Loss of FBXW7 Correlates with Increased IDH1 Expression in Glioma and Enhances IDH1-Mutant Cancer Cell Sensitivity to Radiation

IDH1 is an important enzyme that function in various cellular events, including metabolism, redox homeostasis, and epigenetic regulation. IDH1 mutations have been widely validated in around 80% of low-grade gliomas and secondary glioblastomas (1). The hotspot mutations (such as in codon R132) occur in the catalytically active site of IDH1 enzymes, and the main driver of tumorigenesis is the oncometabolite production of 2-hydroxyglutarate (2HG; ref. 2). The accumulated 2HG leads to tumorigenesis by inhibiting α-ketoglutarate (αKG)-dependent enzymes, causing cellular alterations in a plethora of cellular metabolism. However, multiple clinical studies suggest that patients with glioma with IDH1/2 mutation have a favorable response to conventional radiotherapy and chemotherapy, and longer median survival times than their wild-type (WT) counterparts (3, 4). In addition, IDH1 is the most upregulated NADPH-producing enzyme in GBM compared with normal brain tissue and is further increased following radiation (5). Therefore, elucidation of mechanisms responsible for IDH1 expression is important for providing insight into strategies to perform stratified treatment in glioma.

Ubiquitination regulates many essential cellular processes in eukaryotes (6). This posttranslational modification is typically achieved by E1, E2, and E3 enzymes that sequentially catalyze reactions, leading to covalent attachment of ubiquitin and substrate protein proteasomal degradation (7). FBXW7 is the substrate binding unit of SCF (SKP1-Cullin1-F-box)-type E3 ligase, which recognizes substrates with conserved phosphor-degron motifs through its WD40 repeats domain (8). As an important tumor suppressor, dominant-negative mutations of FBXW7 have been found in a variety of tumors, and multiple key oncogenic substrates of FBXW7 have been identified, including cyclin E, c-MYC, c-JUN, NOTCH that are involved in the regulation of cell division, apoptosis, differentiation, and some classical oncogenic signaling pathways (8). Although mutations in FBXW7 are relatively rare, multiple studies reported that FBXW7 expression is downregulated in human gliomas (9–11). Specifically, Hagedorn and colleagues found that the expression level of FBXW7 was significantly reduced in more than 80% glioblastoma (10). p53 mutation can downregulate FBXW7 expression, which leads to oncogenic protein c-MYC accumulation and promotes the development of gliomas (12). Moreover, FBXW7 expression has also been reported to inversely correlate with glioma histology and positively with patient survival (11).

Aberrant metabolism is considered to be one of the hallmarks of glioma. The tumor suppressor FBXW7 has also been linked to the transformation of metabolism (13, 14). SREBP family of transcription factors, well-established FBXW7 substrates, plays a critical role in lipid metabolism by regulating the expression of a range of enzymes required for lipid synthesis (15–18). IDH1 is a key enzyme that can support lipogenesis either through NADPH production or reductive carboxylation, facilitating the flux of carbon to lipids (19, 20). The IDH1 gene promoter has an identifiable consensus SRE, and SREBP could bind directly to the SRE sequence (21, 22). These findings have prompted us to hypothesize that FBXW7 is a potential key negative regulator of IDH1. Moreover, given that loss of FBXW7 confers radiosensitization by nonhomologous end joining (NHEJ), and IDH1 mutations induce an homologous recombination (HR) defect and NADPH shortage, we predict that IDH1-mutant tumors with lower FBXW7 expression might achieve maximal efficacy of radiotherapy (23–26). In this study, we went on to determine the mechanisms as well as the consequences of FBXW7 depletion in IDH1 WT and mutant cells on therapeutic responses to radiation.

The patient population represents a randomly selected subgroup from a clinical series that includes all 86 patients undergoing surgical resection between January 2012 and December 2015 in the Department of Neurosurgery of the Fifth Affiliated Hospital of Zhengzhou University (Zhengzhou, P.R. China). The criteria of patient selection are described in Supplementary Materials and Methods. The detailed clinicopathologic variables of 86 glioma samples are shown in Supplementary Table S2. This study was approved by the Zhengzhou University Institutional Review Board and was conducted in accordance with the Declaration of Helsinki. Human glioma patient samples were used according to the guidelines of the Ethical Committee of Zhengzhou University (Zhengzhou, P.R. China). Each patient provided written informed consent.

U87, T98G, HT1080 cell lines and U87 IDH1^R132H/+-mutant isogenic line were obtained from ATCC. U87 and T98G cells were cultured in DMEM (Gibco) supplemented with 10% FBS (Life Technologies), 2 mmol/L l-glutamine, penicillin, and streptomycin (Sigma). HT1080 cells were cultured in Eagle Minimum Essential Medium supplemented with 10% FBS, 2 mmol/L l-glutamine, penicillin, and streptomycin. Normal Human Astrocytes were purchased from Lonza and cultured in Astrocyte Growth Medium BulletKit (Lonza). HCT116 parental and IDH1 knock-in clones (R132H/+) were obtained from Horizon Discovery and cultured in DMEM supplemented with 10% FBS, 2 mmol/L l-glutamine, penicillin, and streptomycin. All cell lines were tested for Mycoplasma every 3 months and authenticated by short tandem repeat profiling. The number of passages between cells thawing and use in all the experiments was limited in 10 passages. Doxycycline (D9891), 25-Hydroxycholesterol (H1015), and N-acetyl cysteine (A7250) were purchased from MilliporeSigma. AGI-5198 was purchased from Selleck (S7185).

The single-guide RNA (sgRNA)/Cas9 dual expression vector PX459 (pSpCas9(BB)-2A-Puro) was obtained from Addgene. The sequences for sgRNAs targeting FBXW7 were: sgFBXW7-1, AGTGGAAGTATGCCCATATA; sgFBXW7-2, CAGCTCAGACATGTCGCTAC; sgFBXW7α, TGTGGTAACCGAATAGTTAG; sgFBXW7β, TTCGGCGTCGTTGTTGCCCT; sgFBXW7γ, TTCGGCGTCGTTGTTGCCCT. To generate doxycycline-inducible FBXW7 knockout, TLCV2 plasmid backbone containing P2A-puromycin was used. All the above constructs were confirmed by Sanger sequencing.

Cells were washed with 10% NaCl, collected in ice-cold methanol. Samples were further extracted with addition of chloroform at 4°C for 20 minutes. The organic and aqueous phases were separated by centrifugation and the aqueous phase was collected and evaporated to dryness. The residue was dissolved in 0.1% formic acid and injected for analysis. Analysis was performed employing a Waters Acquity UPLC system coupled with a Waters Xevo triple quadrupole mass spectrometer.

NADPH/NADP+ ratios were immediately measured by using fresh cells and the NADPH/NADP+ Quantitation Colorimetric kit (BioVision), according to the manufacturer's instructions. Protein concentrations of the cells lysates were used for normalization.

Exponentially growing cells were treated with radiation or temozolomide (TMZ) and then replated at cloning densities. Cells were grown for 11 to 14 days and then fixed/stained with methanol-acetic acid and crystal violet, respectively, and scored for colonies of 50 cells or more. The cytotoxicity of FBXW7 knockout in the absence of radiation treatment was calculated as the ratio of surviving treated cells relative to surviving control cells. Radiation survival data from treated cells were corrected for plating efficiency by normalizing to unirradiated control cells. Cell survival curves were fitted using the linear quadratic equation, and the mean inactivation dose was calculated and used to determine the radiation enhancement ratio. Radiosensitization is indicated by a radiation enhancement ratio of significantly greater than 1.

Irradiations were carried out using a Philips RT250 (Kimtron Medical) at a dose rate of approximately 2 Gy/minute. For flank and orthotopic tumor irradiation, animals were anesthetized and positioned such that the apex of each tumor was at the center of an aperture in the secondary collimator, with the rest of the mouse shielded from radiation.

For immunofluorescence experiments, cells were grown and treated on coverslips in 12-well dishes. Following treatment, cells were fixed and stained with antibodies recognizing γH2AX antibody (clone JBW301; Millipore) and DAPI. Samples were imaged with an Olympus IX71 FluoView confocal microscope (Olympus America) with a 60× oil objective. For flow cytometry, cells were trypsinized, washed with ice-cold PBS, and fixed in ice-cold 70% ethanol. Thereafter, samples were incubated with the γH2AX antibody overnight at 4°C followed by incubation with a FITC-conjugated secondary antibody (Sigma). γH2AX positivity was quantified by using a gate arbitrarily setting on the untreated sample to define a region of positive staining (∼5%). This gate was then overlaid on the FBXW7 knockout and/or radiation samples.

The I-SceI-based NHEJ and HR assay was performed as described earlier with modifications (27, 28). Parental and IDH1^R132H U87 cells stably expressing the pEJ5 or DR-GFP reporter were seeded in 12-well plates. Cells with or without FBXW7 deletion were then infected with I-SceI–expressing adenovirus. Cells were harvested after 44 hours and the percentage of GFP-positive cells (indicative of NHEJ or HR repair) was quantitated by flow cytometry.

U87 and HCT116 were washed and solubilized in ice-cold methanol/water (85%/15%). Samples were centrifuged (13,000 rpm) for 10 minutes and evaporated in a vacuum concentrator plus (Eppendorf) at 30°C. The samples were then resolved in 100 μL water methanol/water (50%/50%) and subsequently transferred to the LC/MS system. Quantification of the pentose phosphate pathway (PPP) metabolites was normalized to the protein concentration.

All procedures involving mice were approved by the University Committee on Use and Care of Animals at the Zhengzhou University (Zhengzhou, P.R. China; SYXK2016-0002). HCT116 (WT and IDH1^R132H/+) and U87 IDH1^R132H/+ cells (1 × 10⁶) carrying a doxycycline-inducible shFBXW7 or sgFBXW7 were resuspended in 1:1 PBS:Matrigel (BD Biosciences) and injected into the bilateral dorsal flanks of female nude mice. Once tumor volumes reached around 100 mm³, mice were administrated with doxycycline (2 mg/mL) via drinking water and changed daily. For radiation treatment, animals were anesthetized with isoflurane and positioned such that the apex of each flank tumor was at the center of a 2.4-cm aperture in the secondary collimator, with the rest of the mouse shielded from radiation. Tumor volumes were measured twice weekly using digital calipers and the formula (π/6) (Length × Width²).

A total of 1 × 10⁵ U87 IDH1^R132H/+ cells expressing luciferase in 2 μL of PBS were stereotactically implanted into the brains of female athymic nude mice (6–8 weeks of age). Mice were treated as indicated and then observed until they became moribund, at which point they were sacrificed. Bioluminescent images of orthotopic U87 IDHR132H/+ tumors were taken via bioluminescence (IVIS Spectrum, PerkinElmer).

qPCR and metabolomic data were analyzed by GraphPad Prism 5. Student t tests and one-way ANOVAs were used to determine inferential statistics. Quantitative data are presented as the mean ± SEM.

To gain insight into the expression of FBXW7 in glioma, we first compared the expression levels of FBXW7 in different grades of gliomas and normal brain tissues in The Cancer Genome Atlas (TCGA) database. As expected, the expression of FBXW7 was highest in normal brain tissues and significantly decreases in all grades of gliomas (Supplementary Fig. S1A), consistent with the tumor suppressor function of FBXW7 in glioma (10). Furthermore, we analyzed the FBXW7-associated genes in TCGA glioma cohort using the LinkedOmics database (29). Interestingly, as shown in the volcano plot, we found IDH1 as one of genes exhibiting highly negative correlation with FBXW7 (Fig. 1A), and we observed a statistically significant correlation between FBXW7 and IDH1 by Pearson correlation analysis in both TCGA and publicly available gene expression data from Chinese Glioma Genome Atlas (CGGA) database (Fig. 1B). Compared with the correlation in the TCGA cohort, FBXW7 expression demonstrates stronger negative correlation with IDH1 expression in both CGGA RNA sequencing and microarray data (Fig. 1B). Furthermore, the negative association between FBXW7 and IDH1 expression was also observed in the subsets of each dataset based on the World Health Organization (WHO) grade system (Supplementary Fig. S1B and S1C). However, we did not find a significant correlation between IDH2 and FBXW7 mRNA levels in all analyses of TCGA and CGGA datasets except the CGGA693 (Supplementary Fig. S2A).

Next, we sought to assess whether FBXW7 protein expression is negatively associated with IDH1 in human gliomas, IHC was performed in 86 untreated resected glioma samples. The 86 samples consisted of (i) 10 WHO grade I gliomas, with eight cases of pilocytic astrocytoma and two cases of subependymal giant cell astrocytoma; (ii) 38 WHO grade II gliomas, including 23 cases of astrocytoma and 15 cases of oligodendroglioma; (iii) 26 WHO grade III gliomas, with 14 cases of oligodendroglioma and 12 cases of anaplastic astrocytoma; (iv) 12 WHO grade IV glioblastoma samples. FBXW7 expression levels were determined by H-score, which was further divided into high (score ≥ 100) and low (score < 100) expression. Among the 86 samples, high FBXW7 expression was found in 72 samples (83.7%), with 93.75% (45/48) in low-grade gliomas (WHO grade I and II) and 71.05% (27/38) in high-grade gliomas (WHO grade III and IV), while 14 samples (16.3%) showed low FBXW7 expression, with only three samples (6.25%, 3/48) in low-grade gliomas and 11 samples (28.95%, 11/38) in high-grade gliomas (Supplementary Tables S2 and S3). The percentage of high FBXW7 expression samples in the low-grade tumors was significantly higher related to the high pathologic grade tumors (χ² = 8.017, P = 0.005; Fig. 1C; Supplementary Table S3). The correlation between FBXW7 expression and clinicalpathologic parameters of the glioma samples were also analyzed using the FBXW7 expression level against gender, age, tumor size, dizziness history, epileptic condition, preoperative Karnofsky Performance Status (KPS) score, glioma cell components, and tumor site. We found that FBXW7 expression is positively correlated with tumor diameter (χ² = 10.994, P = 0.001) and epileptic condition (χ² = 14.918, P < 0.001; Supplementary Table S2).

The expression of IDH1 in the same cohort was examined and scored in parallel with FBXW7. On the basis of the IDH1 H-score in each sample (high expression, score ≥ 100; low expression, score < 100), we observed that 86.06% (74/86) samples demonstrated high IDH1 expression, and 13.94% (12/86) gliomas had low IDH1 expression (Fig. 1C; Supplementary Table S4). In contrast to FBXW7, the percentage of high IDH1 samples in high-grade gliomas was significantly higher compared with the percentage in the lower-grade gliomas (94.73%, 36/38 vs. 79.17%, 38/48, χ² = 4.282, P = 0.036; Supplementary Table S4). Importantly, in line with the correlation at the transcription level, we found significant negative correlation between FBXW7 and IDH1 expression levels in the 86 glioma samples (χ² = 30.454, P < 0.001; Table 1). Therefore, the above data clearly show the negative association between FBXW7 and IDH1 expression, suggesting a potential regulation of IDH1 expression by FBXW7 in gliomas.

To determine whether FBXW7 regulates IDH1 expression in glioma cells, U87 and T98G cells were cultured and the effects of CRISPR/Cas9-mediated knockout of FBXW7 on IDH1 expression were measured. Deletion of FBXW7 by two independent FBXW7 sgRNAs significantly increased the protein levels of IDH1 in both cell lines (Fig. 1D). To further examine whether the regulation of IDH1 expression by FBXW7 is at the transcriptional level, we detected the IDH1 mRNAs and found transcript levels of IDH1 were significantly elevated upon FBXW7 knockout (Fig. 1E). Our extended analysis indicates that IDH2 expression was not affected by FBXW7 depletion in both cell lines (Supplementary Fig. S2B and S2C).

FBXW7 encodes three protein isoforms (α, β, and γ) resulting from alternative splicing that contains conserved interaction domains at the C-terminus but isoform-specific domains at the N-terminal region (Supplementary Fig. S2D; ref. 8). Introducing FBXW7α or γ, but not β, isoform-specific siRNAs or sgRNAs greatly increased IDH1 mRNA and protein levels (Fig. 1F and G; Supplementary Fig. S2E–S2G). The WD40 domain of FBXW7 is crucial for subsequent ubiquitination and proteasomal degradation of its substrates. We next determined whether cancer-derived mutations in the FBXW7 WD40 domains would abrogate the negative regulation. In contrast to WT FBXW7, neither R465H or R479Q mutants (in the third WD40 repeat; Supplementary Fig. S2D) was found to decrease IDH1 expression when overexpressed (Fig. 1H). In addition, ERK kinase is recently identified to phosphorylate FBXW7 and promote FBXW7 degradation (30). In line with the effect of FBXW7 deletion, overexpression of WT ERK1, rather than kinase-dead ERK1, significantly lead to reduced FBXW7 expression and higher IDH1 expression compared with control (Fig. 1I and J). Taken together, these results suggest FBXW7 negatively regulates IDH1 expression at the transcriptional level, which is likely through mediating FBXW7 substrate(s) turnover.

We next sought to determine whether one or multiple well-established FBXW7 substrates are involved. FBXW7 controls proteasome-mediated degradation of several oncoproteins, which exhibit regulatory role in cellular metabolism, including c-MYC, c-JUN, and SREBP1. We first found accumulated protein levels of all three substrates upon FBXW7 deletion in both U87 and T98G cells, with predominantly occurring in c-MYC and SREBP1 (Fig. 2A). To discern the contributions of the three substrates to the IDH1 expression regulated by FBXW7, c-MYC, c-JUN, or SREBP1 was individually silenced. While the enhanced expression of IDH1 upon FBXW7 knockout was unaffected in response to c-MYC or c-JUN siRNA, SREBP1 silencing caused a significant reduction of IDH1 expression levels in the FBXW7-knockout U87 and T98G cells, indicating that FBXW7 regulates IDH1 expression through mediating SREBP1 turnover (Fig. 2B; Supplementary Fig. S3A). This was confirmed by using a second independent siRNA targeting SREBP1 (Supplementary Fig. S3B). In support of our findings, Ricoult and colleagues identified IDH1 as a key transcriptional target of SREBP1 across a spectrum of human cancers (21). Because sterol abundance strongly affects SREBP1 processing and activity in cells, we tested the effects of exogenous cholesterol on IDH1 expression in response to FBXW7 deletion by using 25-hydroxycholesterol (25-HC), which inhibits SREBP processing. As expected, 25-HC suppressed the expression of IDH1 transcript and protein in cells knockout of FBXW7, although SREBP1 protein levels were higher than control cells (Fig. 2C).

IDH1 functions in the cytosol to convert isocitrate to aKG. We found a 2- to 3-fold increase of aKG levels upon FBXW7 deletion in both U87 and T98G cells (Fig. 2D). This increase in aKG levels was similar to that following IDH1 overexpression (Supplementary Fig. S3C). Importantly, aKG levels remained unaffected following FBXW7 knockout when cells were depleted with SREBP1 (Fig. 2E), indicating that the increased aKG level is the consequence of SREBP1 accumulation in response to FBXW7 knockout. Conversion of isocitrate to aKG catalyzed by IDH1 also results in the generation of NADPH, and this provides a major source of NADPH for cells in addition to the PPP (2, 31). Thus, we analyzed the ratio of NADPH over NADP+ in FBXW7-deleted cells in medium without glucose but supplemented with glutamine and pyruvate to circumvent the contribution of the PPP to NADPH generation. We found that under these conditions, the NADPH/NADP+ ratio was significantly increased in the FBXW7 knockout cells (Fig. 2F), similar to cells transfected with IDH1 (Supplementary Fig. S3D). Again, this elevation in NADPH/NADP+ ratio was due to FBXW7 knockout-mediated SREBP1 accumulation (Fig. 2G).

Mutant IDH1 catalyzes the neomorphic production of 2HG. We next investigated whether FBXW7 also regulates the expression of mutant IDH1. To this end, we used isogenic U87 IDH1^R132H/+ and HCT116 IDH1^R132H/+ cells, as well as HT1080 cells, a fibrosarcoma-derived cell line that endogenously carries the IDH1 R132C mutation. In these cells, we introduced the doxycycline-inducible two independent sgRNAs targeting FBXW7, and observed that upon FBXW7 knockout, both mutant and total IDH1 transcription is significantly induced (Fig. 3A; Supplementary Fig. S4A and S4B). This elevation is also reflected in the IDH1 protein levels (Fig. 3B). In line with this, FBXW7 expression is decreased in both low grade and high grade of IDH1-mutant gliomas, which is to the similar extent in WT samples compared with normal brain tissues, and demonstrates strong negative correlations with mutant IDH1 expression in the above TCGA and CGGA datasets (Fig. 3C; Supplementary Fig. S4C–S4E). In addition, consistent with the observation in WT IDH1 cells, the induction of mutant IDH1 expression was dramatically eliminated by SREBP1 depletion when FBXW7 was deleted in response to doxycycline in U87 IDH1^R132H/+ and HT1080 cells (Fig. 3D; Supplementary Fig. S4F and S4G).

We further evaluated the levels of 2HG, which were induced by around 3- to 5-fold in the FBXW7 knockout U87 IDH1^R132H, HT1080, and HCT116 IDH1^R132H cells (Fig. 3E; Supplementary Fig. S4H). Again, 2HG increase by FBXW7 deletion was abrogated when the cells were treated with siRNAs targeting SREBP1 (Fig. 3F; Supplementary Fig. S4I). Collectively, our results demonstrate that FBXW7 not only regulates WT IDH1, but mutant IDH1 expression, and importantly the production of the oncometabolite 2HG in IDH1-mutant cells.

IDH1/2-mutated glioma cells are known to confer defects in HR, whereas the other major DNA double-strand break repair pathway NHEJ remains intact (25). On the other hand, inactivation of FBXW7 increases cancer cells sensitivity to radiation through abrogating NHEJ repair (23). We therefore hypothesize that deletion of FBXW7 in IDH1-mutant cells might enhance the radiosensitization effect by inhibiting both NHEJ and HR. We used the IDH1 WT and mutant isogenic U87 and HCT116 cells and found that knockout of FBXW7 renders both cell lines sensitive to radiation treatment in the clonogenic assay, but the radiation enhancement ratios in mutant IDH1 cells were significantly higher compared with WT IDH1 cells (Fig. 4A–C). We next determined the capacity of double-strand DNA (DSB) repair in the FBXW7-deleted IDH1 WT and mutant cells after radiation by analyzing the phosphorylated (S139) histone 2AX (γH2AX), which serves as a well-established biomarker of DSBs (32). As shown in Fig. 4D and E; Supplementary Fig. S5A and S5B, higher endogenous levels of residual γH2AX foci in the FBXW7-deleted IDH1-mutant U87 and HCT116 cells were observed relative to the related parental cells with FBXW7 depletion after radiation. Flow cytometry assay also revealed that, upon FBXW7 knockout, the γH2AX levels were elevated significantly in the IDH1-mutant HCT116 cells compared with WT cells at 6, 16, and 24 hours following radiation (Fig. 4F and G). Normal human astrocytes have been demonstrated to more resistant to radiation than glioma cells (33). Unlike glioma cells, deletion of FBXW7 in NHA cells showed less radiosensitivity, NADPH/NADP+ ratio, and γH2AX focus formation, suggesting a cancer cell–specific effect of radiation (Supplementary Fig. S5C–S5G). To determine whether the persistent DNA damage seen in FBXW7-deleted IDH1 mutant cells was linked to defective DSB repair, we assessed the effects of FBXW7 knockout together with IDH1 status on both HR and NHEJ repair by using IDH1 WT and mutant U87 cells containing a reporter of HR or NHEJ, respectively (27). As expected, deletion of FBXW7 in IDH1 WT cells showed decreased NHEJ repair, but had no effect on HR, whereas the efficiency of DNA DSB repair by both NHEJ and HR declined significantly in response to FBXW7 knockout in the context of mutant IDH1 (Fig. 4H and I; Supplementary Fig. S5H and S5I). Therefore, defective HR and NHEJ repair contributes to the radiosensitization of IDH1-mutant cells upon loss of FBXW7.

Mutant IDH1 proteins not only exhibit decreased activity for the WT reaction to produce NADPH but gain to modify its activity and catalyze the NADPH-mediated reduction of aKG to 2HG (31). Disruption of NADPH homeostasis in IDH1-mutant cells has an increased sensitivity to oxidative stress introduced by radiation treatment (26). We therefore examined whether FBXW7 deletion in the context of IDH1 mutation further increases NADPH consumption due to high expression of IDH1 expression. As expected, the NADPH/NADP+ ratio was reduced in R132H/+ U87 and HCT116 cells compared with WT, and further decreased upon knockout of FBXW7 (Fig. 5A; Supplementary Fig. S6A). The oxidative PPP is thought to be the key pathway through which NADPH is regenerated in the cytosol (34, 35), and IDH1 mutation cells exhibit increased PPP flux (26) (Fig. 5B). We first applied PPP intermediate metabolites analysis to determine the effects of FBXW7 deletion on PPP activity. As shown in Supplementary Fig. S6B, FBXW7 deletion did not change the production of PPP intermediates in parental U87 cells. In line, we did not observe any correlation between expression of FBXW7 and genes encoding some key PPP enzymes (G6PD, PGD, TALDO1, and TKT; Supplementary Fig. S6C). In addition, knockout of FBXW7 had no effect on NADPH oxidase 1 (NOX1) expression and NADPH oxidase activity (Supplementary Fig. S6D and S6E). In contrast, knockout of FBXW7 in IDH1^R132H/+ cells further stimulated a 2-fold increase of the oxPPP intermediate gluconate-6-phosphate (G6P) production (Fig. 5C and D). Treatment of IDH1-mutant cells with AGI-5198, a pharmacologic inhibitor of mutant IDH1, led to a reversal of G6P levels, and FBXW7 deletion showed comparable amount of G6P (Supplementary Fig. S6F). These data indicate that instead of directly regulating PPP activity, FBXW7 depletion increases mutant IDH1 protein levels and creates an NADPH sink, which activates PPP to compensate for the NADPH demand. However, despite the PPP activity is relative increased, the NADPH produced may be used to support 2HG synthesis, leading a lower NADPH/NADP+ ratio. Furthermore, we investigated whether radiosensitivity of FBXW7-deleted IDH1-mutant cells is partially caused by oxidative stress resulting from NADPH shortage, and we treated R132H/+ U87 and HCT116 cells in the presence or absence of FBXW7 with the NADPH surrogate and ROS scavenger N-acetylcysteine (NAC). Application of NAC increased the surviving fractions of FBXW7-deleted R132H/+ cells more than those of control cells after treatment with radiation (Fig. 5E–G). This finding suggests that upon FBXW7 deletion, higher NADPH consumption-induced oxidative stress is another key factor mediating the sensitization to radiation of IDH1-mutant cells.

On the basis of the significant enhancement of radiosensitization upon FBXW7 deletion in the IDH1-mutant cells in vitro, we further tested the therapeutic efficacy of radiation in animal tumor models. We first established flank xenografts using parental or IDH1^R132H/+ HCT116 cells harboring doxycycline-inducible short hairpin RNA (shRNA) targeting FBXW7 in nude mice. An inducible knockdown model was chosen rather than a stable shRNA knockdown or CRISPR/Cas9 knockout because it allowed for initial tumor growth to occur with intact FBXW7, thereby better modeling a therapeutic intervention. Once tumors were about 100 mm³ in size, animals were randomized into eight treatment groups: (i) WT Ctrl, (ii) WT doxycyline (FBXW7 KD), (iii) WT radiation, (iv) WT combined doxycycline and radiation, (v) R132H/+ Ctrl, (vi) R132H/+ doxycycline (FBXW7 KD), (vii) R132H/+ radiation, (viii) R132H/+ combined doxycycline and radiation (Fig. 6A). Consistent with a previously characterized oncogenic role of IDH1^R132H (25), we found a modest growth advantage of IDH1^R132H/+ HCT116 tumors in athymic nude mice (Fig. 6B). We also observed accelerated growth in both WT and mutant IDH1 cells upon doxycycline administration (Fig. 6B), likely due to the elimination of tumor suppressor role of FBXW7. As expected, radiation alone significantly slowed tumor growth compared with untreated tumors in each genetic background as analyzed using a linear mixed effects model (P < 0.01; Fig. 6B). Importantly, combined doxycycline and radiation significantly slowed tumor growth in IDH1^R132H/+ tumors compared with WT tumors (P < 0.01, Fig. 6B). Immunoblotting confirmed that administration of doxycycline had the intended effects on FBXW7 expression in both IDH1 WT and mutant xenografts (Fig. 6C).

We then asked whether depletion of FBXW7 dramatically enhanced radiosensitization in U87 IDH1^R132H/+ xenografts in vivo. Consistent with the findings in HCT116 cells, doxycycline treatment (FBXW7 knockdown) significantly inhibited U87 IDH1^R132H/+ tumor growth in response to radiation in mouse flank, although FBXW7 silenced tumor exhibited growth advantage in the absence of radiation (Fig. 6D). In addition, the size and weight of tumors in mice treated with doxycycline and radiation demonstrated the greatest decrease compared with other groups at 4 weeks after treatment (Fig. 6E; Supplementary Fig. S7A). This effect could be reproduced in U87 IDH1^R132H/+ cells with doxycycline-inducible CRISPR/Cas9-mediated FBXW7 knockout (Supplementary Fig. S7B). To further verify this observation in an orthotopic xenograft model, the U87 IDH1^R132H/+ cells constitutively expressing luciferase were used to monitor orthotopic xenograft growth by using luminescent imaging (Fig. 6F). The data showed that doxycycline and radiation treatment markedly slowed down orthotopic U87 IDH1^R132H tumor growth (Supplementary Fig. S7C), and significantly prolonged the overall survival of intracranial glioma-bearing mice in comparison with control-, doxycycline-, or radiation-treated alone (Fig. 6G). Interestingly, in depth analysis of CGGA data revealed that although lower FBXW7 exhibits unfavorable prognosis in gliomas, patients with lower expression of FBXW7 appeared to gain more benefit from treatment of radiotherapy, especially in the patients harboring the IDH1 mutations (Supplementary Fig. S7D). Collectively, these data demonstrate that silencing of FBXW7 strikingly renders IDH1-mutant tumor xenografts sensitive to radiation, suggesting that IDH1-mutant gliomas with lower FBXW7 expression might be more sensitive to radiotherapy.

Compared with untransformed cells, the metabolic rewiring of cancer cells results in the acquisition of unique characteristics, which allows to selectively target cancer cells while sparing normal cells. IDH1 mutations are such attractive therapeutic targets, and identifying upstream components of IDH1 regulation could be potentially beneficial. In the current study, we demonstrate that the expression of FBXW7 is negatively associated with IDH1 expression in gliomas. FBXW7 knockout in cancer cells increases SREBP1 stability, which activates the expression of WT or mutant IDH1 in different cell settings and elevates production of the oncometabolite 2HG in IDH1-mutant cells. Loss of FBXW7 renders IDH1-mutant cells highly sensitive to radiation treatment, which is attributed by suppression of both NHEJ and HR repair, as well as disrupted NADPH homeostasis (Fig. 6H).

In vitro and in vivo studies have revealed emerging links between FBXW7 and metabolic regulation, and FBXW7 functions as a negative regulator of lipid synthesis and metabolism (18, 36, 37). In line with previous observations that IDH1 promoter is fully activated by SREBP1 in cells, our work validated that FBXW7/SREBP1 is a key negative regulatory axis of IDH1 expression in cancer cells (21). Through this mechanism in IDH1 WT cells, we found a significant increase of NAPDH levels upon deletion of FBXW7. Inhibition of WT IDH1 has been reported to decrease NADPH levels and lead to GBM cells sensitive to radiation (5, 38). On the other hand, FBXW7 is rapidly recruited to the DNA damage lesions in a PARP1- and ATM-dependent manner and facilitates NHEJ repair of DSBs (23, 39). Depletion or inhibition of FBXW7 contributes to genomic instability and causes cancer cells sensitive to radiation (23, 24). Our data show that FBXW7 deletion radiosensitizes IDH1 WT cells, even though the NADPH/NADP+ ratios are increased, suggesting that NHEJ deficiency upon loss of FBXW7 may overwhelm NADPH protection against oxidative damage in response to radiation treatment.

2HG produced by oncogenic mutant IDH1 is necessary and sufficient for inducing HR deficit and BRACness phenotype (25). Loss of FBXW7 in IDH1-mutant cells causes simultaneous suppression of both NHEJ and HR, two major pathways for DSB repair, contributing to higher radiosensitivity in cancer cells harboring IDH1 mutation than WT cells. In addition, the high NADPH demands by mutant IDH1 during 2HG synthesis also confer cells sensitive to radiation by limiting the ability of cells to neutralize radiation-induced oxidative stress (26, 40). Our data support that FBXW7 deletion in IDH1-mutant cells further increased NADPH consumption, while application of NAC partially reversed the radiosensitization mediated by overcoming NADPH shortage (Fig. 5). IDH1-mutant cells increase PPP flux to regenerate NADPH in respond to 2HG synthesis (26). In line with this, FBXW7 knockout further enhanced PPP activity to meet the substrate demands of increased mutant IDH1. Contrary to a previous report suggesting a positive regulation of FBXW7 on PPP in macrophages (41), we did not find a direct modulation PPP activity by FBXW7 in WT cells or IDH1-mutant cells treated with AGI-5198, suggesting that loss of FBXW7 augments PPP activity through promoting mutant IDH1 expression. Notably, our survival analysis in CGGA693 indicates IDH1-mutant patients with lower FBXW7 expression benefit most from radiotherapy. Therefore, our data suggest an urgent need to test FBXW7 as a biomarker to predict radiosensitivity in IDH1-mutant gliomas and other cancers. Given that the addition of TMZ to radiotherapy boosts overall survival of patients with GBM (42), our study briefly extended to test the effect of FBXW7 deletion on cancer cell sensitivity to TMZ. FBXW7 KO sensitizes both WT and IDH1-mutant U87 cells to TMZ to a similar extent (Supplementary Fig. S8A–S8D). In addition, the survival of patients with glioma with lower FBXW7 expression is similar to that of patients expressing higher FBXW7 after TMZ treatment, whereas lower FBXW7-expressing patients tend to have worse prognosis without TMZ therapy (Supplementary Fig. S8E). This appears to imply that tumors with lower FBXW7 expression sensitize to the combination of TMZ and radiation, which deserves further investigation.

Novel biomarkers for patient stratification to cancer therapy are a key pillar of precision medicine to improve therapeutic efficacy. In the work reported here, we demonstrate that lower FBXW7 expression in the settings of IDH1 mutation may represent both DNA repair and metabolic vulnerability that sensitize cells to radiation. Our findings also provide a strong rationale to study other potential combinations of metabolic pathway inhibition with radiation to further improve therapeutic effect under the above circumstances.

No disclosures were reported.

Z. Yang: Formal analysis, validation, investigation, methodology, writing–original draft, writing–review and editing. N. Hu: Formal analysis, validation, investigation, methodology, writing–original draft, writing–review and editing. W. Wang: Formal analysis, validation, investigation, methodology, writing–original draft, writing–review and editing. W. Hu: Validation, investigation, writing–original draft. S. Zhou: Formal analysis, investigation. J. Shi: Resources, data curation, validation. M. Li: Validation, investigation. Z. Jing: Formal analysis, validation, investigation. C. Chen: Resources, formal analysis, validation, investigation. X. Zhang: Resources, formal analysis, validation, investigation. R. Yang: Resources. X. Fu: Conceptualization, formal analysis, supervision, funding acquisition, methodology, writing–original draft, writing–review and editing. X. Wang: Conceptualization, formal analysis, supervision, funding acquisition, validation, writing–original draft, writing–review and editing.

This study is supported by Key Research Projects of Henan Higher Education Institutions: 14A320078 (X. Wang); Henan Natural Science Foundation 182300410379 (X. Fu); National Natural Science Foundation of China: 81874068 (X. Fu), 81972361 (X. Wang).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.