Fumarate Mediates a Chronic Proliferative Signal in Fumarate Hydratase-Inactivated Cancer Cells by Increasing Transcription and Translation of Ferritin Genes

ABSTRACT

Germ line mutations of the gene encoding the tricarboxylic acid (TCA) cycle enzyme fumarate hydratase (FH) cause a hereditary cancer syndrome known as hereditary leiomyomatosis and renal cell cancer (HLRCC). HLRCC-associated tumors harbor biallelic FH inactivation that results in the accumulation of the TCA cycle metabolite fumarate. Although it is known that fumarate accumulation can alter cellular signaling, if and how fumarate confers a growth advantage remain unclear. Here we show that fumarate accumulation confers a chronic proliferative signal by disrupting cellular iron signaling. Specifically, fumarate covalently modifies cysteine residues on iron regulatory protein 2 (IRP2), rendering it unable to repress ferritin mRNA translation. Simultaneously, fumarate increases ferritin gene transcription by activating the NRF2 (nuclear factor [erythroid-derived 2]-like 2) transcription factor. In turn, increased ferritin protein levels promote the expression of the promitotic transcription factor FOXM1 (Forkhead box protein M1). Consistently, clinical HLRCC tissues showed increased expression levels of both FOXM1 and its proliferation-associated target genes. This finding demonstrates how FH inactivation can endow cells with a growth advantage.

INTRODUCTION

Hereditary leiomyomatosis and renal cell cancer (HLRCC) patients carry a germ line-inactivating mutation in one of the fumarate hydratase (FH) alleles and are prone to developing skin leiomyomas, uterine fibroids, and renal cell carcinoma of type 2 papillary morphology (1). HLRCC-associated tumors harbor a loss of heterozygosity at the FH locus, indicating biallelic FH inactivation as the tumor-initiating event (1). However, it remains unclear how FH inactivation drives carcinogenesis.

The most direct consequence of FH inactivation is intracellular fumarate accumulation. Accumulated fumarate can covalently modify cysteine residues of proteins in an uncatalyzed process termed succination and cause many alterations in cellular signaling (2). Succination in HLRCC cells was first discovered on Kelch-like ECH-associated protein 1 (KEAP1), a negative regulator of the nuclear factor (erythroid-derived 2)-like 2 (NRF2) transcription factor (3). Since HLRCC is driven by FH inactivation, chronic succination of KEAP1 results in constitutive NRF2 activation and increased expression of its target genes (3). Besides KEAP1, the Krebs cycle enzyme aconitase 2 (Aco2) was reported to be a succination target in Fh knockout mouse tissues, and succination inhibited its activity (2). Despite dramatic cellular changes induced by protein succination, it remains unclear whether succination contributes to FH inactivation-driven carcinogenesis.

The expansion of FH-inactivated cells into a tumor mass indicates that FH inactivation somehow endows the cells with a chronic proliferative signal, which is a fundamental hallmark of cancer (4). While the mechanisms by which cancer cells with oncogenic mutations in growth signaling genes acquire such a signal are clear, how the loss of a tricarboxylic acid (TCA) cycle enzyme induces proliferative signaling is enigmatic. Analogous to chronic exposure to an electrophilic carcinogen, FH-inactivated cells are chronically exposed to high levels of intracellular fumarate. Thus, understanding the mechanisms by which FH inactivation contributes to carcinogenesis has profound implications in cancer biology.

RESULTS

Fumarate accumulation confers a chronic proliferative signal in HLRCC cells.

To determine whether fumarate accumulation confers a chronic proliferative signal, we created a pair of isogenic FH-reconstituted (FH^res) and control (FH^−/−) HLRCC cell lines by stably transducing the well-characterized HLRCC cell line UOK262 (5) with functional Flag-tagged FH (UOK262-FH^res) or with the corresponding empty vector control (UOK262-FH^−/−). Since increased NRF2 protein levels in HLRCC cells are a direct consequence of KEAP1 succination, the NRF2 protein level and the upregulation of its target genes serve as good indicators of cellular protein succination in these cells (3). Expectedly, FH reconstitution resulted in decreased protein levels of NRF2 and of the NRF2 target gene AKR1B10, indicating that it drove a reduction in the intracellular fumarate concentration and therefore decreased protein succination (Fig. 1A). Using this isogenic pair of HLRCC cell lines in a cell proliferation assay, we found that FH-reconstituted UOK262-FH^res cells had a significantly lower proliferation rate than did control UOK262-FH^−/− cells (Fig. 1B). This reduced growth rate can be abrogated by the addition of a membrane-permeable form of fumarate, monomethyl fumarate (MMF) (Fig. 1C), supporting that fumarate promotes cell proliferation in FH-inactivated cells. Thus, we sought to determine the mechanism by which FH inactivation promotes cell proliferation.

FIG 1.

Fumarate accumulation confers chronic proliferation. (A) Western blots of NRF2 and its transcription target AKR1B10 after reintroduction of FH into the FH^−/− HLRCC tumor cell line UOK262. β-Actin (ACTB) was used as a loading control. (B) Cell viability measured by formazan production in UOK262 cells (FH^−/−) after reintroduction of FH (UOK262-FH^res). Data are presented as means ± standard deviations of results from a representative experiment. Curves were statistically significantly different, as determined by two-way analysis of variance, for the FH genotype (P < 0.001) but not for time (P > 0.05). (C) Cell viability measured by formazan production in UOK262 cells (FH^−/−) after reintroduction of FH (UOK262-FH^res) and 80 μM MMF treatment. Data are presented as means ± standard deviations of results from a representative experiment. Curves were statistically significantly different, as determined by two-way analysis of variance, for MMF treatment (P < 0.001) but not for time (P > 0.05).

Fumarate accumulation increases intracellular ferritin levels.

We focused our investigation on protein succination, as it is one of the most prominent cellular changes induced by FH inactivation (6). To date, the succination of two proteins in the context of fumarate accumulation has been described in detail. These two proteins are KEAP1, which was identified in HLRCC, and Aco2, which was identified in Fh knockout mouse tissue. Succination of KEAP1 resulted in NRF2 activation, while succination of Aco2 inhibited its aconitase activity (2, 3). Aco2 belongs to a family of proteins that includes iron regulatory protein 1 (IRP1) (also known as aconitase 1 [ACO1]) and IRP2 (also known as ACO3) (7). Both IRP1 and IRP2 (collectively denoted IRPs) play central roles in cellular iron signaling, with IRP2 exerting a dominant effect (8). Interestingly, both NRF2 and IRPs interplay to regulate the expression of the ferritin light chain (FTL) and the ferritin heavy chain (FTH1) genes (9). These genes encode different subunits of ferritin, which is a cellular iron storage protein recently shown to promote cancer cell proliferation (10). Specifically, NRF2 promotes FTL and FTH1 transcription, while IRP2 represses the translation of the resulting transcripts by binding to a hairpin structure located in the 5′ untranslated region (UTR) known as an iron response element (IRE) (Fig. 2A). Thus, we hypothesized that fumarate accumulation in FH-inactivated cells drives a chronic growth signal by increasing ferritin expression in a concerted manner: it increases ferritin gene transcription through NRF2 activation and increases ferritin mRNA translation by inhibiting IRP2 (Fig. 2A).

FIG 2.

Fumarate accumulation increases ferritin levels. (A) Ferritin transcriptional and translational regulation. NRF2 can transcribe the FTL and FTH1 genes. IRP2 can bind FTL and FTH1 mRNA iron response element hairpins in the 5′ untranslated regions to repress translation. (B) Western blots of HK2 cells show relative increases in FTL and FTH1 protein levels upon siRNA-mediated FH knockdown. ACTB was used as a loading control. (C) Western blots of HEK293 cells show relative increases in FTL and FTH1 protein levels upon CRISPR-Cas9-mediated FH knockout. Increased NRF2 levels indicate a stable knockout. ACTB was used as a loading control. (D) Western blots of FH-rescued (FH^res) and control (FH^−/−) UOK262 cells show increased FTL and FTH1 levels under conditions of nonfunctional FH. ACTB was used as a loading control. (E) Addition of 80 μM MMF or 40 μM DMF to HEK293 cells increases FTL and FTH1 protein levels. ACTB was used as a loading control.

To determine the effects of FH inactivation on the intracellular ferritin level, we transiently knocked down FH in an immortalized human kidney epithelial cell line, HK2. Knockdown resulted in increased intracellular FTL and FTH1 protein levels (Fig. 2B), indicating that FH inactivation increases the cellular ferritin level. Since the effects of transient small interfering RNA (siRNA)-mediated knockdown could be different from those of biallelic FH inactivation, we further tested the observed effects by creating FH knockout HEK293 cells using clustered regularly interspaced short palindromic repeat (CRISPR) gene-editing technology. The knockout was confirmed by Sanger sequencing (see Fig. S1A and S1B in the supplemental material) and immunoblotting (Fig. S1C). Consistently, the knockout also resulted in higher FTL and FTH1 protein levels (Fig. 2C). Moreover, the reconstitution of functional FH in UOK262 HLRCC cells decreased FTL and FTH1 protein levels (Fig. 2D), supporting that FH inactivation increases the cellular ferritin level. Since the most direct consequence of FH inactivation is intracellular fumarate accumulation, we proceeded to simulate fumarate accumulation in HEK293 cells by treating them with two different membrane-permeable forms of fumarate: MMF and dimethyl fumarate (DMF). Treatments with either DMF or MMF led to increased FTL and FTH1 protein levels (Fig. 2E), supporting that fumarate accumulation increases the intracellular ferritin level.

IRP2 is a succination target.

To date, the only aconitase family member reported to be a succination target has been mouse Aco2 (2). Multiple-protein-sequence alignment of mouse Aco2 with human IRP1 and IRP2 revealed that the succination sites identified in mouse Aco2 are conserved in both human IRP1 and IRP2, suggesting that these proteins may be succination targets as well (Fig. 3A). We focused our investigation on IRP2 because it is the functionally dominant member (8). Using tandem mass spectrometry (MS/MS) analyses, we identified multiple succinated cysteine residues on ectopically expressed Flag-tagged IRP2 immunoprecipitated from DMF-treated HEK293 cells, supporting human IRP2 as a succination target (Fig. 3B; see also Table S1 in the supplemental material). However, peptides derived from the regions of IRP2 spanning residues C512 to C516 and C578 to C581 were not detected in this analysis, as the peptides generated by trypsin digestion were too large to be identified by using our proteomics platform. To more closely interrogate these regions, we generated an IRP2-A523R mutant, which introduced an additional tryptic digestion site. Tandem mass spectrometry analyses of IRP2-A523R revealed that C512 and C581 are succinated. These residues are conserved between human IRP2 and mouse Aco2 (Fig. 3C and Table S2). C512 was previously shown to be critical in maintaining the translational repression function of IRP2 (11). Thus, succination of IRP2 may partly contribute to the increased ferritin protein levels in FH-inactivated cells.

FIG 3.

IRP2 is a succination target. (A) Multiple-sequence alignment of mouse Aco2, human IRP2 (ACO3), and IRP1 (ACO1). Succination sites reported for Fh^−/− mouse Aco2 are conserved in human IRP1 and IRP2. These sites are highlighted in red. (B) Adduct sites on Flag-tagged wild-type IRP2, as determined by LC-MS/MS. Black bars represent peptides detected by mass spectrometry analysis. (C) Adduct sites on Flag-tagged IRP2-A523R, as determined by LC-MS/MS. Black bars represent peptides detected by mass spectrometry analysis.

Fumarate inhibits IRP-mediated translational repression.

To quantify the effects of fumarate on IRP-mediated FTL and FTH1 translational repression, we developed constructs that allow the expression of a chimeric gene consisting of a firefly luciferase open reading frame carrying either the 5′ UTR of FTL (pECE-FTL-IRE-LUX) or the 5′ UTR of FTH1 (pECE-FTH1-IRE-LUX) (see Fig. S2 in the supplemental material). The transcription of this chimeric gene is controlled by a simian virus 40 (SV40) promoter, which allows constitutive transcription in transfected cells. Together with these vectors, we also created a transfection control vector, pECE-RL (Fig. S2). pECE-RL has the Renilla luciferase gene controlled by the same SV40 promoter as pECE-FTL-IRE-LUX and pECE-FTH1-IRE-LUX. The 5′ UTRs of the FTL and the FTH1 mRNAs each contain an IRE. By fusing them to the firefly luciferase gene, we enabled IRPs to control the translation of the firefly luciferase gene, allowing the quantification of IRP-IRE binding-mediated translational repression by luciferase reporter assays. Under conditions of high intracellular iron levels (supplemented with 200 μM iron citrate [Fe]), IRPs do not bind to IREs, resulting in higher firefly luciferase activity in HEK293 cells transfected with the constructs (Fig. 4A and B). Inversely, in the presence of an iron chelator (200 μM deferoxamine [DFO]), IRP represses the translation of firefly luciferase, resulting in lower activity (Fig. 4A and B). Expectedly, Western blot analysis showed that iron and DFO treatments increased FTL and FTH1 levels and decreased IRP2 levels accordingly (Fig. 4C). The cellular IRP2 level is regulated by the cellular iron sensor F-box and leucine-rich repeat protein 5 (FBXL5), which is stabilized under conditions of high iron levels (12, 13). FBXL5 mediates IRP2 ubiquitylation, marking it for proteasomal degradation. Hence, iron treatment increases the FBXL5 protein level, while DFO treatment decreases it (Fig. 4C). Using the luciferase reporter constructs, we found that DMF or MMF treatments, which simulate intracellular fumarate accumulation, can decrease the ferritin translational-repression activity of IRPs (Fig. 4D to G). Accordingly, Western blot analysis of DMF- and MMF-treated samples also showed increases in ferritin levels in the DMF- and MMF-treated groups compared to the vehicle control groups (Fig. 4H and I). DMF and MMF treatments also led to increased FBXL5 levels (Fig. 4H and I); however, treatment also resulted in concurrently increased IRP2, FTL, and FTH1 protein levels. Since in canonical iron signaling, IRP2 represses the translation of FTL and FTH1, the concurrent increases in FBXL5, FTL, FTH1, and IRP2 levels upon DMF and MMF treatments indicate that fumarate accumulation disrupts normal iron signaling. We also evaluated levels of iron signaling proteins in FH-inactivated cells. Consistently, UOK262-FH^−/− cells had higher protein levels of FBXL5, FTL, FTH1, and IRP2 than did UOK262-FH^res cells (Fig. 4J), indicating that FH inactivation can alter iron signaling.

FIG 4.

Fumarate inhibits IRP-mediated translational repression. (A and B) IRE reporter assays using the Dual-Glo luciferase assay (Promega) indicate appropriate responsiveness to iron levels for both pECE-FTL-IRE-LUX (A) and pECE-FTH1-IRE-LUX (B) cotransfected with pECE-RL into HEK293 cells. Cells were treated with ferric citrate (Fe; 200 micromolar) or an iron chelator (DFO; 200 micromolar) to modulate iron levels. Data are presented as means ± standard deviations. * indicates a P value of <0.05 relative to vehicle (Veh) treatment, as determined by Student's t test. RLU, relative light units. (C) Immunoblots of HEK293 cells treated with DFO show increased IRP2 and decreased FTL, FTH1, and FBXL5 protein levels. Treatment with Fe resulted in opposite effects. ACTB was used as a loading control. (D to G) IRE reporter assays using the Dual-Glo luciferase assay (Promega) with dimethyl fumarate and monomethyl fumarate treatments. HEK293 cells were cotransfected with pECE-RL and either pECE-FTL-IRE-LUX (D and F) or pECE-FTH1-IRE-LUX (E and G). Cells were treated with increasing concentrations of either dimethyl fumarate (D and E) or monomethyl fumarate (F and G). Data are presented as means ± standard deviations. * indicates a P value of <0.05 relative to vehicle treatment, as determined by Student's t test. (H and I) Western blots of HEK293 cells treated with increasing concentrations of DMF (H) or MMF (I) show dysregulated iron signaling, as indicated by simultaneously high protein levels of IRP2, FBXL5, FTL, and FTH1. ACTB was used as a loading control. (J) Western blots of UOK262-FH^−/− control cells and UOK262-FH^res cells. FH reconstitution leads to decreased NRF2, IRP2, FBXL5, FTH1, and FTL protein levels. ACTB was used as a loading control.

In canonical iron signaling, IRP2 is marked for ubiquitylation and subsequent proteasomal degradation by FBXL5. The paradoxical concurrent increases in FBXL5 and IRP2 levels following fumarate treatment or FH inactivation may be due to an impaired ubiquitylation of IRP2 by FBXL5, as evidenced by the decreased IRP2 ubiquitylation following MMF treatment (Fig. S3).

Sustained NRF2 activation partly contributes to increased intracellular ferritin levels.

Both FH inactivation and fumarate accumulation activate NRF2, and both FTL and FTH1 are transcription targets of NRF2. Thus, NRF2 activation may partly contribute to the observed increased ferritin gene expression levels in FH-inactivated cells by increasing their transcription. Quantitative PCR (qPCR) analyses revealed that the transcript levels of both FTL and FTH1 were higher in UOK262-FH^−/− cells than in UOK262-FH^res cells, suggesting that FH inactivation promotes the transcription of the FTL and FTH1 genes (Fig. 5A). To determine the contribution of NRF2 to promoting the transcription of the FTL and FTH1 genes in the context of fumarate accumulation, we generated NRF2 knockout HEK293 cells (HEK293-NRF2^−/−) using CRISPR gene-editing technology. The knockout was confirmed by Sanger sequencing and immunoblotting (see Fig. S4 in the supplemental material). Using this cell line, we found that DMF treatment increased FTL and FTH1 transcript levels in HEK293-NRF2^+/+ but not in HEK293-NRF2^−/− cells, indicating that NRF2 mediates fumarate-induced FTL and FTH1 transcription (Fig. 5B). Moreover, HEK293 cells transfected with Myc-tagged NRF2 showed increased FTH1 and FTL protein levels compared to those in cells transfected with the empty vector, indicating that NRF2 plays an important role in modulating the expression of these genes (Fig. 5C). In Myc-NRF2-transfected cells, MMF treatment further increased FTL protein levels without further activating NRF2 (Fig. 5C). This effect may be attributed to the inhibition of IRP2 activity.

FIG 5.

NRF2 activation by fumarate contributes to ferritin accumulation. (A) Quantitative reverse transcription-PCR shows that FH reconstitution in UOK262 (UOK262-FH^res) cells decreases relative FTL and FTH1 transcript levels. Changes in expression levels were normalized to values for UOK262-FH^res cells. GCLM is an NRF2 target gene known to be upregulated in HLRCC. Error bars represent standard deviations. * indicates a P value of <0.05 as determined by Student's t test. (B) Quantitative reverse transcription-PCR of HEK293-NRF2^−/− and HEK293-NRF2^+/+ cells treated with 40 μM DMF or the vehicle (Veh). Changes in expression levels were normalized to values for vehicle treatment. Error bars represent standard deviations. * indicates a P value of <0.05 as determined by Student's t test. (C) Western blot of HEK293 cells transfected with pIRESpuro3 (vector) or NRF2-Myc and treated with the vehicle or 80 μM MMF. ACTB was used as a loading control. (D) Western blot of HEK293-NRF2^−/− and HEK293-NRF2^+/+ cells treated with the vehicle, 80 μM MMF, or 40 μM DMF. ACTB was used as a loading control. (E) Western blot of HEK293-NRF2^−/− cells treated with increasing concentrations of DMF, 200 μM DFO, or 200 μM Fe.

To better understand how fumarate-mediated IRP2 inhibition contributes to increased ferritin protein levels independently of NRF2, we treated both HEK293-NRF2^+/+ and HEK293-NRF2^−/− cells with MMF and DMF. As expected, MMF and DMF treatments lead to increased FTL and FTH1 protein levels in HEK293-NRF2^+/+ cells (Fig. 5D). However, in HEK293-NRF2^−/− cells, MMF and DMF treatments increased only the FTL protein level (Fig. 5D). As anticipated, increasing concentrations of DMF in HEK293-NRF2^−/− cells increased FTL but not FTH1 protein levels in a dose-responsive manner (Fig. 5E). These results indicate that in the absence of NRF2, fumarate can still mediate an increase in the FTL protein level; this effect is likely mediated by IRP2 inhibition and demonstrates that both increased transcription and translation cooperate to increase the FTL protein level, while the FTH1 protein level is more dependent on NRF2-mediated transcription.

UOK262 cells cannot be transfected efficiently by using routine DNA transfection strategies. Furthermore, UOK262-FH^−/− and UOK262-FH^res cells exhibit large differences in transduction and transfection efficiencies, making the direct evaluation of FTL and FTH1 translation using the dual-luciferase method impossible. Therefore, we quantified the increased FTH1 and FTL protein levels in UOK262-FH^−/− by immunoblot densitometry analysis and normalized the values by their relative transcript levels (Fig. S5). Consistently, this analysis revealed that increases in FTL protein levels could not be accounted for by increased transcription alone, implying that both increased transcription and increased translation contributed to the fumarate-mediated increased FTL protein level, while the increase in the FTH1 protein level could be accounted for by increased transcription alone.

FH inactivation activates the FOXM1 transcription factor.

Ferritin confers a chronic growth signal by activating FOXM1 (Forkhead box protein M1) signaling (10). FOXM1 controls the expression of genes regulating progression through the G₂/M phases of the cell cycle, and overexpression of FOXM1 is associated with cancer progression (14). Hence, we evaluated the activities of FOXM1 signaling in the context of FH inactivation and fumarate accumulation. Immunoblot analyses showed that UOK262-FH^−/− cells have a higher level of FOXM1 than do UOK262-FH^res cells, suggesting that FH inactivation activates FOXM1 signaling (Fig. 6A). qPCR analyses revealed increases in the levels of FOXM1 transcripts as well as the transcripts of the FOXM1 target genes AURKA, AURKB, and CDK1 in UOK262-FH^−/− cells compared to UOK262-FH^res cells, supporting that FOXM1 is activated in FH-inactivated cells (Fig. 6B). To determine if the observed activation of FOXM1 signaling was a result of fumarate accumulation, we simulated fumarate accumulation in UOK262-FH^res cells by MMF treatment. Immunoblot analyses revealed that MMF-treated cells had higher levels of FOXM1 (Fig. 6C). To ascertain the role of ferritin in mediating the observed activation of FOXM1 signaling, we overexpressed FTL and FTH1 by cotransducing UOK262-FH^res and HEK293 cells with lentiviruses carrying these genes. When ferritin was overexpressed, both cell lines showed increases in FOXM1 protein levels, indicating that ferritin increases the FOXM1 protein level (Fig. 6D). Consistently, sequential FTL and FTH1 knockdowns in UOK262 cells using short hairpin RNAs (shRNAs) also resulted in decreased FOXM1 levels, supporting that ferritin activates FOXM1 signaling in FH-inactivated cells (Fig. 6E). To evaluate if ferritin contributes to a progrowth phenotype in UOK262-FH^−/− cells, we assessed proliferation following FTL or FTH1 knockdown. The knockdown of either FTL or FTH1 decreased the proliferation rate in UOK262 cells (Fig. 6F), indicating that ferritin can mediate growth in UOK262 cells.

FIG 6.

FH inactivation promotes FOXM1 signaling. (A) Western blotting of the FH-inactive UOK262-FH^−/− cell line and the FH-reconstituted UOK262-FH^res cell line. (B) Quantitative reverse transcription-PCR shows that FH reconstitution in UOK262 cells decreases the relative transcript levels of FOXM1 and its downstream targets AURKA, AURKB, and CDK1. Changes in expression levels were normalized to values for UOK262-FH^res cells. GCLM is an NRF2 target gene known to be upregulated in HLRCC. Data are presented as means ± standard deviations of results from a representative experiment. * indicates a P value of <0.05 as determined by Student's t test. (C) Western blots of UOK262-FH^res cells treated with the vehicle or 80 μM monomethyl fumarate show increased FOXM1 signaling with MMF. (D) Western blots of UOK262-FH^res cells and HEK293 cells transduced with the pLKO-CMV empty vector (control) or pLKO-CMV-FTL and pLKO-CMV-FTH1 (ferritin) show that ferritin overexpression increases FOXM1 protein levels. (E) Western blots of UOK262 cells transduced with shRNAs targeting FTL (shRNA-FTL) or FTH1 (shRNA-FTH1) show decreased FOXM1 signaling relative to shRNA-Scr, a nontargeting shRNA control. (F) Cell viability measured by formazan production in UOK262 cells after transduction with shRNA-Scr, shRNA-FTL, and shRNA-FTH1. Data are presented as means ± standard deviations of results from a representative experiment. shRNA-FTL and shRNA-FTH1 knockdown curves were statistically significantly different from the shRNA-Scr curve for shRNA treatment (P < 0.001) but not for time (P > 0.05), as determined by two-way analysis of variance. (G) Relative mRNA levels of ferritin genes (FTL and FTH1), FOXM1, FOXM1 target genes (CDK1, AURKA, and AURKB), and nonferritin NRF2 targets (GCLM and AKR1B10) in HLRCC tumors and in normal kidney tissues. FDR, false discovery rate.

Previous reports indicated that FOXM1 was activated by ferritin through STAT3 signaling (10). Although UOK262-FH^−/− cells had increased STAT3 phosphorylation relative to UOK262-FH^res cells (see Fig. S6A in the supplemental material) and shRNA-mediated knockdown of either FTL or FTH1 decreased FOXM1 levels, only FTL knockdown reduced STAT3 phosphorylation (Fig. S6B). Additionally, ferritin overexpression increased FOXM1 levels (Fig. 6D) without altering STAT3 phosphorylation in UOK262-FH^res cells, indicating that ferritin may activate FOXM1 through a different mechanism (Fig. S6C).

To evaluate the validity of the ferritin-FOXM1 signaling pathway in clinical samples, we reanalyzed previously reported gene expression microarray data derived from HLRCC tumors and normal kidney tissues (GEO accession no. GSE26574 and GSE20896) (3, 15). Concordant with our cell culture models, HLRCC tumors showed significantly increased expression levels of FTL, FTH1, and FOXM1 as well as the FOXM1 target genes AURKA, AURKB, and CDK1 compared to those in normal kidney tissues (Fig. 6G and Table S3).

DISCUSSION

Studies of hereditary cancers have led to the discovery and characterization of many physiologically and pathologically important tumor suppressors, including TP53 (16), RB (17), PTEN (18), and APC (19). Many of these classical tumor suppressors, such as the cell cycle's retinoblastoma protein, RB, mediate cell growth. Indeed, sustaining chronic proliferation is one of the most fundamental hallmarks of cancer (4). While it is clear how somatic mutations in genes involved in growth signaling mediate the chronic proliferation signal, mechanisms by which mutations of metabolic tumor suppressors such as FH confer such a signal were not known. In this study, we show that fumarate, which accumulates as a result of FH inactivation, functions as a mitogen to activate chronic proliferative signaling. Mechanistically, fumarate increases ferritin gene transcription through NRF2 activation and increases ferritin gene translation by inhibiting IRP2. In turn, ferritin activates the progrowth transcription factor FOXM1 (Fig. 7). Previous reports on ferritin-induced growth signaling described how glioblastoma cancer cells had high ferritin levels due to increased iron intake (10). In contrast, we demonstrate how cancer cells use an electrophilic oncometabolite to hijack ferritin regulation in an iron-independent mechanism to increase ferritin levels.

FIG 7.

FH inactivation confers chronic proliferative signaling. The diagram shows how FH inactivation increases ferritin levels to promote proliferation. FH, a TCA cycle enzyme that catalyzes the hydration of fumarate (represented by green hexagons) to form malate (represented by yellow hexagons), is inactivated (represented by a red cross) in HLRCC. FH inactivation causes fumarate accumulation. The accumulated fumarate activates the NRF2 transcription factor, which transcribes the ferritin genes FTL and FTH1. Concurrently, the accumulated fumarate also inhibits IRP2, which normally represses the translation of FTL and FTH1, leading to a net increase in the ferritin level and subsequent FOXM1-dependent proliferative signaling.

The implications of increased ferritin levels in FH-deficient cells for signaling networks require further investigation. The established function of ferritin is to sequester free iron. Fumarate-mediated, iron-independent upregulation of ferritin could induce an intracellular iron deficiency, yet the concurrent accumulation of FBXL5 would indicate otherwise. These counterintuitive changes in IRP2, ferritin, and FBXL5 levels should be explored in the context of intracellular signaling; because iron is an important cofactor in heme- and iron-sulfur-containing enzymes, many carcinogenesis-associated pathways could be altered by iron-independent ferritin upregulation, including DNA synthesis (20, 21), DNA repair (22–24), metabolism (25, 26), and iron-dependent dioxygenase activity (27). Moreover, iron-independent functions of ferritin, such as its role in chemokine receptor-mediated cell migration or proinflammatory phosphatidylinositol 3-kinase (PI3K) signaling (28, 29), could be altered in FH-deficient cancers. Considering that we have shown that FOXM1 can be activated when ferritin is overexpressed, without changes to STAT3 signaling, as was previously reported, future inquiries into this pathway should evaluate if high ferritin levels activate FOXM1 through modulation of iron levels or through iron-independent signaling pathways.

Our results indicate that the thiol reactivity of fumarate underlies the mechanism by which it activates the observed chronic proliferative signaling, suggesting that other thiol-reactive compounds may also exert a similar effect. In addition to fumarate, fumarylacetoacetate (FAA) is another known cysteine-reactive oncometabolite (30). FAA accumulates following fumarylacetoacetate hydrolase (FAH) inactivation in hereditary tyrosinemia type 1 (HT1), a genetic disorder that can progress to hepatocellular carcinoma (HCC) (31, 32). Inhibition of FAA formation can significantly decrease cancer incidences, indicating that FAA accumulation drives HCC formation in HT1 patients (33). Our data imply that FAA may provide a chronic growth signal in HT1-associated HCC through a similar mechanism. Oncometabolites are not the only thiol-modifying compounds implicated in cancer. DMF, which is used clinically to treat psoriasis and multiple sclerosis, can induce kidney tumors akin to HLRCC at sufficiently high concentrations (200 and 400 mg/kg of body weight/day) (34).

Additionally, some environmental toxicants or their activated metabolites are also electrophiles capable of modifying cysteine residues: toxicants like acrolein can form cysteine adducts (35) and may also elicit growth modulation through concerted NRF2 activation and IRP2 inhibition. However, while NRF2 activation may be a common outcome of thiol-modifying compounds and Michael acceptors (36, 37), any generalizable inhibition of IRP2 repression by endogenous compounds and xenobiotics is premature. Indeed, in opposition to our description of fumarate-mediated decreases in IRP2-IRE binding activity, mice exposed to cigarette smoke showed increased IRP2-IRE binding activity, and a thiophene derivative has been shown to enhance IRP2 binding to the FTL IRE (38, 39). Furthermore, fumarate seems to differentially affect the translation of FTL and FTH1. Looking forward, a “cysteine code” may exist for IRP2, whereby increases or decreases to IRP2-IRE binding interactions are determined by which IRP2 cysteines are modified, what the structures of those modifications are, and how the adducts interact with IREs of diverse nucleotide compositions. Given the multitude of genes under IRE control, the alterations to iron signaling networks and downstream pathways may vary widely.

In conclusion, our results show how FH inactivation disrupts cellular iron signaling and induces a chronic proliferative phenotype, providing a mechanistic explanation for how the inactivation of FH can give rise to a fundamental cancer hallmark.

MATERIALS AND METHODS

Alignment of aconitase protein family members.

Amino acid sequences for mouse Aco2 (NCBI RefSeq accession no. NP_542364.1), human IRP2 (IREB2/ACO3) (accession no. NP_004127.1), and human IRP1 (IREB1/ACO1) (accession no. NP_001265281.1) were aligned by using Clustal Omega (40).

Reagents and chemicals.

Stock solutions of 50 mM DMF (catalog no. sc-239774; Santa Cruz Biotechnology, Dallas, TX) and 100 mM MMF (catalog no. 651419; Sigma, St. Louis, MO) were dissolved in dimethyl sulfoxide (DMSO) and serum-free medium (pH adjusted to 7.4 with HEPES buffer) (catalog no. 15630080-1M; Thermo Fisher, Waltham, MA), respectively. Stock solutions of 20 mM Fe (catalog no. F3388; Sigma) and 50 mM DFO (catalog no. D9533; Sigma) were dissolved in water. MG132 (catalog no. M7449; Sigma) was used as a proteasome inhibitor.

Cell culture conditions.

UOK262 cells were a generous gift from Marston Linehan (National Cancer Institute, NIH, Bethesda, MD) (5). HEK293 and HK2 cell lines were obtained from the ATCC (Manassas, VA). HEK293FT cells were purchased from Life Technologies (Thermo Fisher). UOK262, HK2, and derivative cells were cultured in RPMI 1640 supplemented with 10% fetal bovine serum (FBS). All FBS was heat inactivated for 30 min at 56°C. HEK293 cells were cultured in Dulbecco's modified Eagle's medium (DMEM) with high glucose (4.5 g/liter) (no pyruvate) supplemented with 10% FBS. HEK293FT cells were cultured in DMEM with no pyruvate and high glucose supplemented with 10% FBS and 500 μg/ml Geneticin. Cells were cultured at 37°C in atmospheric air enriched with 5% CO₂. For both long-term culture and experiments, HEK293 and HEK293FT cells were cultured on vessels coated with poly-d-lysine.

Construction of Flag-tagged FH vectors.

Transient expression of FH in UOK262 cells was achieved by using a lentiviral construct based on the pLKO.5 vector (Sigma). The U6 promoter was swapped out for a cytomegalovirus (CMV) promoter to generate pLKO.5-CMV. The FH open reading frame was cloned into pLKO.5-CMV to give rise to pLKO-FH.

Construction of FTL, FTH1, and Flag-tagged IRP2 expression vectors.

Total RNA was extracted from HEK293 cells by using TRIzol reagent (Life Technologies, Thermo Scientific) according to the manufacturer's protocols. First-strand cDNA synthesis was performed by using SuperScript IV reverse transcriptase (Life Technologies), using FTL-, FTH1-, or IREB2-specific primers. Subsequent second-strand syntheses, amplifications, and the introduction of an N-terminal Flag tag to IREB2 were performed by using Phusion DNA polymerase (NEB, Ipswich, MA). The resulting IREB2 open reading frame PCR product was cloned into the pIRESpuro3 vector (Clontech, Mountain View, CA) to make the pIRESpuro3-N-Flag-IRP2 construct. Site-directed mutagenesis was performed with the Q5 site-directed mutagenesis kit (catalog no. E0552S; NEB) to generate the A523R mutant pIRESpuro3-N-Flag-IRP2. FTL and FTH1 open reading frame PCR products were cloned into pLKO.5-CMV to generate pLKO-CMV-FTL and pLKO-CMV-FTH1, respectively.

Construction of FTL and FTH1 5′-UTR luciferase vectors.

DNA fragments corresponding to chromosome 19, 48965309, through chromosome 19, 48965507 (human genome reference version GRCh38/hg38), and chromosome 11, 61967426 through chromosome 11, 61967660 (GRCh38/hg38), which correspond to the 5′ UTRs of the FTL and FTH1 genes, respectively, were amplified from human genomic DNA (gDNA) isolated from HEK293 cells. The firefly luciferase open reading frame was amplified from pGL4.21 (Promega, Madison, WI) by PCR using appropriate forward primers to allow splice by overlap extension PCR with the FTL and FTH1 5′ UTRs isolated previously. The resulting PCR products were cloned into the pECE vector (Addgene, Cambridge, MA), giving rise to pECE-FTL-IRE-LUX and pECE-FTH1-IRE-LUX, respectively. The Renilla luciferase open reading frame derived from pRL (Promega) was cloned into pECE to produce pECE-RL. This vector was used as a transfection control in dual-luciferase assays.

Construction of the Myc-NRF2 vector.

pCDNA3-Myc3-Nrf2 was made available from Addgene following its characterization (41). Myc-tagged NFE2L2 was isolated from pCDNA3-Myc3-Nrf2 and cloned into pIRESpuro3 to make NRF2-Myc. Cells were transfected with this plasmid by using Attractene transfection reagent (Qiagen, Valencia, CA).

shRNA-Scr, shRNA-FTL, and shRNA-FTH1 vectors.

The pLKO.5 vector (Sigma) was used to generate the following plasmids: shRNA-FTL (21-bp sequence of 5′-CTG GAG ACT CAC TTC CTA GAT-3′), shRNA-FTH1 (21-bp sequence of 5′-GCC GAA TCT TCC TTC AGG ATA-3′), and the shRNA-Scr scrambled control (21-bp sequence of 5′-GAC TAG AAG GCA CAG AGG GAT-3′). shRNA-Scr was used as a nontargeting control plasmid and targets no known human genes.

Lentivirus production and transient transduction.

Lentivirus for all pLKO plasmids was produced by using the ViraPower lentiviral expression system (Thermo Fisher) according to the manufacturer's protocols. For transient lentiviral transduction, cells were used for experiments 48 h after transduction. Transduction efficacy was verified via Western blotting.

Stable cell line production for UOK262-FH^res and UOK262-FH^−/− cells.

UOK262 parental cells were transduced with pLKO.5-CMV and pLKO.5-FH to generate UOK262-FH^−/− and UOK262-FH^res cells, respectively. Cells with FH stably incorporated into their genome were selected by using 2 μg/ml puromycin. Long-term cultures of the resulting stably transfected cells were maintained in puromycin medium, while experiments using these cells were conducted without puromycin.

Generation of NRF2 and FH knockout HEK293 lines.

NRF2 was knocked out in cells by using a commercial CRISPR-Cas9 NRF2 knockout system (catalog no. sc-400017; Santa Cruz) according to the manufacturer's recommendations. The knockout was verified by using Western blotting and Sanger sequencing. FH was knocked out in cells by using methods described previously (42). Two sets of oligonucleotides were used to generate single-guide RNAs (sgRNAs) to target Cas9 to FH gene loci: set A primers SiteAfor (5′-CAC CGG GAG GCA CTG CTG TTG GTA C-3′) and SiteArev (5′-AAA CGT ACC AAC AGC AGT GCC TCC C-3′) and set B primers SiteBfor (5′-CAC CGG AGC TCA TAG ATT CTT GGC A-3′) and SiteBrev (5′-AAA CTG CCA AGA ATC TAT GAG CTC C-3′). sgRNAs were cloned into pSpCas9(BB)-2A-GFP (Addgene) to generate pSpCas9-GFP-A and pSpCas9-GFP-B. To generate a homology-directed repair (HDR) template plasmid, we isolated the puromycin resistance cassette from pGL4.21 (Promega) using BamHI and SalI restriction sites and cloned the cassette into pUC19 (Addgene) to generate pUC19-Puro; DNA sequences flanking the sgRNA sites were added to the 5′ and 3′ ends of the puromycin cassette to generate homology-directed repair (HDR) arms. To generate a 5′-HDR arm, a DNA fragment corresponding to chromosome 1, 241668363, through chromosome 1, 241669362 (human genome reference version GRCh37/hg19), was isolated from HEK293 gDNA. To generate a 3′-HDR arm, a DNA fragment corresponding to chromosome 1, 241669413, through chromosome 1, 241670412 (GRCh37/hg19), was similarly isolated. Isolated 5′-HDR arm and 3′-HDR arm PCR products were subsequently cloned into pUC19-Puro to generate pUC19-Puro-FHHDR. HEK293 cells were transfected with pSpCas9-GFP-A, pSpCas9-GFP-B, and pUC19-Puro-FHHDR. Cells with FH stably knocked out were selected by using 2 μg/ml puromycin, and single colonies were screened for FH knockout by using immunoblotting and PCR amplification of the FH gene from HEK293-FH^−/− cell gDNA. Primers used for gDNA amplification were 5′-CTG GTA GAT TTT AAT GGC ATG CTG-3′ and 5′-AAC CCT CAT CCT TCC CTA TAC TTT G-3′. Long-term cultures of the resulting stably transfected cells were maintained in puromycin medium, while experiments using these cells were conducted without puromycin.

MMF and DMF treatments.

For MMF and DMF treatments, cells were treated with the specified concentrations of MMF or DMF and were refreshed every 24 h for at least 48 h before being harvested for analyses.

UOK262-FH^−/−, UOK262-FH^res, and UOK262-shRNA cell proliferation assays.

Early-passage UOK262-FH^−/− and UOK262-FH^res cells were seeded into 96-well plates at 3,000 cells/well in medium containing 80 μM MMF or vehicle controls. Cell proliferation was measured by using the CellTiter 96 AQueous One Solution cell proliferation assay [3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium, inner salt (MTS)] (Promega) according to the manufacturer's recommended protocols. For all groups, media were refreshed every 24 h. For shRNA proliferation assays, parental UOK262 cells were transduced with specified shRNA viruses on a 10-cm dish. The following day, medium was refreshed. After an additional 24 h, cells were seeded into 96-well plates at 3,000 cells/well, and cell proliferation was measured daily.

Immunoblotting.

Primary antibodies to β-actin (ACTB) (1:10,000 in milk) (catalog no. A1978; Sigma), AKR1B10 (1:1,000 in bovine serum albumin [BSA]) (catalog no. sc-100501; Santa Cruz), FBXL5 (1:2,500 in milk) (catalog no. 672602; BioLegend, San Diego, CA), FH (1:1,000 in BSA) (catalog no. 4567; Cell Signaling, Danvers, MA), Flag (1:1,000 in milk) (catalog no. 8146; Cell Signaling), FOXM1 (1:500 in milk) (catalog no. sc-502; Santa Cruz), IRP2 (1:1,000 in BSA) (catalog no. sc-33682; Santa Cruz), FTH1 (1:1,000 in BSA) (catalog no. sc-25617; Santa Cruz), FTL (1:1,000 in BSA) (catalog no. sc-74513; Santa Cruz), Myc (1:1,000 in milk) (catalog no. 2276; Cell Signaling), NQO1 (1:1,000 in milk) (catalog no. sc-32793; Santa Cruz), NRF2 (1:1,000 in milk) (catalog no. sc-13032; Santa Cruz), phospho-STAT3 (1:1,000 in milk) (catalog no. 9145; Cell Signaling), STAT3 (1:1,000 in milk) (catalog no. 9139; Cell Signaling), and ubiquitin (1:1,000 in milk) (catalog no. 3936; Cell Signaling) were used for immunoblotting experiments. Band densitometry was quantified by using Image Lab software (Bio-Rad).

Immunoprecipitation and mass spectrometry analysis of Flag-tagged IRP2.

The Flag-tagged IRP2 protein was overexpressed in HEK293 cells by transiently transfecting the cells with the pIRESpuro3-N-Flag-IRP2 vector. On days 2 and 3 posttransfection, cells were cotreated with 40 μM dimethyl fumarate and 200 μM deferoxamine. Cells were harvested for immunoprecipitation on day 4 posttransfection. Immunoprecipitation experiments were performed by using a magnetic DYKDDDDK immunoprecipitation kit (catalog no. 635696; Clontech) according to the manufacturer's recommendations. Samples were eluted in 1× Laemmli sample buffer with a protease inhibitor and β-mercaptoethanol and then resolved on 7% SDS-PAGE gels. A band of approximately 105 kDa, as visualized by Coomassie blue staining, was excised for liquid chromatography (LC)-MS/MS analysis.

Excised bands were reduced, alkylated, and digested in gel by using trypsin as previously described (43). In-gel digests were then desalted, fractionated online by reversed-phase chromatography using a Thermo Fisher easy-nLC 1000 liquid chromatography system, and analyzed by tandem mass spectrometry on a Thermo Fisher Q-Exactive mass spectrometer (44, 45). Database searching was performed by using the MSGF⁺ search algorithm and a human protein database and considered differential modification of cysteine residues by carbamidomethylation (+57.021464 atomic mass units [amu]), 2-succination (+116.010959 amu), 2-monomethyl-succination (+130.026609 amu), and 2-dimethyl-succination (+144.042259 amu) (46). Peptide spectral matches were filtered by using a percolator-derived q value of 0.01 (47).

Ubiquitylation of IRP2.

HEK293 cells were transiently transfected with the pIRESpuro3-N-Flag-IRP2 vector. On days 2 and 3 posttransfection, cells were treated with 80 μM MMF. On day 4, cells were cotreated with 10 μM MG132, 80 μM MMF, or 200 μM Fe, as indicated, for 4 h. Whole-cell extracts were harvested, the protein level was normalized by a bicinchoninic acid (BCA) protein assay, and extracts were utilized for downstream immunoprecipitation and SDS-PAGE analyses as described above.

Luciferase assays.

HEK293 cells were plated onto 10-cm dishes in DMEM supplemented with 10% FBS and transfected the same day with plasmids pECE-RL and either pECE-FTL-IRE-LUX or pECE-FTH1-IRE-LUX by using the Attractene transfection reagent (Qiagen). The following day, transfected cells were transferred into 6-well plates. After cells were allowed to adhere for 24 h, the cells were dosed with appropriate concentrations of compounds diluted in RPMI plus 10% FBS. Media and chemicals were refreshed every 24 h. Following 48 h of exposure, cells were lysed and used in the dual-luciferase reporter (DLR) system (Promega) according to the manufacturer's recommendations.

RNA interference.

siRNA transfections were performed with Oligofectamine reagent (Thermo Fisher) according to the manufacturer's protocols. Two different species of FH-targeting siRNAs were used: siFH-1 (catalog no. FH HSS103687; Thermo Fisher) and siFH-2 (catalog no. FH HSS103688; Thermo Fisher). A nontargeting siRNA silencer negative control (catalog no. AM4637; Thermo Fisher-Ambion) was used as a nontargeting scrambled siRNA (siScr) transfection control. Cells were harvested for immunoblotting at 72 h posttransfection.

qPCR.

Total RNA was extracted from adherent cells by using TRIzol reagent according to the manufacturer's recommended protocols. cDNA was synthesized by using a High-Capacity cDNA reverse transcription kit (catalog no. 4368814; Thermo Fisher/Applied Biosystems), and qPCR was performed by using TaqMan Fast Advanced master mix (Applied Biosystems, Foster City, CA) with an appropriate TaqMan probe. TaqMan probes (Thermo Fisher/Applied Biosystems) targeting ACTB (catalog no. 4352935), AURKA (catalog no. Hs01582072_m1), AURKB (catalog no. Hs00945855_g1), CDK1 (catalog no. Hs00938777_m1), FOXM1 (catalog no. Hs01073586_m1), FTH1 (catalog no. Hs01000476_g1), FTL (catalog no. Hs00830226_gH), and GCLM (catalog no. Hs00157694_m1) were used in this study. Results were analyzed by using the 2^−ΔΔC_T method (48).

Statistical and microarray analyses.

All statistical analyses were performed in the R statistical environment (49). Gene expression microarray data were processed by using a robust multichip averaging algorithm (50). Differentially expressed genes between tumors and normal kidney tissues were determined by using the linear model for microarray analysis algorithm (51).