Cyclin F-Dependent Degradation of RBPJ Inhibits IDH1^R132H-Mediated Tumorigenesis Free

Cyclin F is the founding member of the F-box family of proteins and is a substrate recognition component of the SCF E3 ubiquitin ligase complex. It has been reported to mediate the ubiquitylation and proteasomal degradation of several proteins to maintain genomic stability (1). However, the functions and regulation of cyclin F under conditions of altered metabolic homeostasis remain unexplored.

Recombination signal binding protein for immunoglobulin kappa J region (RBPJ) is an effector of the Notch pathway (2). RBPJ has been linked to regulation of cellular metabolic processes. A previous study suggests that RBPJ regulates glucose uptake in memory T cells. Thus, RBPJ is critical for memory T-cell maintenance and secondary immune responses (3). RBPJ has also been reported to promote mTOR complex stability thereby augmenting lipogenesis in a Srebp1c (sterol regulatory element binding transcription factor 1c)-dependent manner (4). However, the role of RBPJ under metabolic stress conditions is yet unclear.

Isocitrate dehydrogenase 1 (IDH1) is a key enzyme that catalyzes the interconversion of isocitrate and 2-oxoglutarate (2-OG). IDH1 has been observed to be mutated in diverse cancers (5–8). Among the cancer-associated mutations, the most recurrent one is the missense mutation that results in a histidine residue instead of arginine (R132) at the active site. The mutant enzyme produces high levels of the metabolite 2-hydroxyglutarate (2-HG; ref. 9). Furthermore, it has been reported that 2-HG is a competitive inhibitor of 2-OG, and accumulation of 2-HG results in the inhibition of the dioxygenase family of enzymes including histone demethylases and DNA demethylases that require 2-OG as a cofactor (10). This results in epigenetic alterations that promote tumorigenesis. Thus, it is critical to delineate mechanisms underlying IDH1 regulation.

In this study, we examined the regulation and functions of cyclin F under metabolic stress conditions. We observed that metabolic stress–induced cyclin F brings about the ubiquitylation and proteasomal degradation of RBPJ. In addition, IDH1 was identified as a novel RBPJ transcriptional target. Thus, cyclin F inhibits mutant IDH1^R132H–promoted tumorigenesis in a RBPJ-dependent manner.

U251 and 293A cells were cultured in DMEM containing FBS (Invitrogen), 100 U/mL penicillin, and 100 μg/mL streptomycin at 37°C. BT142 cells were cultured as described previously (11). The cell lines were obtained from ATCC, authenticated, and checked for Mycoplasma contamination. Recombinant adenoviruses were amplified and titrated as reported previously (12). Cells were grown to approximately 50%–70% confluency and were infected with recombinant adenovirus at a multiplicity of infection (MOI) of 10–20 for the indicated time. GFP-expressing adenovirus (Ad-GFP) was used as a negative control. To induce glucose starvation, cells were grown to approximately 50% confluency and were then cultured in glucose-free DMEM (Invitrogen) containing dialyzed FBS (Invitrogen) and 0.5 mmol/L glucose (Sigma-Aldrich). Early passage MEFs were cultured and retroviral transduction was performed as described previously (13). Reagents including metformin and MG132 were purchased from Sigma-Aldrich. Palbociclib and MK2206 were purchased from Selleckchem. AGI5198 was purchased from MedChem Express. AGI5198 (1 μmol/L) was added to the culture medium 6 hours prior to the start of starvation time period.

Cells were lysed in lysis buffer [20 mmol/L Tris·HCl (pH 7.4), 5 mmol/L EDTA, 10 mmol/L Na₄P₂O₇, 100 mmol/L NaF, 2 mmol/L Na₃VO₄, 1% (v/v) Nonidet P-40, 1 mmol/L phenylmethylsufonylfluoride, 1 × Protease inhibitor mixture (Roche)]. Equal amounts of proteins per sample were subjected to SDS/PAGE and transferred to a polyvinylidene difluoride membrane (Millipore). The antibodies used are described in Supplementary Table S1. Imaging of all Western blots was performed using a UVP ChemiDoc-it imager equipped with VisionWorksLS software (v7.1; UVP).

Immunoprecipitations were performed using 500 μg cell extracts pretreated with MNase as described previously (14) and incubated with antibodies as indicated. Subsequent immunoblots were performed as described above.

Total RNA was extracted using TRIzol (Invitrogen). cDNA synthesis was performed using iScript cDNA synthesis kit (Bio-Rad) following the manufacturer's protocol. qPCR was performed using Maxima SYBR Green Mastermix (Fermentas) in an Eppendorf real-time PCR machine. 18S rRNA was used as internal control for all samples. ΔΔC_t method was used to analyze RT-qPCR data. Error bars are mean ± SD of three independent experiments with triplicate samples. Primer sequences are listed in Supplementary Table S2.

All animal protocols were approved by an Institutional Animal Care and Use Committee. Female nude (nu/nu) mice were subcutaneously injected in the right flank with 4 × 10⁶ cells/0.15 mL of PBS. The mice were randomly assigned to groups for injecting different cell lines. The tumor size was assessed for 4 weeks. No blinding was used. Tumor volume was determined by direct measurement with a caliper and was calculated using the formula: (widest diameter × smallest diameter²)/2. At the end of the experiment, the mice were euthanized, tumors were harvested and weighed, and metabolite and Western blot analyses were performed. For IHC analysis, tumors were harvested, fixed in formalin, and embedded in paraffin for sectioning. Five-micron–thick paraffin sections were dewaxed and immunostained with 1:100 Ki-67 antibody (Santa Cruz Biotechnology) or 1:100 CD 31 antibody (Santa Cruz Biotechnology). Secondary detection was performed using anti-mouse Alexa Fluor 488 (Molecular Probes) or anti-rabbit Alexa Fluor 555 antibody (Molecular Probes). Counterstaining was done with DAPI. Slides were imaged using a Zeiss ApoTome microscope, and images were analyzed with the Zeiss Zen Blue software. Quantitative analysis of immunofluorescence staining was performed using ImageJ software. The average signal intensity from four random fields was used for the analysis.

BT142^Luc2 cells were generated by stably transfecting (pooled neomycin-resistant population) BT142 cells with the pGL4.51[luc2/CMV/Neo] plasmid (Promega). BT142^Luc2 cells were stably transfected (pooled zeomycin-resistant population) with control (scrambled), cyclin F, RBPJ, or cyclin F along with RBPJ shRNA. BT142^Luc2 RBPJ knockdown cells were stably transfected (pooled G418-resistant population) with FLAG-tagged RBPJ^K315R. The cell suspension was injected orthotopically into the brain of female nude mice. The mice were randomly assigned to groups for injecting different cell lines. The tumor size was assessed for 4 weeks. No blinding was used. For weekly in vivo bioluminescence imaging, mice were anesthetized using ketamine (80 mg/kg) and xylazine (10 mg/kg) by intraperitoneal injection. Anesthetized animals were intraperitoneally administered 150 mg/kg d-luciferin in PBS. Imaging was performed in a Kodak FX Pro imaging system, and images were analyzed using Carestream imaging software. An X-ray image was taken followed by acquisition and overlay of the pseudocolor image representing the spatial distribution of detected photon counts emerging from active luciferase within the animal. Camera settings were kept constant during all measurements. At the end of the experiment, the mice were euthanized, the lungs and liver were harvested, and ex vivo imaging was performed.

Genomic DNA was isolated from blood obtained from euthanized mice for measurement of circulating tumor cells. qPCR was performed using primers specific to the human Alu repeat sequence (forward: 5′-ACGCCTGTAATCCCAGCACTT-3′; reverse: 5′-TCGCCCAGGCTGGAGTGC-3′), while mouse actin served as the control (forward: 5′-GCTTCTTTGCAGCTCCTTCGTTG-3′; reverse: 5′-TTTGCACATGCCGGAGCCGTTGT-3′).

All experiments were conducted independently at least three times. Results were expressed as mean ± SD. Sample size was chosen to give sufficient power for calling significance with standard statistical tests. Statistical analyses were performed by a standard two-tailed Student t test or one-way ANOVA. P < 0.05 was considered significant.

To investigate the function of cyclin F upon metabolic stress, we first examined its effect on cyclin F levels. Our results indicated that, unlike cyclins D, E, A, and B, cyclin F levels were induced upon metabolic stress (Fig. 1A; Supplementary Fig. S1A). We also observed a congruent increase in cyclin F transcript levels (Fig. 1B; Supplementary Fig. S1B). Similar results were obtained in other cell types (Supplementary Fig. S1C–S1F). We did not observe any change in the levels of cyclins D, E, A, B, and F in unstressed conditions (25 mmol/L glucose; Supplementary Fig. S1G). We further examined the half-life of cyclin F protein (Supplementary Fig. S1H). Our results indicate that there is no significant change in the half-life of cyclin F protein under metabolic stress conditions. We also observed that metformin treatment-induced metabolic stress results in cell-cycle arrest, as has been reported previously (15). However, induction of cell-cycle arrest alone by treatment with palbociclib (a selective inhibitor of the cyclin-dependent kinases CDK4 and CDK6) does not result in upregulation of cyclin F levels (Supplementary Fig. S1I and S1J). These results suggest that the upregulation in cyclin F levels is dependent on alteration in metabolic homeostasis. To delineate the mechanism of cyclin F induction upon metabolic stress, we analyzed its promoter region and identified a putative binding site for FOXO1, a key member of the FOX (Forkhead box) family of transcription factors involved in metabolic stress response (16, 17) (Fig. 1C). We next examined cyclin F transcript levels upon metabolic stress, in the presence or absence of FOXO1. We observed that in absence of FOXO1, there was no induction in cyclin F upon metabolic stress (Fig. 1D and E). Similar results were obtained when other shRNAs were used for FOXO1 depletion to exclude off-target effects of shRNA (Supplementary Fig. S2A and S2B). Furthermore, FOXO1 levels remain unaltered under unstressed and starvation conditions (Fig. 1E; Supplementary Fig. S2C). Moreover, cyclin F is specifically regulated by FOXO1 but not by other family members like FOXO3a, FOXO4, or FOXO6 (Supplementary Fig. S2D). We further examined the binding of FOXO1 to cyclin F promoter. Upon metabolic stress, increasing levels of FOXO1 were detected at cyclin F promoter (Fig. 1F). FOXO1 is phosphorylated by Akt at Ser256 and by Cdk1/2 at Ser249, which determine its cellular distribution and hence its activity (18–20). We observed that levels of phosphorylated FOXO1 (Ser256), which is cytoplasmic, declined upon metabolic stress concomitant to an increase in nuclear FOXO1 levels (Fig. 1G). However, no phosphorylation at Ser249 site was detected. Thus, our results suggest that Akt plays a key role in determining FOXO1 subcellular localization under metabolic stress conditions. To further corroborate the role of FOXO1 in cyclin F transcriptional regulation, we performed reporter assays. Our results indicated that cyclin F promoter was robustly transactivated over the time course of starvation period but abrogation of FOXO1 expression inhibited transactivation of cyclin F promoter (Supplementary Fig. S2E). Moreover, deletion of the FOXO1-binding site in the reporter assay abolished FOXO1-mediated transactivation. Furthermore, CRISPR-Cas9–mediated disruption of the FOXO1-binding site in the cyclin F promoter region abrogates FOXO1-mediated regulation of cyclin F expression (Supplementary Fig. S2F; Fig. 1H and I). Taken together, these results suggest that FOXO1 induces cyclin F upon metabolic stress.

To gain insights into its functions, we employed a biochemical approach for identifying cyclin F–interacting proteins (Fig. 2A). Many novel as well as known interactors of cyclin F were identified including CP110 and NUSAP1 (21, 22). Among the novel interactors, RBPJ was of particular interest as little is understood about its regulation. Thus, we further examined cyclin F–RBPJ interaction. We observed that RBPJ coimmunoprecipitated with cyclin F (Fig. 2B and C). To confirm whether cyclin F directly interacts with RBPJ, GST pull-down experiments were performed (Supplementary Fig. S3A). Our results indicated that cyclin F directly interacts with RBPJ. We further observed that RBPJ specifically interacts with cyclin F but not with other F-box proteins (Supplementary Fig. S3B). We then mapped the domain of RBPJ, to which cyclin F binds (Fig. 2D). We observed that cyclin F immunoprecipitated with FLAG-RBPJ segment (RBPJ-3D) encompassing the C-terminal RHR domain. Previous reports have suggested that cyclin F binds to the Cy motif (RxL/RxI) in its substrates (21). The C-terminal region of RBPJ has one Cy motif located between residues 304 and 306. Disruption of this motif led to loss of cyclin F–RBPJ interaction. Using FLAG-tagged cyclin F constructs we also mapped the domain of cyclin F that binds to RBPJ (Fig. 2E). Our data indicate that RBPJ binds to FLAG–cyclin F segment (cyclin F–2D) containing the cyclin domain. Earlier reports indicate that the hydrophobic patch in the cyclin domain is involved in substrate binding (23). We also observed that cyclin F mutated in this hydrophobic patch, cyclin F^M309A, cannot interact with RBPJ. Next, we examined the interaction between endogenous cyclin F and RBPJ. Our results suggested that although RBPJ coimmunoprecipitated with cyclin F, the interaction was attenuated over the time course of metabolic stress as RBPJ protein levels were downregulated (Fig. 2F and G; Supplementary Fig. S3C and S3D). Thus, our results indicate that cyclin F binds to RBPJ, but upon metabolic stress, the interaction was lost due to downregulation of RBPJ.

We further investigated the mechanism of RBPJ downregulation under metabolic stress conditions. Our results indicated that there was no significant change in RBPJ transcript levels upon metabolic stress (Fig. 3A). Because ubiquitin-mediated proteasome degradation pathway is frequently involved in regulation of protein levels, we determined the effects of the proteasome inhibitor, MG132, on RBPJ protein levels. We observed that MG132 treatment attenuated the decline in RBPJ protein levels upon metabolic stress while there was no significant change in its transcript levels (Fig. 3B). Moreover, no decline in interaction with cyclin F was observed upon metabolic stress in presence of MG132 (Supplementary Fig. S3E). We next examined RBPJ protein levels in presence or absence of cyclin F. We observed that cyclin F depletion abrogates RBPJ downregulation upon metabolic stress (Fig. 3C; Supplementary Fig. S4A). Furthermore, MG132 treatment inhibited the downregulation of RBPJ levels upon metabolic stress, while in the absence of cyclin F, MG132 treatment had no significant effect on RBPJ levels. Similar results were obtained in other cell types (Supplementary Fig. S4B and S4C). We also examined whether cyclin F has any effect on the half-life of RBPJ protein (Supplementary Fig. S4D). In the presence of cycloheximide, there was a significant cyclin F–dependent decline in RBPJ protein levels under starvation conditions with a distinct reduction in half-life. We also observed that unlike wild-type RBPJ, cyclin F noninteracting mutant RBPJ^RxI/AxA levels are not downregulated under metabolic stress conditions (Supplementary Fig. S4E). We next investigated whether cyclin F mediates RBPJ ubiquitylation. We observed that RBPJ was polyubiquitylated upon metabolic stress that was inhibited upon abrogation of cyclin F expression (Fig. 3D). Moreover, RBPJ polyubiquitylation was found to be K48-linked (Fig. 3E; Supplementary Fig. S4F). Similar results were obtained in other cell types (Supplementary Fig. S4G and S4H). Similar observations were made when other shRNAs were used for cyclin F depletion to exclude off-target effects of shRNA (Supplementary Fig. S4I and S4J). Furthermore, polyubiquitylation of RBPJ was abrogated in the presence of K48R ubiquitin mutant (Supplementary Fig. S4K).

To identify the site of polyubiquitylation, we performed GST pull-down assays under glucose starvation conditions. The GST-RBPJ fusion protein comprising the amino acids 301–500 was observed to be polyubiquitylated (Fig. 3F). We further mutated each of the three lysine residues within this region. Our results indicated that the mutation of K315 residue specifically abrogated K48-linked polyubiquitylation (Fig. 3G). Moreover, protein sequence analysis of RBPJ region encompassing amino acids 305–318 suggests that K315 site is conserved across diverse organisms (Fig. 3H). Furthermore unlike wild-type RBPJ, RBPJ^K315R levels are not downregulated under glucose starvation conditions (Supplementary Fig. S4L). These results indicate that metabolic stress–induced cyclin F promotes K48-linked polyubiquitylation of RBPJ at K315, resulting in its proteasomal degradation.

To determine the effect of RBPJ downregulation on cellular metabolic processes, we examined a list of genes involved in energy metabolism in cyclin F knockdown and cyclin F/RBPJ double knockdown cells by RT-qPCR. We observed that in U251 control cells, IDH1 expression was downregulated upon metabolic stress concomitant to decline in RBPJ levels. However, in U251 cyclin F knockdown cells, which constitutively express high levels of RBPJ, IDH1 levels were elevated and there was no significant change upon metabolic stress. Upon depletion of both cyclin F and RBPJ, IDH1 expression was constitutively downregulated (Fig. 4A and B; Supplementary Fig. S5A). Similar results were obtained in other cell types (Supplementary Fig. S5B–S5E). Similar observations were made when other sets of shRNAs were used for cyclin F and RBPJ depletion to exclude off-target effects of shRNA (Supplementary Fig. S5F and S5G). Furthermore, upon ectopic expression of ubiquitylation-resistant RBPJ mutant (RBPJ^K315R), there was no decline in IDH1 levels upon metabolic stress in contrast to the downregulation observed in case of wild-type RBPJ (Fig. 4C; Supplementary Fig. S5H). Similar results were obtained in BT142 cells (Supplementary Fig. S5I).

We next analyzed the IDH1 promoter region and identified a putative RBPJ binding site approximately 5.8 kb upstream of the transcription start site (Fig. 4D; ref. 24). We investigated the binding of RBPJ to IDH1 promoter. In U251 control cells, RBPJ bound to the IDH1 promoter but the levels detected at the promoter declined upon metabolic stress (Fig. 4E). However, upon abrogation of cyclin F expression, the decline in binding of RBPJ to IDH1 promoter under starvation conditions was not observed. Upon abrogation of expression of both cyclin F and RBPJ, no significant levels of RBPJ could be detected at the IDH1 promoter. Furthermore, in case of ubiquitylation-resistant RBPJ mutant (RBPJ^K315R), metabolic stress had no effect on its ability to bind IDH1 promoter. We further performed reporter assays. We observed that transactivation of IDH1 promoter was downregulated upon metabolic stress concomitant to decline in RBPJ levels. Moreover, abrogation of RBPJ expression as well as deletion of the RBPJ-binding site inhibited IDH1 promoter transactivation (Supplementary Fig. S5J). In addition, presence or absence of Notch1 did not have any effect on IDH1 levels (Supplementary Fig. S5K). These data suggest that IDH1 is a Notch-independent RBPJ transcriptional target.

We further analyzed the effects of Akt signaling on RBPJ and IDH1 levels under metabolic stress conditions (Supplementary Fig. S5L). Our results indicate that under metabolic stress conditions, Akt-dependent FOXO1 phosphorylation is downregulated, which promotes cyclin F induction, leading to RBPJ degradation and IDH1 downregulation. Thus, inhibition of Akt abrogates FOXO1 phosphorylation, resulting in constitutively elevated levels of cyclin F and reduced levels of RBPJ and IDH1. These data suggest that Akt signaling modulates cyclin F–RBPJ–IDH1 axis. Previous reports suggest that metabolic stress results in p53-dependent cell-cycle arrest (13). Hence, we further examined the effect of metabolic stress on the cell-cycle progression as well as levels of cyclin F, RBPJ, and IDH1 in U251 control and p53 knockdown cells (Supplementary Fig. S5M and S5N). As reported previously, we also observed that metabolic stress triggers cell-cycle arrest in presence of p53, but not in absence of p53. However, metabolic stress resulted in induction of cyclin F and consequent downregulation of RBPJ and IDH1 levels in both U251 control and p53 knockdown cells. Moreover, refeeding with glucose resulted in a reversal of these effects in both U251 control and p53 knockdown cells. Taken together, these results suggest that alteration in metabolic homeostasis determines cyclin F–mediated regulation of RBPJ and consequently IDH1 levels.

Because our results indicated that mutant IDH1 (IDH1^R132H) in BT142 cells is regulated by RBPJ in a cyclin F–dependent manner, we investigated the effect of cyclin F on IDH1^R132H activity. IDH1^R132H-generated 2-HG levels were downregulated upon metabolic stress, but upon abrogation of cyclin F expression, no decline in 2-HG levels was observed (Fig. 5A). Upon simultaneous abrogation of cyclin F and RBPJ expression, 2-HG levels were downregulated, irrespective of metabolic stress, as compared with control cells. On the other hand, upon ectopic expression of RBPJ^K315R, there was no downregulation of 2-HG levels upon metabolic stress, but treatment with AGI5198, an inhibitor of IDH1^R132H, resulted in decline of 2-HG levels irrespective of metabolic stress. 2-HG inhibits cellular dioxygenases including DNA and histone demethylases (10). Hence, we analyzed the effect of cyclin F on 5-hydroxymethyl cytosine (5hmC) levels. In BT142 control cells, 5hmC levels were upregulated upon metabolic stress. Upon abrogation of cyclin F expression, there was no significant change in 5hmC levels upon metabolic stress (Fig. 5B). On the other hand, upon simultaneous abrogation of cyclin F and RBPJ expression, 5hmC levels were elevated as compared with control cells. Moreover, there was no significant change in 5hmC levels upon metabolic stress in BT142 cyclin F/RBPJ double knockdown cells. In presence of RBPJ^K315R, metabolic stress did not trigger significant alterations in 5hmC levels, but AGI5198 treatment resulted in an increase in 5hmC levels. Similar observations were made in IDH1^R132H-expressing U251 cells (Supplementary Fig. S6A–S6C). Similar results were obtained when other sets of shRNAs were used for cyclin F and RBPJ depletion to exclude off-target effects of shRNA (Supplementary Fig. S6D and S6E). We next investigated the effects of cyclin F on the methylation status of H3K4, H3K9, and H3K27. In BT142 control cells, metabolic stress led to a decline in H3K9me3 and H3K27me3 levels, concomitant to RBPJ and IDH1 downregulation. But similar decrease was not observed in BT142 cyclin F knockdown cells (Fig. 5C). In BT142 cyclin F/RBPJ double knockdown cells, the H3K9me3 and H3K27me3 levels were lower as compared with the control cells, and there was no discernable change upon metabolic stress. Upon ectopic expression of RBPJ^K315R, H3K9me3 and H3K27me3 levels were not significantly altered upon metabolic stress. But when ectopic expression of mutant RBPJ (RBPJ^K315R) was combined with AGI5198 treatment, it resulted in a decline in H3K9me3 and H3K27me3 levels, as compared with the control cells. H3K4me3 levels remained unaltered in these conditions. These results indicate that cyclin F-dependent RBPJ degradation inhibits IDH1^R132H-mediated epigenetic reprogramming.

To explore the physiological significance of cyclin F–mediated regulation of IDH1^R132H epigenetic program, we analyzed the transcript levels of key tumor suppressor genes (TSG) reported to be epigenetically silenced in gliomas (25–28). Our data indicate that the levels of these TSGs are upregulated upon metabolic stress, which is inhibited upon abrogation of cyclin F expression (Fig. 5D). In BT142 cyclin F/RBPJ double knockdown cells, the expression of the TSGs was elevated as compared with control cells, and there was no significant alteration in the levels upon metabolic stress. Ectopic expression of RBPJ^K315R inhibited the upregulation of the TSGs upon metabolic stress while simultaneous treatment with AGI5198 abrogated the inhibition. We further analyzed the promoter regions of these genes by MeDIP assays (Supplementary Fig. S6F). Our data indicate that in BT142 control cells, metabolic stress–induced cyclin F promotes demethylation of the CpG islands in the promoters of the TSGs and depletion of cyclin F resulted in a loss of this effect. However, in BT142 cyclin F/RBPJ double knockdown cells, the extent of promoter methylation was lower as compared with control cells, and there was no significant alteration upon metabolic stress. In the presence of RBPJ^K315R, promoter demethylation upon metabolic stress was attenuated, whereas simultaneous treatment with AGI5198 resulted in reduced promoter methylation, with no discernable change under metabolic stress conditions. These results indicate that cyclin F–dependent RBPJ degradation inhibits IDH1^R132H-mediated epigenetic silencing of TSGs.

We next investigated the effect of cyclin F on the malignant phenotype of IDH1^R132H-expressing cells. Our results indicated that metabolic stress downregulated the invasiveness, migration potential, and anchorage-independent growth (Fig. 6A–C). Abrogation of cyclin F expression enhanced the invasiveness, migration potential, and anchorage-independent growth, with no significant change upon metabolic stress. However, the invasiveness, migration potential, and anchorage-independent growth of BT142 cyclin F/RBPJ double knockdown cells was lower as compared with control cells, and there was no significant alteration upon metabolic stress. In presence of RBPJ^RxI/AxA and RBPJ^K315R, invasiveness, migration potential and anchorage-independent growth was elevated with no significant change under metabolic stress conditions. However, treatment with AGI5198 led to a reduction in the invasiveness, migration potential, and anchorage-independent growth of RBPJ^K315R-expressing cells with no significant change under metabolic stress conditions. Similar results were obtained in U251 cells expressing IDH1^R132H (Supplementary Fig. S6G–S6I). Similar observations were made when other sets of shRNAs were used for cyclin F and RBPJ depletion to exclude off-target effects of shRNA (Supplementary Fig. S6J and S6K). These results indicate that cyclin F–mediated RBPJ degradation inhibits IDH1^R132H-dependent malignant transformation.

We next examined the effect of cyclin F on tumorigenicity of IDH1^R132H-expressing cells. Under metabolic stress conditions, cyclin F depletion resulted in significantly larger tumors as compared with control cells (Fig. 6D and E; Supplementary Fig. S7A). However codepletion of RBPJ resulted in reduced tumor growth. Presence of RBPJ^K315R promoted tumor growth while simultaneous administration of AGI5198 attenuated the tumor growth. To further investigate the differences in tumor phenotype, we checked the 2-HG levels in these tumors. Cyclin F knockdown resulted in elevated 2-HG levels while concomitant knockdown of RBPJ lowered 2-HG levels (Supplementary Fig. S7B, left). In presence of RBPJ^K315R, 2-HG levels were elevated while simultaneous AGI5198 treatment reduced 2-HG levels. Concurring observations were made when 5hmC levels were examined (Supplementary Fig. S7B, right). Furthermore, increased Ki67 and CD31 staining, which are indicative of proliferation rate and angiogenesis, respectively, was observed upon cyclin F knockdown while simultaneous knockdown of RBPJ mitigated these effects. In presence of RBPJ^K315R, increased proliferation and angiogenesis was observed while simultaneous AGI5198 administration resulted in decreased proliferation and angiogenesis (Supplementary Fig. S7C and S7D). These results suggest that cyclin F suppresses IDH1^R132H-mediated tumorigenesis.

We further investigated the tumor-suppressive effects of cyclin F on the metastatic potential of IDH1^R132H-expressing cells. We observed that under metabolic stress conditions, BT142 cyclin F knockdown cells formed much larger tumors as compared with control cells (Fig. 7A and B). Depletion of cyclin F also resulted in numerous metastatic nodules in the liver as well as lungs (Fig. 7C and D). Similar observations were made with cells expressing RBPJ^K315R. However, treatment with AGI5198 resulted in reduced tumor size and metastasis. Moreover, concomitant depletion of both cyclin F and RBPJ also resulted in reduced tumor size and metastasis. We also observed higher number of circulating tumor cells (CTC) in mice implanted with cells depleted for cyclin F or cells expressing RBPJ^K315R (Fig. 7E). However, AGI5198 administration led to lowered CTC count. Moreover, the levels of the epithelial marker E–cadherin were lower, while those of the mesenchymal markers vimentin and fibronectin were elevated in tumors arising from cells depleted for cyclin F, or cells expressing RBPJ^K315R, which was reversed upon AGI5198 treatment (Supplementary Fig. S7E). Thus, we concluded that cyclin F suppresses IDH1^R132H-mediated tumor growth and metastasis in a RBPJ-dependent manner.

Several previous studies have reported increased levels of IDH1^R132H in gliomas (29). Thus, we next examined whether there is any association between cyclin F and IDH1^R132H levels in different grades of gliomas. We observed that cyclin F levels were lower in gliomas as compared with normal brain tissue while RBPJ and IDH1^R132H levels were elevated (Fig. 7F). Furthermore, with increase in tumor grades, cyclin F levels declined while RBPJ and IDH1^R132H levels increased (Supplementary Fig. S7F–S7H). These results indicate that dysregulation of cyclin F expression and consequent upregulation of IDH1^R132H is critical for glioma progression.

Cyclin F by virtue of its role in maintaining genomic stability is critical for normal cell physiology as well as transformation. However, little is known about cyclin F regulation. A previous study suggests that hypoxia-induced miR-210 downregulates several mitosis-related genes including cyclin F by directly binding to the 3′-UTR. Thus, miR-210 perturbs mitotic progression by inducing chromosome misalignment and mis-segregation (30). Oct4 has been reported to transactivate cyclin F and NIPP1 (nuclear inhibitor of protein phosphatase 1; ref. 31). The Oct4-cyclin F/Nipp1 pathway inhibits the protein phosphatase 1 complex–mediated dephosphorylation of Rb to promote embryonic stem cell self-renewal (32). Here we report that under metabolic stress conditions cyclin F levels are induced. Our results further indicate that FOXO1 binds to cyclin F promoter and triggers its expression. Thus, our findings provide mechanistic insights into regulation of cyclin F expression.

Despite being a key transcription factor regulating diverse physiologic processes, the regulation of RBPJ remains poorly understood. A recent report suggests that miR-133a-3p downregulates RBPJ expression to inhibit the tumorigenic properties (33). RBPJ has also been reported to be induced under hypoxic conditions (34). However, the mechanism of RBPJ induction under hypoxic conditions remains unclear. Furthermore, emerging reports also indicate an important role of RBPJ in controlling cellular metabolism (3, 4). Nonetheless, the regulation of RBPJ in response to metabolic cues remains unexplored. Our findings establish the role of cyclin F in the posttranslational regulation of RBPJ under metabolic stress conditions. Cyclin F mediates RBPJ ubiquitylation at the conserved K315 residue, leading to its proteasomal degradation. Thus, our findings provide a new dimension to the repertoire of cyclin F substrates.

IDH1 is a frequently mutated metabolic enzyme in diverse cancer types (35). However, molecular mechanisms underlying regulation of IDH1 expression are yet unclear. In pancreatic cancer cells, HuR, a RNA-binding protein binds to and stabilizes IDH1 transcript. Elevated IDH1 levels promote chemoresistance in these cells by augmenting antioxidant defense mechanism (36). A recent report suggests that exogenously expressed FOXO3A transactivates IDH1. FOXO3A depletion reduces 2-HG levels and downregulates proliferation capacity (37). Our results suggest that there is no FOXO1-dependent alteration in IDH1 levels under normal physiologic conditions (Supplementary Fig. S8A and S8B). Under metabolic stress conditions, FOXO1 induces cyclin F, which in turn mediates RBPJ degradation, leading to IDH1 downregulation. Consequently, depletion of FOXO1 results in elevated levels of RBPJ and IDH1 under metabolic stress conditions, while codepletion of FOXO1 and RBPJ leads to IDH1 downregulation. Thus, our data indicate that IDH1 is a direct RBPJ transcriptional target. In cancer cells expressing the oncogenic mutant IDH1^R132H, cyclin F–mediated RBPJ degradation under low-nutrient conditions results in downregulation of IDH1^R132H levels. Decline in IDH1^R132H levels diminishes the malignant phenotype. Thus, our findings delineate a mechanism for regulation of IDH1^R132H expression.

No potential conflicts of interest were disclosed.

Conception and design: R.S. Deshmukh, S. Das

Development of methodology: R.S. Deshmukh, S. Das

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): R.S. Deshmukh, S. Sharma, S. Das

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): R.S. Deshmukh, S. Sharma, S. Das

Writing, review, and/or revision of the manuscript: R.S. Deshmukh, S. Sharma, S. Das

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R.S. Deshmukh, S. Das

Study supervision: S. Das

We thank the members of the Molecular Oncology Lab for helpful discussions. The authors would also like to acknowledge the support from National Institute of Immunology. R.S. Deshmukh and S. Sharma were supported by a fellowship from the Department of Biotechnology, Ministry of Science and Technology, Government of India. This work was supported by a grant from the Department of Biotechnology, Government of India [BT/HRD/NBA/37/01/2015 (i) to S. Das].

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.