Targeting eIF4A triggers an interferon response to synergize with chemotherapy and suppress triple-negative breast cancer

Protein synthesis is frequently dysregulated in cancer and selective inhibition of mRNA translation represents an attractive cancer therapy. Here, we show that therapeutically targeting the RNA helicase eIF4A with zotatifin, the first-in-class eIF4A inhibitor, exerts pleiotropic effects on both tumor cells and the tumor immune microenvironment in a diverse cohort of syngeneic triple-negative breast cancer (TNBC) mouse models. Zotatifin not only suppresses tumor cell proliferation but also directly repolarizes macrophages toward an M1-like phenotype and inhibits neutrophil infiltration, which sensitizes tumors to immune checkpoint blockade. Mechanistic studies revealed that zotatifin reprograms the tumor translational landscape, inhibits the translation of Sox4 and Fgfr1, and induces an interferon (IFN) response uniformly across models. The induction of an IFN response is partially due to the inhibition of Sox4 translation by zotatifin. A similar induction of IFN-stimulated genes was observed in breast cancer patient biopsies following zotatifin treatment. Surprisingly, zotatifin significantly synergizes with carboplatin to trigger DNA damage and an even heightened IFN response, resulting in T cell–dependent tumor suppression. These studies identified a vulnerability of eIF4A in TNBC, potential pharmacodynamic biomarkers for zotatifin, and provide a rationale for new combination regimens consisting of zotatifin and chemotherapy or immunotherapy as treatments for TNBC.


Mass cytometry and flow cytometry
Trp53-null tumors treated with vehicle or Zotatifin were dissociated in 1 mg/ml Collagenase A for 2 hrs at 37°C with 125 rpm rotation followed by 3 short centrifugations to enrich for tumor stromal cells from supernatants.
Red blood cells were removed using RBC lysis buffer (Biologend #420301).The remaining single cell suspension was analyzed using mass cytometry or flow cytometry.
In mass cytometry, after staining with cisplatin viability dye, cells were surface stained using a cocktail of antibodies conjugated to metal isotopes, fixed, permeabilized using Foxp3 staining buffer set (eBioscience #00552300), and stained with antibodies for intracellular markers.The cells were then stained with Cell-ID Intercalator-Ir (Fluidigm #SKU201192A) overnight at 4°C and analyzed using a Helios CyTOF Mass Cytometer (Fluidigm).The normalized FCS files were first processed using FlowJo to remove beads, dead cells, and doublets.Equal numbers of CD45+ single cells from biological replicates of each treatment group were concatenated and subjected to analysis in Cytobank.Data were dimensionally reduced using viSNE (https://search.r-project.org/CRAN/refmans/CytobankAPI/html/viSNE-class.html) and cell clusters were identified using FlowSOM (https://bioconductor.org/packages/release/bioc/html/FlowSOM.html).
For cell cycle analysis, 2153L cells cultured in regular medium in vitro were allowed to grow to 30% confluence and treated with 40 nM Zotatifin for 48 hrs before trypsinization and fixation in 4% PFA.Next, the cells were pelleted, washed with PBS, and resuspended in DAPI staining buffer (Thermo Fisher #R37606) before flow cytometry.
All flow cytometry data were acquired using an Attune NxT Flow Cytometer (Thermo Fisher) and analyzed using FlowJo software (version 10.7.1).The cell cycle distribution was analyzed using the Watson Pragmatic algorithm provided by FlowJo.

Luminex cytokine analysis
2153L tumor tissues from the same batch with TMT MS were pulverized to powder under liquid nitrogen and lysed in MILLIPLEX MAP Lysis buffer (Millipore #43040).Lysates were cleared by centrifugation at 10,000 g three times at 4°C.The protein concentration of the supernatant was measured using a BCA Protein Assay Kit and adjusted to 1 mg/ml using lysis buffer.The abundance of cytokines was determined using a MILLIPLEX Mouse 32-Plex Cytokine Panel 1 kit (Millipore # MCYTMAG-70K-PX32) according to the manufacturer's instructions.

Separation of tumor-associated macrophage (TAM)
Untreated 2153L and 2151R tumors were dissociated in 1 mg/ml Collagenase A for 2 hrs at 37°C and TAMs were separated using EasySep™ Mouse F4/80 Positive Selection Kit (STEMCELL technologies, #100-0659) following manufacture's protocol.TAMs were cultured in BMDM culture medium and treated with either vehicle or 40 nM Zotatifin for 24 hrs before immunoblotting.
For TMT labeling, 120 μg digested peptide from each sample as well as RefMix, which is a mixture of equal amounts of peptide from each sample, were labeled with the TMT10plex Label Reagent Set (Thermo Fisher #90110) for 1 hr at RT.After confirming the labeling efficiency for each channel using quality control MS runs, the reaction was quenched by adding 5% hydroxylamine (Thermo Fisher #90115) for 15 min at RT.All samples were then combined and freeze dried using SpeedVac.Each plex was reconstituted with 1 ml of 3% ACN/0.1% formic acid and desalted using Sep-Pak Vac 3cc tC18 cartridges (Waters #WAT054925).The elute was dried using SpeedVac.
TMT-labeled peptides were fractioned offline using an Agilent 300Extend-C18 column (4.6 mm X 250 mm, 5 µm) coupled to an Agilent 1260 Infinity II system at 1 ml/min for 96 min.The 96 fractions were concatenated into 24 peptide pools and a flow-through pool and acidified with 0.1% formic acid.The peptides were separated on an online nanoflow Easy-nLC-1200 system (Thermo Fisher) coupled to an Orbitrap Fusion Lumos ETD mass spectrometer (Thermo Fisher).Proteome fractions (1 µg each) were loaded onto pre-column (2 cm x 100 µm I.D.) and separated on in-line 5 cm x 150 µm I.D. column (Reprosil-Pur Basic C18aq, Dr. Maisch) equilibrated with 0.1% formic acid.Peptide separation was performed at a flow rate of 750 nl/min over a 90 min gradient time with different concentrations of solvent B (4-32% for 85 min, followed by 5 min wash at 90% B).
The peptides were ionized at a positive spray voltage of 2.4 kV and the ion transfer tube temperature was set at 300 °C.The mass spectrometer was operated in data-dependent mode with 2 sec cycle time.MS1 was acquired in Orbitrap (60000 resolution, scan range 350-1800 m/z, AGC 5e5, 50 ms injection time), followed by MS2 in Orbitrap (50000 resolution, AGC 1e5, 105 ms injection time, HCD 38%).Dynamic exclusion was set to 20 sec and the isolation width was set to 0.7 m/z.
To process the proteomics data, raw files were converted to mzML using MSConvert (2).MASIC (3) was used to calculate precursor ion intensities (derived from the area under each elution curve) and to extract reporter ion intensities using default high resolution MS parameters.The Butterworth smoothing method was used with a sampling frequency of 0.25 and an SIC tolerance of 10 ppm.Reporter ion tolerance was set to 0.003 Da with reporter ion abundance correction enabled.Raw spectra were searched with MSFragger (v3.2) using both mass calibration and parameter optimization (4,5).Peptide validation was performed using a semi-supervised learning procedure in Percolator (6) as implemented in MokaPot (7).The peptides were grouped and quantified into gene product groups using gpGrouper (8).Only the gene products identified in both TMT multiplexes were retained for downstream analyses.Samples were first normalized to the internal reference within each TMT multiplex and then normalized by their median peptide abundance before subsequent data analyses.

Gene set enrichment analysis (GSEA)
GSEA (v3.0) was performed using hallmark gene sets (v7.0) from the Molecular Signature Database (MSigDB) using default settings after mapping mouse genes to their human homologs using the HomoloGene system.Genes without mapping were excluded, and the median value was taken when multiple mouse genes mapped to a single human gene.Pathway enrichment P values were calculated using gene set permutation.

Quantitative real-time PCR (qPCR)
Total RNA was extracted using TRIzol reagent (Thermo Fisher #15596026) following the manufacturer's instructions.Total RNA (1 µg) was converted to cDNA using the High-Capacity cDNA Reverse Transcription Kit (Thermo Fisher #4368814).The mRNA levels were detected using amfisure qGreen Q-PCR Master Mix (GenDEPOT #Q5602).18S was used as the internal reference gene for patient biopsy samples and Actb was used as the internal reference for all other samples.The levels of target genes were normalized to those of internal reference gene to calculate the 2 -ΔΔCt value.The sequences of all the qPCR primers are listed in Supplementary Table S6.

Supplementary Figure 4 .Supplementary Figure 5 .
treated with CHX or MG132 for different time periods.F and G, QPCR analysis for SOX4 (F) and FGFR1 (G) in BT549 cells that were treated with 40 nM Zotatifin for different time periods.H, Immunoblotting analysis of 2153L cells that were treated with indicated drugs for 24 hrs.CPT, camptothecin.I, QPCR analysis of HAP1 cells that were treated with 40 nM Zotatifin for 6 hrs.In C, D, F, G, and I, data are representative of two independent experiments and are presented as mean ± SD of technical triplicates.J, Distribution of Actb and Gapdh mRNAs across the different fractions in polysome profiling analysis of 2153L cells that were treated with vehicle or 40 nM Zotatifin for 2 hrs.Data are presented as mean ± SEM of three biological replicates.K, Immunoblotting analysis of 2153L cells that were treated with vehicle or Zotatifin in the presence of CHX or MG132 for 2 hrs.Data are representative of two independent experiments.Zotatifin induces interferon response genes.A, QPCR analysis of 2153L cells that were treated with different concentrations of Zotatifin for 6 hrs.B, QPCR analysis of 2153L cells that were treated with 40 nM Zotatifin for different time periods.In A and B, data are representative of two independent experiments and are presented as mean ± SD of technical triplicates.C, QPCR analysis of 2153L cells that were treated with indicated drugs for 24 hrs.Data are presented as mean ± SD of technical triplicates.D, QPCR analysis of paired ER+ breast cancer biopsies from pre-treatment (pre) and on Zotatifin treatment (on) patients.The mRNA levels of pre-treatment samples were set as 1 and fold changes were calculated for each paired sample.Data are analyzed using two-tailed unpaired Student's t-test.n=8 patient biopsy pairs.Sox4 inhibition by Zotatifin contributes to Zotatifin induced interferon response genes.A, The mRNA levels of Eif4a1, Eif4a2, and Ddx3x in normal mammary glands of BALB/c mice and Trp53-null preclinical models.The RNA levels for each gene were averaged RNA-seq signals from 1-12 biological replicates.B, The Chronos dependency scores of EIF4A1 and EIF4A2 in CRISPR knockout screens.A lower Chronos score indicates higher essentiality.A score of 0 indicates a gene is not essential and a score of -1 is the median scores of all pan-essential genes.C, The RNA and protein levels of EIF4A1, EIF4A2, and A Figure 2I