Blocking NRF2 Translation by Inhibition of Cap-Dependent Initiation Sensitizes Lymphoma Cells to Ferroptosis and CAR T-cell Immunotherapy

Abstract

Cancers co-opt stress response pathways to drive oncogenesis, dodge immune surveillance, and resist cytotoxic therapies. Several of these pathways also provide protection from ferroptosis, an iron-dependent oxidative cell death pathway triggered by clinically available drugs, including chemotherapies, rheumatologic agents, and novel ferroptosis inducers under evaluation in clinical trials. In this study, we found that disrupting cap-dependent translation initiation in diffuse large B-cell lymphoma (DLBCL) sensitizes cells to ferroptosis. Specifically, the eIF4A1 inhibitor zotatifin synergized with pharmacologic ferroptosis inducers primarily through suppression of glutathione production, which protects polyunsaturated fatty acids from ferroptotic oxidation. Loss of nuclear factor erythroid 2-related factor 2 (NRF2) translation, a master regulator of antioxidant genes, was a key consequence of rocaglates, including zotatifin, and other disruptors of cap-dependent initiation. Although NRF2 loss alone was insufficient to trigger ferroptosis, it lowered the antioxidant threshold, sensitizing cells to lipid peroxidation and ferroptotic death under additional oxidative stress. In vivo, combining zotatifin with the optimized ferroptosis inducer imidazole ketone erastin significantly reduced tumor burden in DLBCL patient-derived xenografts. Treatment with zotatifin in combination with chimeric antigen receptor (CAR) T cells, a vital treatment modality for patients with DLBCL, revealed that zotatifin preexposure sensitized DLBCL tumors to CD19-directed CAR T cells in vitro and extended survival of CAR T-cell-treated immunocompetent mice bearing syngeneic DLBCL tumors in vivo. Overall, eIF4A1 inhibition-induced translational disruption provides opportunities to leverage the therapeutic impacts of ferroptosis inducers, including cytotoxic immunotherapies.

Significance: Translational disruption causes NRF2 loss that sensitizes lymphomas to ferroptosis and enhances CAR T-cell and drug efficacy, highlighting eIF4A1 targeting as a promising therapeutic strategy for treating cancer.

Figure 1:. Upregulation of ferroptosis protection during rocaglate exposure.

A) TMT-pSILAC workflow to assess zotatifin-induced changes in protein synthesis over 8 h in SU-DHL10 cells. Created in BioRender. Manara, P. (2025) https://BioRender.com/7mfy9vi. B) Proteins upregulated (log₂FC ≥ +0.5) across three independent TMT-pSILAC replicates. C) Scatter plot showing 587 genes significantly altered (pValue < 0.01) at both the protein (TMT-pSILAC) and mRNA (RNA-seq) levels in zotatifin-treated vs. untreated SU-DHL10 cells. Axes represent log₂FC for RNA-seq (y-axis) and TMT-pSILAC (x-axis). Seven labeled genes met the significance threshold (log₂FC ≥ +0.5 or ≤ −0.5) in both datasets. D) Enrichr KEGG pathway analysis of genes significantly up- or downregulated (log₂FC ≤−0.5 or log₂FC ≥+0.5 in all three replicates). Top 10 pathways are ranked by −log₁₀(adjusted P-value). E) Heatmap of ferroptosis-related proteins (TFRC, SLC3A2, ACSL4, CBS) with log₂FC ≥ +0.5 across replicates. Matrix_6 (mat1) shows Z-score normalized intensities across six TMT channels (126–131), with a color scale from pink (downregulated) to purple (upregulated). F) Immunoblot time course of SLC7A11, CBS, SLC3A2, ACSL4, and MYC in SU-DHL10 (left), BJAB (center), and OCI-Ly1 (right) cells treated with 12.5 nM zotatifin. Cyclophilin B (CYPB) serves as a loading control. Data represent three independent experiments.

Figure 2:. Sensitization to ferroptosis by zotatifin.

A) Linear regression of CCLE mRNA expression for SLC3A2, NFS1, GPX4, FTH1, ACSL4, TFRC, and CBS vs. CR-1–31B sensitivity (AUC, CTD2) from DepMap. B) Immunoblot of SU-DHL10 cells with doxycycline (doxy)-inducible SLC3A2 (low/high exposure) vs. non-induced controls, treated with 12.5 nM zotatifin for 24–72 h. CYPB = loading control. Two independent experiments. C-D) Effects of SLC3A2 overexpression after 48 h zotatifin (12.5 nM) treatment, measured by (C) ATP-based viability assay and (D) Annexin-V/PI flow cytometry. Two-way ANOVA with Sidak’s (C) and Bonferroni’s (D) multiple comparisons (*ns, **P < 0.001, ****P < 0.0001). For C) statistical analysis compared 48 h cell viability across zotatifin doses between doxy-induced SLC3A2-overexpressing and control cells. E) Dose-response curves of U266 WT and U266 SLC3A2 KO cell lines treated with varying drug concentrations. The % growth inhibition (GI) is plotted against the log-transformed drug concentration. GI₅₀ values were determined using nonlinear regression (Log(inhibitor) vs. normalized response – Variable slope model). Statistical comparison of GI₅₀ values between U266 WT and U266 KO showed a significant difference (p = 0.0202). Data represent the mean ± SD of 4 replicates per group. F) BODIPY C11 oxidation ratio (oxidized/non-oxidized [Ox/non-Ox]) Mean Fluorescence Instensity (MFI) measured by flow cytometry and G) GSH/GSSG ratio calculated using the concentrations of GSH and GSSG at different zotatifin concentrations (3.12 nM to 100 nM) in SU-DHL10 (left) and OCI-Ly1 (right) cells. Ferrostatin-1 was used to rescue ferroptosis. F-G) Statistical analysis was performed using repeated measures two-way ANOVA with Bonferroni’s multiple comparison test (ns, *P < 0.05, **P < 0.01, **P < 0.001, ****P < 0.0001). Data are mean ± standard error of the mean (SEM), of three independent experiments. H) Schematic representation of molecular ferroptosis protection mechanisms and targeted pathways of ferroptosis. Created in BioRender. Manara, P. (2025) https://BioRender.com/im5jhqc. I) Bliss δ synergy score heatmap for zotatifin in combination with various ferroptosis inducers (erastin, SASP, RSL3, DMF, PRLX93936, altretamine) and inhibitors (NAC, ferrostatin-1) in RIVA, TMD8, BJAB, SU-DHL10, and OCI-Ly1 DLBCL cell lines. Bliss δ > 10 = synergy; −10 to 10 = additive; < −10 = antagonism.

Figure 3:. Zotatifin Enhances Oxidation-Driven Damage Induced by Ferroptosis Inducers.

A) GSH/GSSG ratio in SU-DHL10 (left) and OCI-Ly1 (right) cells treated with zotatifin alone or with ferroptosis inducers (erastin, RSL3, DMF) or the inhibitor (ferrostatin-1). Statistical analysis was performed using repeated measures two-way ANOVA with Bonferroni’s multiple comparison test (ns, *P < 0.05, **P < 0.01, **P < 0.001, ****P < 0.0001). Data are mean ± SEM of three independent experiments. B) Immunoblot for SLC3A2, SLC7A11, cleaved caspase 3, XIAP, MYC, and GPX4 in SU-DHL10 (left) and OCI-Ly1 (right) cells treated with erastin (600 nM) + zotatifin (12.5 nM) for 24 h. CYPB was used as a loading control. Data from two independent experiments. C) BODIPY C11 oxidation ratio (Ox/non-Ox MFI) and ROS in SU-DHL10 (left) and OCI-Ly1 (right) cells stained with (D) MitoSox and (E) DHE measured by flow cytometry in SU-DHL10 (left) and OCI-Ly1 (right) cells treated with zotatifin alone or with ferroptosis inducers (erastin, RSL3, DMF) or inhibitor (ferrostatin-1). (A), (C-E) statistical analysis was performed using repeated measures two-way ANOVA with Bonferroni’s multiple comparison test (ns, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001).

Figure 4:. eIF4A1-mediated loss of NRF2 expression upon rocaglate treatment.

A) Immunoblots for NRF2, MYC, and HO-1 in SU-DHL10 and OCI-Ly1 cells treated with indicated zotatifin doses for 24 h. CYPB was used as a loading control. Data from three independent experiments. B) Polysome profiling of SU-DHL10 cells treated with vehicle or 50 nM zotatifin for 24 h. Mean ± SEM, two independent experiments. C) Translation efficiency measured as the ratio of polysome-associated to free/monosome-associated mRNA in SU-DHL10 cells treated with 50 nM zotatifin for 24 h. Mean ± SEM, two experiments. Unpaired t tests: *P < 0.05, ****P < 0.0001. D) Schematic dual-luciferase assay in SU-DHL10 cells transfected with either wild-type or A4-mutant hNFE2L2 5′UTR:Rluc:IRES:Fluc reporters (left). Created in BioRender. Manara, P. (2025) https://BioRender.com/hgmv38s. Cells treated for 1 h with zotatifin (50 nM), RocA (25 nM), silvestrol (25 nM), hippuristanol (100 nM), 4EGi-1 (50 μM), or DMSO (right). Repeated measures two-way ANOVA with Bonferroni’s test: ***P < 0.001, ****P < 0.0001. Mean ± SEM, three independent experiments. E) Immunoblots of HAP1 cells expressing EIF4A1 WT, F163F, or F163L alleles, treated with 100 nM zotatifin for 0–6 h. CYPB was used as a loading control. Data are representative of two independent experiments. F) Left: schematic of 6 h pull-down assay using NRF2 5′UTR RNA baits following treatment with zotatifin (12.5 nM) or RocA (6.25 nM). Generic and AG sequences used as negative/positive controls. Right: diagram of NRF2 5′UTR showing four baits (Bait 1–4, marked by increasing asterisks). Created in BioRender. Manara, P. (2025) https://BioRender.com/z5j8wtw. G) Immunoblot of pull-down (top) and input lysates (bottom) from U87MG cells treated as in (F). Blots probed for eIF4A1 and DDX3 to assess NRF2 5′UTR binding. CYPB was used as a loading control. Data representative of two independent experiments.

Figure 5:. eIF4A1-mediated loss of NRF2 expression sensitizes to ferroptosis.

A) Bliss δ synergy heatmap for zotatifin + ML385 in DLBCL cell lines (RIVA, TMD8, BJAB, SU-DHL10, OCI-Ly1). Bliss δ > 10 = synergy; −10 to 10 = additive; < −10 = antagonism. B) BODIPY C11 oxidation ratio (Ox/non-Ox MFI) by flow cytometry in SU-DHL10 cells treated with zotatifin + ML385 (1–50 μM). Two-way ANOVA with Bonferroni’s test (ns, **P < 0.01, ****P < 0.0001). Mean ± SEM, three experiments. C) Immunoblot for NRF2, HO-1, MYC, GPX4, and cleaved caspase 3 in SU-DHL10 cells treated with 5 μM ML385 and 12.5 nM zotatifin for 24 h. CYPB was used as a loading control. Data are representative of two independent experiments. D) BODIPY C11 oxidation (Ox/non-Ox MFI) in HAP1 EIF4A1 WT or F163L cells treated with zotatifin + erastin (left), RSL3 (center), or DMF (right). Mean ± SEM, three experiments. Two-way ANOVA with Bonferroni’s test (ns; **P < 0.01; ***P < 0.001; ****P < 0.0001). E) Immunoblot of HAP1 EIF4A1 WT (left) or F163L (right) cells treated with erastin (1 μM) ± zotatifin (50 nM) for 24 h. Blots probed for HO-1, cleaved caspase 3, and GPX4. CYPB was used as a loading control. Data from two independent experiments. F) Bliss δ synergy heatmap for zotatifin + erastin, RSL3, or DMF in HAP1 EIF4A1 WT or F163L cells. Bliss δ interpretation as in (A).

Figure 6:. In vivo efficacy and safety of zotatifin plus the ferroptosis inducer IKE.

A) Forty GCB DLBCL PDX tumors were implanted in NSG mice, randomized into four groups (n = 10): control, zotatifin (1 mg/kg IP, 2×/week), IKE (40 mg/kg IP, 5×/week for 13 days), and combination. Dosing was reduced on day 10 to zotatifin (2×/week) and IKE (1×/week). Created in BioRender. Manara, P. (2025) https://BioRender.com/0qob6g5. B) Average body weights ± SEM, measured twice weekly (n = 10). C) Kaplan–Meier survival curves for each group (n = 10). Mantel-Cox test (*P < 0.05, **P < 0.01, ***P < 0.001). Red arrows indicate four mice lost in the combo group due to liver toxicity. D) Average tumor volume ± SEM, measured twice weekly by ultrasound (n = 10). One-way ANOVA with Bonferroni’s test (*P < 0.05, ****P < 0.0001). E) Mean tumor weight at necropsy ± SEM by treatment group (n = 6). One-way ANOVA with Bonferroni’s test (**P < 0.01, ***P < 0.001, ****P < 0.0001 vs. vehicle). F) IHC staining for 4-HNE and cleaved caspase 3 in tumors from vehicle, zotatifin, IKE, or combo groups (20× magnification). G-H Quantification of IHC intensity for G) 4-HNE and H) cleaved caspase-3. Index calculated using Fiji after background subtraction (ROI mean intensity). One-way ANOVA with Bonferroni test (ns, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 vs. vehicle). Mean ± SEM, five experiments.

Figure 7:. Zotatifin sensitizes lymphoma tumors to killing by CD19 CAR-T cells.

A) Bliss δ synergy plots for zotatifin + IFNγ in SU-DHL10 (left) and OCI-Ly1 (right). Bliss δ > 10 = synergy; −10 to 10 = additive; < −10 = antagonism. B) ATP levels in RD114-derived CD28-CD19 CAR T cells at 24 h (left), 48 h (center), and 72 h (right). C) Schematic of co-culture experiment (see Results). D) Co-culture (see schematic in C) of RD114-derived CD28-CD19 CAR T cells as effector (E) cells with SU-DHL10 (left) or OCI-Ly1 (center) target (T) cells, and mouse CAR T cells (E) with A20-Luciferase cells (T, right). Target cells were pretreated with DMSO or 12.5 nM zotatifin and labeled with CFSE. MFI was quantified by flow cytometry. Created in BioRender. Manara, P. (2025) https://BioRender.com/lmedq71. E) IFNγ levels from co-cultures (see schematic in C) of RD114-derived CD28-CD19 CAR T cells with SU-DHL10 (top) or OCI-Ly1 (bottom) at a 1.25:1 E:T ratio. Mean ± SEM of three experiments. Two-way ANOVA with Bonferroni’s test (ns). F) CD3 and IRF1 staining in co-cultures (as in C) at 1.25:1 E:T ratio. Cells were fixed, permeabilized, and analyzed by flow cytometry. Mean ± SEM of three experiments. Two-way ANOVA with Bonferroni’s test (ns, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001). G) BODIPY C11 oxidation ratio (Ox/non-Ox MFI) in SU-DHL10 (top) and OCI-Ly1 (bottom) cells pretreated with DMSO or 12.5 nM zotatifin for 36 h, then IFNγ (0–25 ng/mL) for 8 h. Mean ± SEM of three experiments. Two-way ANOVA with Bonferroni’s test (*P < 0.05). H) CD3 and SLC3A2 staining in co-cultures (as in C) of SU-DHL10 (top) and OCI-Ly1 (bottom) with CAR T cells at 1.25:1 E:T ratio. Mean ± SEM of three experiments. Two-way ANOVA with Bonferroni’s test (ns, *P < 0.05, **P < 0.01, ****P < 0.0001). I) Schematic of in vivo CAR T experiment. 40 BALB/c mice were engrafted with A20-expressing luciferase cells (1×10⁶ cells/mouse) via tail vein injection and randomized into four groups of 10 mice each: vehicle, zotatifin (1 mg/kg, IP. every three days), CAR T, and combo (zotatifin+CAR T). CAR T cells (6×10⁶ cells/mouse) were injected at day 5 into CAR T and combo groups. CAR T cells were collected from spleen 14 days before injection. Created in BioRender. Manara, P. (2025) https://BioRender.com/1xzacsd. J) Total bioluminescence (photons/sec) over time. Mean ± SEM, five experiments. Two-way ANOVA with Bonferroni’s test (*P < 0.05). K) Kaplan–Meier survival curves from five experiments. Mantel–Cox test (ns).