Targeting NRF2 uncovered an intrinsic susceptibility of acute myeloid leukemia cells to ferroptosis

Bone marrow (BM) samples were obtained from sixteen patients with newly diagnosed AML (characteristics provided in the Supplemental Table 1). Peripheral blood (PB) samples were obtained from eight healthy individuals (HIs). All samples were obtained with written informed consent in accordance with the ethical committee of the First Affiliated Hospitals of Jinan University.

AML cell lines HL60, MV4;11, KG1a, U937, HEL, Molm13, NB4, KG1 and Kasumi-1 were cultured in RPMI 1640 (Invitrogen) containing 10% fetal bovine serum (FBS). HEK293T cells were cultured in DMEM (Invitrogen) medium containing 10% FBS. All these cell lines were cultured in a humidified atmosphere containing 5% CO₂ at 37 °C. Ferrostatin-1, ML385, FIN56, and RSL3 were purchased from Selleck Chemicals (Houston, TX, USA).

Specific oligonucleotides directed against human NRF2 were designed and cloned into the pLKO.1-puro plasmid according to the protocol recommended by Addgene, and empty pLKO.1-puro plasmid served as control. Lentiviruses carrying pLKO.1-puro were prepared by the transfection of standard packaging vectors into HEK293T cells using lipofectamine 3000 (Life Technologies, USA). Viral supernatant was harvested 24 and 48 h after transfection. A total of 2 × 10⁵ MV4;11 cells were co-cultured with virus supernatant plus 8 µg/mL polybrene (Sigma-Aldrich, St Louis, USA). Then 5 µg/mL puromycin was applied to select positively infected cells for at least 5 days. The shRNA sequences are listed in Supplemental Material 1. For ectopic expression of NRF2, lentiviruses overexpressing NRF2 (LV-NRF2) and empty viruses (LV-NC) were purchased from Genechem (Shanghai, China). A total of 5 µg/mL puromycin was used to select positively infected cells for at least 5 days.

For analysis of cell death, cells were harvested, and the Annexin-V-APC/PI Detection Kit (MultiSciences, Shanghai, China) was used following the manufacturer’s protocols and performed as previously described [25, 26]. For cell cycle assay, the Cell Cycle Staining Kit (MultiSciences, Shanghai, China) was used according to the manufacturer’s instructions. Cells were collected by CytoFLEX, and data were analyzed by Flowjo software.

Total RNA from cells were extracted using TRIzol reagent (Invitrogen, CA, USA) following the manufacturer’s instructions and was reversed transcribed into cDNA using High-Capacity cDNA Reverse Transcription Kits (Applied Biosystems, CA, USA). Quantitative real-time PCR (qRT-PCR) was performed with SYBR Green (Tiangen, China) in a total volume of 20 µl using the following program: 95°C for 5 min followed by 40 cycles of 95℃ for 30s, 60℃ for 30s, and 72℃ for 30s. Relative target gene expression levels were normalized to the mRNA level of β-actin. The primers for qRT-PCR are shown in Supplemental Material 1.

Cells were collected and lysed in 1х SDS loading buffer containing protease inhibitors. Protein lysates were separated using 12% SDS-PAGE and subsequently transferred to polyvinylidene fluoride membrane (Millipore, USA) following by blocking with QuickBlock™ blocking buffer (Beyotime, China) and incubation with antibodies directed against GPX4 (1:1000, 52,455, CST, MA, USA) or NRF2 (1:800, sc-365,949, Santa cruz) overnight at 4℃. Anti-β-actin antibody (1:5,000, Beyotime, China) served as control. After incubation with a secondary antibody, ECL (Beyotime, China) was used to amplify the binding antibody signal, and images were obtained with a UVITEC photo documenter. The full gel images in the paper figures are displayed in Figure S1.

The Neon® Transfection System (Invitrogen) was used to transfect 100 pmol oligonucleotides into MV4;11 cells in a total volume of 10 µl [27, 28]. In brief, 2 × 10⁵ cells were suspended in a volume of 10 µl with 100 pmol oligonucleotides and transfected three times in each 12-well plate. After electroporation, cells were cultured in RPMI 1640 medium supplemented with 10% FBS at 37 °C and 5% CO₂ for 24 h, and subsequently divided into two groups with or without drug treatment. The cells were then collected for protein level detection and evaluation of cell death. Supplemental Material 1 lists the siRNA sequences targeting GPX4.

To detect LPO, C11-BODIPY 581/591 (Abclonal, China) was employed following the manufacturer’s instructions. In brief, cells were labeled with C11-BODIPY (2.5 µM) in 400 µL RPMI 1640 medium and incubated at 37 °C and 5% CO₂ in a cell incubator away from light for 60 min. The cells were then washed twice with PBS, resuspended in 200 µL of fresh PBS, and analyzed using flow cytometry.

RNA from LV-NC and LV-NRF2 MV4;11 cells was extracted using TRIzol reagent. Amplified cDNA was sequentially converted into short-read sequencing libraries using the VAHTS Universal V6 RNA-seq Library Prep Kit for Illumina® Hisat2 (version: 2.0.4). Sequencing was performed by Novogene Co. Ltd., Tianjin Biotechnology Corporation. The expression level of each gene was calculated as fragments per kilobase of exon model per million mapped reads (FPKM).

Unless otherwise stated in the figure legends, all data are presented as the mean ± SD of three independent experiments. Statistical analysis was conducted using SPSS 22.0 software on data from independent biological replicates. Student’s t-test was used to determine the significance of differences between two groups. One-way ANOVA followed by either Bonferroni or Dunnett’s post hoc test was applied to compare three or more groups, as specified in the figure legends. A p value < 0.05 was considered significant. *, p < 0.05, **, p < 0.01, ***, p < 0.001.

Our recent studies have demonstrated that NRF2 promotes the expression of key molecules involved in AML development and drug sensitivity. To identify potential therapeutic targets for AML, we performed a broad screening of direct NRF2 target genes (Fig. 1a). Our hypothesis was that NRF2 target genes would contain ChIP-Seq binding peaks in their vicinity and show altered expression levels in RNA-seq studies. To generate a list of NRF2 target genes, we initially utilized several ChIP-Seq profiling datasets (ENCSR707IUN_1, GSM2394418, GSM2423706, GSM2423560) and identified 349 unique overlapping genes. These represent a high-confidence set of NRF2-bound genomic regions (Fig. 1b). The majority of the identified binding sites were found near potential NRF2 target genes, suggesting that these genes could be regulated by NRF2. We further performed KEGG pathway analysis on these potential NRF2 target genes and found that they were significantly enriched in several terms, including ferroptosis and glutathione metabolism, both of which are closely related to ferroptosis (Fig. 1c, Supplemental Material 2). Next, we conducted RNA-seq to identify NRF2 target genes in AML. We identified over 3,000 upregulated genes in NRF2-overexpressing MV4;11 cells (log2 fold change ≥ 0.25). Our results revealed the upregulation of numerous antioxidative and ferroptosis genes in response to NRF2 overexpression. Subsequently, we identified 70 unique genes overlapping between ChIP-Seq and RNA-seq (Fig. 1d), and compiled them in Supplemental Material 3. Consistent with the previous findings, we found that the NRF2 targets are mainly involved in the processes of ferroptosis and glutathione metabolism (Fig. 1e, Supplemental Material 4). Both NRF2 and ferroptosis have been shown to be linked to cancer therapy resistance. It is intriguing to speculate that these NRF2 targets may act in concert to function as anti-ferroptosis mediators upon drug treatment in AML.

Using qRT-PCR analysis, we validated that several of these genes are indeed transcriptional targets of NRF2 in AML cells (Fig. 1f-g). Among the selected genes, including SLC7A11, GCLC, GPX4, FTH1, HMOX1, GCLM, and FTL, GPX4 mRNA was upregulated upon lentivirus-mediated NRF2 overexpression and downregulated upon shRNA-mediated NRF2 (shNRF2) knockdown. Furthermore, GPX4 was also upregulated in response to NRF2 activation by arsenic trioxide (ATO) and cytarabine (Ara-C) treatment (Figure S2). Given the known activation of NRF2 in AML and its resistance to therapies, we propose that GPX4 may be one of the key targets that confer an anti-ferroptosis role to NRF2.

To confirm the regulatory role of NRF2 on GPX4 in AML, we used western blotting to detect changes in GPX4 expression after manipulating NRF2 expression. As shown in Fig. 2a and b, overexpression of NRF2 led to increased GPX4 expression, while knockdown of NRF2 resulted in decreased GPX4 expression. Moreover, we observed higher expression of NRF2 and GPX4 in the BM of AML patients compared with HIs (Fig. 2c). Among several AML cell lines, we found high expression of GPX4 in MV4;11 and Kasumi-1 cells (Fig. 2c, right panel), which were subsequently used in our experiments. We also analyzed the expression of NRF2 and GPX4 in 15 AML patients and found a positive correlation between the two (Fig. 2d). Furthermore, survival analysis of AML patients included in The Cancer Genome Atlas (TCGA) dataset showed that both NRF2 and GPX4 expression were significantly associated with poor overall survival (OS) (Fig. 2e and Figure S3a-b), indicating that their increased expression is associated with poor prognosis in AML. In addition, the NRF2 inhibitor ML385 reduced GPX4 protein levels in a concentration-dependent manner (Fig. 2f). We also found that inhibiting NRF2 with ML385 inhibited the viability of various AML cell lines (Fig. 2g), suggesting that NRF2 and GPX4 could represent promising targets for the treatment of AML cancer cells.

It has been reported that RSL3 and FIN56 can specifically inhibit GPX4 to induce the ferroptosis of tumor cells [14, 29, 30]. To investigate the synergistic effects of NRF2 and GPX4 inhibition in AML cells, we first explored the use of these specific GPX4 inhibitors to induce ferroptosis in AML cells. Our experiments showed that treatment with GPX4 inhibitors resulted in a dose-dependent suppression of GPX4 expression in MV4;11 cells (Fig. 3a). We also found that RSL3 inhibited the viability of various AML cell lines in a concentration-dependent manner, while having little effect on peripheral blood mononuclear cells (PBMCs) from HIs, indicating its relative safety (Fig. 3b). Next, AML cell lines MV4;11, Kasumi-1, and KG1α were treated with NRF2 inhibitor ML385 alone, a GPX4 inhibitor (FIN56 or RSL3) alone, or ML385 combined with a GPX4 inhibitor. We found that combined administration of ML385 and GPX4 inhibitor markedly inhibits the proliferation of AML cell lines compared with either inhibitor alone, and the combination index (CI) indicating that these two inhibitors have synergistic effects (Fig. 3c and d and S4a).

Further studies revealed that treatment with ML385 and a GPX4 inhibitor significantly increases the cell death of MV4;11 and Kasumi-1 cells (Fig. 4a, b). In line with these findings, it was observed that ML385 acted in collaboration with Erastin, another ferroptosis inducer, to induce cell death in MV4;11 cells (Figure S4b). Furthermore, our study has uncovered that RSL3 enhances the effectiveness of the primary drugs used to treat AML like Ara-C (Figure S4c). Additionally, we have observed that treatment with a combination of ML385 and GPX4 inhibitors results in the arrest of the cell cycle of MV4;11 cells in the G0/G1 phase, as demonstrated in Fig. 4c. We further examined if the co-administration of ML385 and a GPX4 inhibitor would augment the cytotoxic effects in cells obtained from newly diagnosed AML patients. Comparable to the findings in AML cell lines, the combined use of ML385 and FIN56/RSL3 inhibited the viability and stimulated the cell death of primary AML cells, as depicted in Fig. 4d and e. In summary, these outcomes imply that the combination of inhibiting NRF2 and GPX4 produces a more robust cytotoxic effect on AML cells.

Previous studies have shown that GPX4 plays a crucial role in ferroptosis and chemoresistance in human cancers. Activation of the NRF2 pathway plays a dominant role in protecting human cancer cells against cell death. To investigate potential synergistic effects of ML385 and GPX4 inhibitors on ferroptosis, we assessed the combined impact of these agents on lipid peroxidation levels using the C11-BODIPY probe. As depicted in Fig. 5a and b, co-treatment with ML385 and GPX4 inhibitors markedly elevated lipid-based ROS accumulation in MV4;11 and Kasumi-1 cells. Consistent with these results, simultaneous treatment with ML385 and RSL3 led to the highest cell death compared to ML385 or RSL3 alone in MV4;11 cells (Fig. 5c). Furthermore, we confirmed that the enhanced cytotoxic effects of ML385 and RSL3 in AML cell lines was blocked by ferrostatin-1 (Fer-1, an inhibitor of ferroptosis) (Fig. 5d). The above data suggested that ML385 may act synergistically with FIN56/RSL3 to induce ferroptosis via the NRF2/GPX4 pathway.

To further investigate the synergistic effects of NRF2 and GPX4 inhibition, MV4;11 and Kasumi-1 cells were treated with ML385 or GPX4 inhibitor alone or ML385 combined with GPX4 inhibitor, and the expression of NRF2 and GPX4 was detected by western blotting. As shown in Fig. 6a-b, the GPX4 inhibitor combined with the NRF2 inhibitor resulted in the lowest expression of GPX4 and NRF2. At the same time, transfection of shNRF2 sharply increased the sensitivity of MV4;11 cells to GPX4 inhibitors (Fig. 6c). We next measured the level of intracellular lipid ROS in shNRF2-transfected MV4;11 cells to confirm whether knockdown of NRF2 could increase the oxidation of lipids. As expected, down-regulation of NRF2 promoted GPX4 inhibitors-induced lipid oxidation (Fig. 6d). To investigate the impact of NRF2 expression on AML cells treated with GPX4 inhibitors, we overexpressed NRF2 in MV4;11 cells and treated them with RSL3 at the indicated dose. Our results showed that NRF2 overexpression partially restored the cytotoxic effects of RSL3 on MV4;11 cells, as depicted in Fig. 6e. Additionally, we employed two small interfering RNAs (siRNAs) to target GPX4 and explored its effects on the reactivity of AML cells to ferroptosis (Fig. 6f). Our findings indicated that transfection of siRNA targeting GPX4 significantly increased the sensitivity of MV4;11 cells to RSL3 and FIN56, as demonstrated in Fig. 6g. Overall, our results suggest that targeting NRF2 and GPX4 may enhance the sensitivity of AML cells to ferroptosis.