data are shown as mean ± SD in (a−d, g, i, k). Source data are provided as a Source Data file.

Fig. 4. AURKA inhibition drives PPARA signaling leading to an increase of PGC1α in a partial c-Myc dependent manner.

a Parental or chronically alisertib treated GBM22 cells were subjected to transcriptomic analysis and followed by GSEA. Shown are the enrichment plots of Hallmark_Myc_Target and Biocarta_PPARA_Pathway. NES: normalized enrichment score. FDR: false discovery rate. b Real-time PCR analysis of PGC1α mRNA levels in SF188, GBM22, and U87 treated with alisertib or DMSO (n = 8 in SF188 and U87, n = 4 in GBM22 independent samples) (*p = 0.0116, ****p < 0.0001). c Real-time PCR analysis of PGC1α mRNA levels in parental or alisertib chronically exposed SF188, GBM22, and U87 cells (n = 8 independent samples) (****p < 0.0001). d Real-time PCR analysis of PDK4 mRNA levels of parental or alisertib chronically exposed SF188, GBM22, and U87 cells (n = 3 in SF188 and SF188AR, n = 4 in GBM22 and GBM22AR, n = 3 in U87 and n = 4 in U87AR independent samples) (****p < 0.0001). FC: fold change. e Protein capillary electrophoresis for the indicated proteins of GBM22 and SF188 cells treated with alisertib for 24 h. Vinculin is used as a loading control. f Protein capillary electrophoresis for the indicated proteins of parental or chronically alisertib treated GBM22 or SF188 cells. g GBM22, SF188, and U87 cells were transduced with scrambled or shARKA and the whole-cell lysates were subjected to protein capillary electrophoresis with the indicated antibodies. h SF188 cells were transduced with scramble or sgAURKA and the whole-cell lysates were subjected for protein capillary electrophoresis with the indicated antibodies. i ChIP-sequencing (H3K27ac) and ATAC-sequencing were performed in parental or chronically alisertib treated GBM22 cells. Shown are the respective tracks around the PPARGC1A locus. j Parental or chronically alisertib treated GBM22 and SF188 cells were subjected to CHIP with an IgG as a negative control or a c-Myc specific antibody. The PGC1α region was amplified by PCR (n = 3 independent samples) (****p < 0.0001). k Parental or chronically alisertib treated GBM22 and SF188 cells were subjected to CHIP with an IgG as a negative control or a H3K27ac specific antibody. The PGC1α region was amplified (n = 3 independent samples) (****p < 0.0001). l Real-time PCR analysis of PGC1α mRNA levels of GBM22 cells transfected with c-Myc-WT and c-Myc mutant (T58A) followed by treatment with increasing concentrations of alisertib (n = 4 independent samples). m GBM22 and SF188 cells were transfected with c-Myc-WT and c-Myc mutant (T58A), treated with alisertib for 24 h, and the whole-cell lysates were subjected to protein capillary electrophoresis with the indicated antibodies. n GBM22 cells were transfected with non-targeting or two specific siRNAs targeting Myc, treated with 10 µM alisertib for 24 h, and the whole-cell lysates were subjected to protein capillary electrophoresis. o GBM22 and SF188 cells were transfected with non-targeting or specific PGC1α siRNA, treated with increasing concentration of alisertib for 72 h, and cellular viability was analyzed (n = 4 independent samples). IC₅₀ in µM range in GBM22 and in nM range in SF188. p Parental or chronically alisertib treated SF188 cells were transfected with non-targeting siRNA or PGC1α specific siRNA and analyzed for oxygen consumption rate (OCR) on a Seahorse XFe24 device. The graph (right panel) shows the OCR level (n = 3 independent samples) (***p = 0.0007, ****p < 0.0001). Statistical significance was assessed by a two-tailed student’s t-test in (b−d, j, k) or ANOVA with Dunnett’s multiple comparison test in p. Data are shown as mean ± SD in b−d, j−l, o, p). Source data are provided as a Source Data file.

Fig. 5. AURKA inhibition impacts central carbon metabolism resulting in enhanced labeling of TCA cycle metabolites by long-chain fatty acids.

a Summary of key reactions related to the tracer experiment. Blue circles indicate ¹³C carbons from glutamine, brown circles indicate ¹³C carbons from palmitic acid, whereas red circles highlight ¹³C carbons from glucose. Black circles are used to display ¹²C carbons. b, c Parental or chronically alisertib treated GBM22 cells were cultured in DMEM media containing 25 mM U-¹³C glucose, 4 mM glutamine and 10% dialyzed FBS and were subjected to LC/MS analysis (n = 3 independent samples) (pyruvic acid: *p = 0.0492, lactic acid: *p = 0.019). d Parental or chronically alisertib treated GBM22 cells were cultured in DMEM media containing 25 mM glucose, 4 mM U-¹³C glutamine and 10% dialyzed FBS and were subjected to LC/MS analysis (n = 3 independent samples). e Parental or chronically alisertib treated GBM22 cells were cultured in DMEM media containing 5 mM glucose, 1 mM glutamine, 100 µM U-¹³C palmitic acid, and 10% dialyzed FBS and were subjected to LC/MS analysis. Shown are relative percentages of the isotopologues for each metabolite (n = 3 independent samples). Data are shown as mean ± SD in (b−e). Source data are provided as a Source Data file.

Fig. 6. Dual inhibition of FAO and AURKA elicits a synergistic reduction in cellular viability of GBM cells.

For a, b parental or chronically alisertib treated GBM22 cells were subjected to transcriptomic analysis and followed by GSEA. Shown are a graphical plot of NES vs FDR-q value of microarray data in (a) and an enrichment plot of Go_Regulation_of_Fatty_Acid_Oxidation in (b). NES: normalized enrichment score. FDR: false discovery rate. c Real-time PCR analysis of genes related to fatty acid oxidation (mRNA levels) of parental or chronically alisertib treated GBM22 cells (n = 8 in GBM22 and n = 4 in GBM22AR independent samples). FC: fold change. d Non-polar metabolite analysis of parental or chronically alisertib treated GBM22 cells. A heatmap of parental or chronically alisertib treated GBM22 cells is shown. AcCa: acyl carnitine; Cer: ceramides; CerG1-3: neutral glycosphingolipids. CerP: phosphosphingolipids; ChE: cholesterol Ester; DG: diglyceride; LPC: lysophosphatidylcholine; LPE lysophosphatidylethanolamine; LPI: lysophosphatidylinositol; LPS: lysophosphatidylserine; PC: phosphatidylcholine; PE: phosphatidylethanolamine; PG: phosphatidylglycerol; PI: phosphatidylinositol; PS: phosphatidylserine; SM: sphingomyelin; So: sphingosine; TG: triglyceride. For e, f acyl-carnitine and lysophosphatidyl levels in the non-polar metabolite analysis of parental or chronically alisertib treated GBM22 cells (n = 5 independent samples). For g, h Parental or chronically alisertib treated GBM22 and SF188 cells were treated with alisertib, etomoxir, or the combination of both for 72 h, and cellular viability was analyzed. Isobolograms are shown in (g) and quantification of cell viability is shown in (h) (n = 4 independent samples) (**p = 0.0047, ***p = 0.0002, ****p < 0.0001). i Parental or chronically alisertib treated GBM22 cells were treated with alisertib, etomoxir, or the combination of both for 24 h and analyzed for oxygen consumption rate (OCR) on a Seahorse XFe24 device. The graph (right panel) shows the OCR level (n = 5 independent samples) (*p = 0.0195, ****p < 0.0001). j SF188 cells were transfected with non-targeting siNT or siAURKA (single or pool), treated with etomoxir for 72 h, and cellular viability was analyzed (n = 4 independent samples) (siAURKA-2: *p = 0.0181, siAURKA-4: *p = 0.1015). For k, l SF188 cells were transfected with non-targeting siNT or CPT1A specific siRNA (siCPT1A), treated with alisertib for 72 h, and cellular viability was analyzed (n = 7 in siNT + DMSO, n = 8 in siCPT1A + DMSO, n = 4 in siNT + Ali 20 nM and siCPT1A + Ali 20 nM independent samples) (*p = 0.0495, ***p = 0.0009). Protein capillary electrophoresis confirms the silencing of CPT1A is shown in l. Vinculin is used as a loading control. Statistical significance was assessed by a two-tailed student’s t-test in (j, k) or ANOVA with Dunnett’s multiple comparison test in (h, i). Data are shown as mean ± SD in (c, e, h−k). Source data are provided as a Source Data file.

Fig. 7. Dual inhibition of FAO and AURKA extends animal survival in orthotopic patient-derived xenograft models of human GBM.

For a, b GBM12 cells were implanted into the subcutis of immunocompromised Nu/Nu mice. After the tumors were established, the mice were randomized to four treatment groups: vehicle, alisertib (30 mg/kg), etomoxir (20 mg/kg), and the combination treatment. The tumor volume over time is shown in (a) and the tumor volume on the last day of the experiment is shown in (b) (n = 8 in alisertib and etomoxir, n = 9 in vehicle and combination independent tumors) (*p = 0.0438, **p = 0.0085, ***p = 0.0003). For c, d GBM43 cells were implanted into the subcutis of immunocompromised Nu/Nu mice. After the tumors were established, the mice were randomized to four treatment groups: vehicle, alisertib (30 mg/kg), etomoxir (20 mg/kg), and combination treatment of both. The tumor volume over time is shown in (c) and the tumor volume on the last day of the experiment is shown in (d) (n = 9 in vehicle and etomoxir, n = 12 in alisertib and combination independent tumors) (Ali vs Combination: *p = 0.0462, etomoxir vs combination: *p = 0.0162). e GBM12 cells were implanted and treated with alisertib, etomoxir, or the combination of both as described in (a). On the last day of the experiment, mice were injected with 100 µM U-¹³C-palmitic acid for 4 h and tumors were subjected to LC/MS. The graph shows the U-¹³C-palmitic acid labeling of the TCA cycle. FC: fold change. f Tumors from the experiment in (a) were fixed and stained with PGC1α. Scale bar: 50 µM. For g, h GBM12 and GBM22 cells were implanted in the right striatum of nude mice. Four groups were randomly assigned: vehicle, alisertib, etomoxir, and combination of both, seven days after the implantation. Mice were treated three times per week and animal survival is provided (Kaplan−Meier-curve): vehicle: GBM12: 20d, GBM22: 20d; alisertib: GBM12: 28d, GBM22: 22d; etomoxir: GBM12: 25.5d, GBM22: 19.5d; combination: GBM12: 37.5d, GBM22: 26d. The log-rank test was used to assess statistical significance (n = 5 in vehicle and alisertib, n = 4 in etomoxir, n = 6 in combination). For (i−k) The brain tumors from the experiment in (h) were fixed and stained with H&E, TUNEL, or Ki67. Scale bar: 50 µm. l Quantification of TUNEL and Ki67 positive cells in (j, k), respectively (n = 5 in vehicle, etomoxir, combination, n = 6 in alisertib independent high-power field microscopy) (***p = 0.0001, ****p < 0.0001). Statistical significance was assessed by a two-tailed student’s t-test in e or ANOVA with Dunnett’s multiple comparison test in (b, d, l). Data are shown as mean ± SEM in (a, c) and as mean ± SD in (b, d, e, l). Source data are provided as a Source Data file.