GLUT3 upregulation promotes metabolic reprogramming associated with antiangiogenic therapy resistance

Abstract

Clinical trials revealed limited response duration of glioblastomas to VEGF-neutralizing antibody bevacizumab. Thriving in the devascularized microenvironment occurring after antiangiogenic therapy requires tumor cell adaptation to decreased glucose, with 50% less glucose identified in bevacizumab-treated xenografts. Compared with bevacizumab-responsive xenograft cells, resistant cells exhibited increased glucose uptake, glycolysis, ¹³C NMR pyruvate to lactate conversion, and survival in low glucose. Glucose transporter 3 (GLUT3) was upregulated in bevacizumab-resistant versus sensitive xenografts and patient specimens in a HIF-1α-dependent manner. Resistant versus sensitive cell mitochondria in oxidative phosphorylation-selective conditions produced less ATP. Despite unchanged mitochondrial numbers, normoxic resistant cells had lower mitochondrial membrane potential than sensitive cells, confirming poorer mitochondrial health, but avoided the mitochondrial dysfunction of hypoxic sensitive cells. Thin-layer chromatography revealed increased triglycerides in bevacizumab-resistant versus sensitive xenografts, a change driven by mitochondrial stress. A glycogen synthase kinase-3β inhibitor suppressing GLUT3 transcription caused greater cell death in bevacizumab-resistant than -responsive cells. Overexpressing GLUT3 in tumor cells recapitulated bevacizumab-resistant cell features: survival and proliferation in low glucose, increased glycolysis, impaired oxidative phosphorylation, and rapid in vivo proliferation only slowed by bevacizumab to that of untreated bevacizumab-responsive tumors. Targeting GLUT3 or the increased glycolysis reliance in resistant tumors could unlock the potential of antiangiogenic treatments.

Figure 1. Increased glycolytic metabolism in a xenograft model of antiangiogenic therapy resistance.

(A) Bevacizumab treatment (10 mg/kg twice per week until volumetric endpoint reached) lowered intratumoral glucose compared with IgG control antibody treatment of subcutaneous U87 xenografts. P = 0.03, Wilcoxon-Mann-Whitney test, n = 4 replicates/group. (B) Cultured bevacizumab-resistant glioma cell line–derived xenograft (U87-Bev^R) cells exhibited more glucose uptake than cultured bevacizumab-sensitive (U87-Bev^S) cells in normoxia (P = 0.03) and hypoxia (P = 0.03). Wilcoxon-Mann-Whitney test, n = 4 replicates/group. 2NBDG, fluorescently labeled deoxyglucose analog. (C) Cultured U87-Bev^R cells exhibited the same baseline extracellular acidification rate (ECAR) as U87-Bev^S cells (P = 0.1–0.2, Wilcoxon-Mann-Whitney test), but greater ECAR when treated with inhibitors of mitochondrial oxidative phosphorylation, consistent with greater stress-associated glycolysis (range: P < 0.001 to P = 0.03, Wilcoxon-Mann-Whitney test). n = 46 samples/cell type/time point. (D) Cultured U87-Bev^R cells produced more pyruvate than U87-Bev^S cells in hypoxia (P = 0.008) but not normoxia (P = 0.2). Wilcoxon-Mann-Whitney test, n = 5 replicates/group. (E) ¹³C NMR spectroscopic measurement of pyruvate to lactate conversion revealed greater conversion in U87-Bev^R relative to U87-Bev^S. (F) ATP production assessed by a chemiluminescence assay revealed U87-Bev^R to produce more ATP production than U87-Bev^S in normoxic cells in DMEM (P = 0.03), with the difference persisting in hypoxia (P = 0.03). Wilcoxon-Mann-Whitney test, n = 4 replicates/group. For ECAR curves, error bars represent SDs. For box-and-whisker plots, the horizontal line in the box is the median, while the box extends from the 25th to 75th percentile and the whiskers from minimum to maximum values. *P < 0.05 and ***P < 0.001.

Figure 2. Increased GLUT3 expression in bevacizumab-resistant xenografts in a HIF-1α–dependent manner.

(A) Among glycolytic enzymes and regulatory factors, Western blot revealed increased GLUT3, increased lactate dehydrogenase (LDH), and decreased HK3 in bevacizumab-resistant glioma cell line–derived xenograft (U87-Bev^R) cells as the only significant changes (see Supplemental Figure 1 for quantification and statistical analysis) occurring between subcutaneous U87-Bev^R vs. bevacizumab-sensitive (U87-Bev^S) xenografts (n = 5 of each xenograft). Band sizes are indicated next to each lane. (B) By Western blot, bevacizumab increased GLUT3 expression in intracranial U87-Bev^R, not U87-Bev^S, xenografts. Bevacizumab lowered HIF-1α expression in intracranial U87-Bev^S xenografts. Bevacizumab tended to increase HIF-1α expression in intracranial U87-Bev^R xenografts, but this tendency did not reach significance (see Supplemental Figure 4 for quantification and statistical analysis). (C) Treatment of cultured U87-Bev^S and U87-Bev^R cells with 10 μM YC-1 (HIF-1α inhibitor) lowered GLUT3 protein expression in hypoxic U87-Bev^R cells and in normoxic U87-Bev^S cells.

Figure 3. Increased GLUT3 expression in isogeneic bevacizumab-resistant xenografts.

(A) Immunostaining of intracranial xenografts (n = 5/group) and (B) quantification of the immunostaining intensity revealed decreased HIF-1α staining in bevacizumab- versus IgG-treated bevacizumab-sensitive glioma cell line–derived xenograft (U87-Bev^S) (P = 0.008) and increased HIF-1α (P = 0.008) and GLUT3 (P = 0.008) staining in bevacizumab- versus IgG-treated bevacizumab-resistant (U87-Bev^R) cells (Wilcoxon-Mann-Whitney test). Magnification, ×20. Scale bars: 50 μm. For the box-and-whisker plot, the horizontal line in the box is the median, while the box extends from the 25th to 75th percentile and the whiskers from minimum to maximum values. **P < 0.01.

Figure 4. Increased GLUT3 expression in bevacizumab-resistant patient-derived xenografts.

(A) Immunostaining of intracranial patient-derived xenografts (PDXs) and (B) quantification of the immunostaining revealed that bevacizumab did not alter HIF-1α (P = 0.3) and lowered GLUT3 (P = 0.03) expression in bevacizumab-responsive GBM43 PDXs but increased HIF-1α (P = 0.04) and GLUT3 (P = 0.006) expression (Wilcoxon-Mann-Whitney test), with colocalized staining, in bevacizumab-resistant SF7796 PDXs. Magnification, ×20. Scale bars: 50 μm. (C) Glucose assay revealed that bevacizumab treatment decreased tumoral glucose in responsive GBM43 PDXs (P=0.03), but increased glucose in resistant SF7796 PDXs (P = 0.03). Wilcoxon-Mann Whitney test, n = 4/group. For box-and-whisker plots, the horizontal line in the box is the median, while the box extends from the 25th to 75th percentile and the whiskers from minimum to maximum values. *P < 0.05, **P < 0.01.

Figure 5. Increased GLUT3 expression in bevacizumab-resistant patient specimens.

(A) Absolute quantification PCR revealed increased GLUT3 transcript levels in all regions of patient bevacizumab-resistant glioblastomas (n = 3) compared with bevacizumab-naive glioblastomas (n = 5) (P < 0.001). For the dot plot, the horizontal line represents the mean and vertical line represents the SD. (B) Quantification of immunostaining of patient glioblastomas revealed 8-fold more GLUT3 in bevacizumab-resistant patient glioblastomas compared with before treatment (n = 5), with no change in bevacizumab-naive glioblastomas at recurrence compared with at diagnosis (n = 5). P = 0.008, paired Student’s t test. (C) Shown are representative images from each group at ×20 magnification (scale bars: 100 μm) with ×63 zoom (scale bars: 20 μm). For the box-and-whisker plot, the horizontal line in the box is the median, while the box extends from the 25th to 75th percentile and the whiskers from minimum to maximum values. *P < 0.05, **P < 0.01.

Figure 6. Decreased oxidative phosphorylation in a xenograft model of antiangiogenic therapy resistance.

(A) Flow cytometry of bevacizumab-resistant glioma cell line–derived xenograft (U87-Bev^R) and bevacizumab-sensitive (U87-Bev^S) cells treated with MitoTracker revealed no difference in number of mitochondria per cell between the cells (P = 0.07, Student’s t test). (B) Incubating cultured U87-Bev^S and U87-Bev^R cells with JC-1 fluorochrome revealed the red-to-green ratio, an indicator of mitochondrial membrane potential, to be lower in U87-Bev^R than U87-Bev^S cells in normoxia (P = 0.03) and hypoxia (P = 0.03). Furthermore, normoxic U87-Bev^R cells had a remarkably reduced membrane potential, on par with hypoxic U87-Bev^S cells (P = 0.2), indicating the level of mitochondrial impairment in cells from bevacizumab-resistant xenografts. Wilcoxon-Mann-Whitney test, n = 4/group. (C) Oxygen consumption rate (OCR) analysis over time when treating cells with 4 different mitochondrial inhibitors, as per the Seahorse extracellular flux analyzer protocol: oligomycin at 18 minutes, FCCP at 36 minutes, and antimycin A+rotenone at 54 minutes. U87-Bev^S had greater basal respiration before adding oligomycin (P = 0.03) and ATP production (basal respiration minus the proton leak between the post-oligomycin curve and the post-antimycin A+rotenone curve; P = 0.008) than U87-Bev^R, without a change in spare respiratory capacity (maximal respiration after adding FCCP minus basal respiration; P = 0.8) or proton leak (P = 0.3). Wilcoxon-Mann-Whitney test. For OCR curves, error bars represent SDs. (D) ATP production was higher in mitochondria isolated from U87-Bev^S cells and grown in 10 mM galactose and normoxia (20% oxygen), conditions selecting for oxidative phosphorylation, than mitochondria from U87-Bev^R cells (P = 0.03), a difference that was eliminated in hypoxia (P = 0.1). Wilcoxon-Mann-Whitney test, n = 4/group. (E) Western blot revealed decreased HSP60 expression in U87-Bev^R xenografts compared with U87-Bev^S xenografts (n = 5 tumors/group). For box-and-whisker plots, the horizontal line in the box is the median, while the box extends from the 25th to 75th percentile and the whiskers from minimum to maximum values. *P < 0.05.

Figure 7. Changes in triglyceride levels in a xenograft model of antiangiogenic therapy resistance.

(A) Oil Red O staining revealed increased numbers of neutral lipids in bevacizumab-resistant glioma cell line–derived (U87-Bev^R) subcutaneous xenografts compared with bevacizumab-sensitive (U87-Bev^S) subcutaneous xenografts (P = 0.03). Wilcoxon-Mann-Whitney test, n = 4/group. Magnification, ×63. Scale bars: 500 μm. (B) Oil Red O staining revealed increased numbers of neutral lipids in intracranial U87-Bev^R xenografts compared with intracranial U87-Bev^S xenografts (P = 0.003). Wilcoxon-Mann-Whitney test. Magnification, ×40. Scale bar: 750 μm. (C) TLC of intracranial xenografts (n = 6/group) revealed increased levels of triglycerides (P = 0.009) and decreased cholesterol (P = 0.002) with unchanged free fatty acids (P = 0.4) in U87-Bev^R xenografts compared with U87-Bev^S xenografts (Wilcoxon-Mann-Whitney test). (D) AdipoRed revealed increased intracellular triglyceride accumulation in U87 cells stressed with the mitochondrial inhibitors oligomycin, FCCP, antimycin A, and rotenone (P = 0.005). Wilcoxon-Mann-Whitney test, n = 22 replicates/group. (E) Western blot revealed no changes in triglyceride synthesis enzymes in U87-Bev^R xenografts compared with U87-Bev^S xenografts. (F) Oxygen consumption rate (OCR) measured to assess fatty acid oxidative metabolism using the Seahorse extracellular flux analyzer revealed that the basal respiration due to utilization of endogenous fatty acids was 18-fold higher in U87-Bev^R cells than U87-Bev^S cells (P = 0.008). Wilcoxon-Mann-Whitney test, n = 6/group. For box-and-whisker plots, the horizontal line in the box is the median, while the box extends from the 25th to 75th percentile and the whiskers from minimum to maximum values. *P < 0.05, **P < 0.01.

Figure 8. Bevacizumab-resistant xenograft cells exhibit greater survival in low glucose than bevacizumab-responsive cells, a phenotype replicated with GLUT3 upregulation.

(A) Culturing bevacizumab-sensitive glioma cell line–derived xenograft (U87-Bev^S) and bevacizumab-resistant (U87-Bev^R) cells in severely low (0.1 g/l) glucose revealed greater survival of U87-Bev^R than U87-Bev^S (P = 0.002). Wilcoxon-Mann-Whitney test, n = 6/group. (B) Culturing of U87-Bev^S cells for 24 hours in severely low glucose levels replicating those seen with antiangiogenic therapy (0.1 g/l) increased GLUT3 transcription relative to that seen at high glucose (4.5 g/l) or moderately low glucose levels, replicating those in tumors not treated with antiangiogenic therapy (0.3 g/l). P = 0.02, Student’s t test, n = 3/group. Error bars represent SDs. (C) Culturing of U87-Bev^S and U87-Bev^R cells in 4.5, 0.3, and 0.1 g/l glucose revealed that 24 hours in 0.1 and 0.3 g/l glucose induced GLUT3 protein expression assessed by Western blot in U87-Bev^S but still not to the levels seen in U87-Bev^R at any of the glucose concentrations. (D) Glucose uptake was greater in U87-Bev^S/GLUT3 cells than U87-Bev^S/empty vector (EV) cells. P = 0.03, Wilcoxon-Mann-Whitney test, n = 42/group. (E) U87-Bev^S/GLUT3, U87-Bev^S/GLUT1, and U87-Bev^S/EV cell survival in 4.5, 0.3, and 0.1 g/l glucose after 96 hours was affected by both the glucose concentration (P < 0.0001) and the nature of the overexpressed glucose transporter, with GLUT3 promoting more survival than GLUT1 (P < 0.0001), and interaction occurring between these variables. P < 0.0001, ANOVA, n = 6/group. For box-and-whisker plots, the horizontal line in the box is the median, while the box extends from the 25th to 75th percentile and the whiskers from minimum to maximum values. *P < 0.05, ***P < 0.001.

Figure 9. GLUT3 upregulation drives the metabolic and therapeutic response changes associated with resistance to antiangiogenic therapy in a targetable manner.

(A) Extracellular acidification rate (ECAR) over time, a marker of glycolysis, was higher in cultured bevacizumab-sensitive glioma cell line–derived xenograft cells transfected with GLUT3 (U87-Bev^S/GLUT3) cells vs. U87-Bev^S/empty vector (EV) cells treated with various oxidative phosphorylation inhibitors (range: P < 0.0001–0.02). Wilcoxon-Mann-Whitney test, n = 46/group. (B) Oxygen consumption rate (OCR) over time after treatment with 4 different mitochondrial inhibitors, as per the Seahorse extracellular flux analyzer protocol: oligomycin at 18 minutes, FCCP at 36 minutes, and antimycin A+rotenone at 54 minutes. Basal respiration (rate prior to oligomycin injection minus nonmitochondrial respiration) and ATP production (basal respiration minus proton leak) were higher in U87-Bev^S/EV than U87-Bev^S/GLUT3. P < 0.0001, Wilcoxon-Mann-Whitney test, n = 46/group. For ECAR and OCR curves in A and B, error bars represent SDs. (C) Glycogen synthase kinase-3β inhibitor GSK-3 IX (10 μM) caused more cell death in bevacizumab-resistant (U87-Bev^R) cells than U87-Bev^S cells after 24 (P = 0.04), 48 (P = 0.005), and 72 (P = 0.006) hours. Student’s t test, n = 6/group. For box-and-whisker plots, the horizontal line in the box is the median, while the box extends from the 25th to 75th percentile and the whiskers from minimum to maximum values. *P < 0.05, **P < 0.01. (D) Subcutaneous xenografts derived from U87-Bev^S/GLUT3 and U87-Bev^S/EV cells were treated with bevacizumab or control IgG antibody (n = 6/group). After 20 days, GLUT3 overexpression promoted tumor growth in vivo (P = 0.006) while bevacizumab reduced tumor growth (P < 0.0001), with no interaction between these variables (P = 0.1, ANOVA). For tumor volumes, values and error bars represent mean ± SEM.