Immunosuppressive Myeloid Cells Induce Nitric Oxide–Dependent DNA Damage and p53 Pathway Activation in CD8⁺ T Cells

Microbial infections result in heightened production of myeloid cells that leave the bone marrow before they have fully matured, a process referred to as “emergency hematopoiesis” (1, 2). Human cancers also substantially alter hematopoiesis through release of key regulators of myelopoiesis (including GM-CSF, G-CSF, and M-CSF) and proinflammatory mediators (including IL6, IL1β, and TNFα; refs. 3, 4). Such alterations in myelopoiesis result in accumulation of immature myeloid cells with a pathologic activation state in the circulation, lymphoid organs, and tumors. These immature myeloid cells are collectively referred to as myeloid-derived suppressor cells (MDSC), and the number of circulating MDSCs represents an independent indicator of poor survival outcomes in a number of human cancers (5–7). In both humans and mouse models, MDSCs include populations of monocytic (M-MDSC) and polymorphonuclear (PMN-MDSC) that can be distinguished based on morphologic characteristics and cell-surface markers (8). MDSCs potently inhibit T cell–mediated tumor immunity and circulating MDSCs negatively correlate with clinical responses to checkpoint blockade (9).

In vitro and in vivo studies have demonstrated that MDSCs potently inhibit T-cell proliferation (10) through multiple mechanisms (Supplementary Table S1; refs. 11–19). A number of studies have shown that the release of nitric oxide (NO) by MDSCs is an important contributor to inhibition of T-cell proliferation (20). NO is synthesized by inducible nitric oxide synthase (iNOS) in MDSCs and NO readily diffuses across cell membranes (10, 20). The production of superoxide by NADPH oxidase (Nox2) has also been implicated in inhibition of effector T-cell function (21, 22). NO can react with superoxide, resulting in the formation of peroxynitrite. It was previously proposed that peroxynitrite inhibits the first steps of T-cell activation through nitrosylation of T-cell receptor (TCR) and CD8 molecules (23). Other mechanisms implicated in the inhibition of T-cell function by MDSCs include the depletion of essential amino acids (including L-arginine, L-tryptophan, and L-cystine), release of inhibitory cytokines (such as IL10), and expression of inhibitory surface ligands (including PD-L1; refs. 17, 24–27).

The present study was designed to further interrogate the molecular changes that occur in T cells in the presence of MDSCs. Single-cell RNA-sequencing (scRNA-seq) analysis demonstrated that T cells cocultured with dendritic cells (DC) and MDSCs initially acquired transcriptional features of an activated state. However, the presence of MDSCs induced a p53 transcriptional signature and DNA damage in T cells at later time points, resulting in T cell apoptosis. DNA damage was prevented by inactivation of the Nos2 gene in MDSCs, implicating production of NO by iNOS in this process. Furthermore, KPC pancreatic tumors implanted into Nos2-deficient mice were infiltrated by a greater number of CD8⁺ T cells, and DNA damage was substantially reduced in tumor-infiltrating CD8⁺ T cells from Nos2-deficient mice compared with those in wild-type mice.

Flow cytometry: Hoechst 33342 (Thermo; #R37165); Vybrant CFDA SE Cell Tracer (Thermo; #V12883); CellTrace CFSE Cell Proliferation Kit (Thermo; #C34553); Zombie UV (BioLegend; #423108); Zombie Violet (BioLegend; #423114); TruStain FcX (BioLegend; anti-mouse CD16/CD32; Clone 93; #101320); Annexin V–Pacific Blue (Thermo; #A35122); CD8α-BV650 (BioLegend; Clone 53-6.7; #100741); CD8α-PE (BioLegend; Clone 53-6.7; #100708); CD69-PE (BioLegend; Clone H1.2F3; #104508); CD25-APC-Cy7 (BioLegend; Clone PC61; #102026); IFNγ-PE (BioLegend; Clone XMG1.2; #505808); TNFα-PE-Dazzle594 (BioLegend; Clone MP6-XT22; #506346); CD11b-APC (BioLegend; Clone M1/70; #101262); Gr-1-PE (BioLegend; Clone RB6-8C5; #108408); Ly6G-FITC (BioLegend; Clone 1A8; #127606); Ly6G-PE-Cy7 (BioLegend; Clone HK1.4; #128018); Ly6C-PerCP-Cy5.5 (BioLegend; Clone HK1.4; #128028); Ly6G-FITC (BioLegend; Clone 1A8; #127606); CD45-PerCP-Cy5.5 (BioLegend; Clone 30-F11; #103130); CD3-BV786 (BioLegend; Clone 17A2; 100232); CD3-BV421 (BioLegend; Clone 17A2; #100228), NK1.1-APC (BioLegend; Clone PK136; 108710); CD107a-BV421 (BioLegend; Clone 1D4B; #121618); Granzyme B-APC (Thermo; Clone NGZB; #17-8898-82).

Western blotting: anti-mouse p53 (Rat IgG2b; Clone 19764; R&D Systems; #AF1355); anti-phospho-p53 (S15; Abcam; Polyclonal; #ab1431); goat anti-rat IgG-HRP (H+L; Thermo; Polyclonal; #31470); goat anti-rabbit IgG-HRP (Cell Signaling Technologies; Polyclonal; #7074).

Neutralization: anti-mouse IL10 (Bio X Cell; Clone JES5-2A5; #BE0049); anti-mouse PD1 (CD279; Bio X Cell; Clone 29F.1A12; #BE0273).

Microscopy: anti-γH2AX-AF488 (Abcam; Clone EP854(2)Y; #ab195188); anti-phospho-p53 (S15; Abcam; Polyclonal; #ab1431); anti-rabbit IgG-AF488 (Abcam; Polyclonal; #ab150077).

N-acetyl-L-cysteine (NAC; Sigma; #A7250) was prepared in PBS to a working concentration of 1.25 mmol/L. N^G-monomethyl-L-arginine (L-NMMA; Sigma; #M7033) was prepared in PBS to a working concentration of 3.125 mmol/L. Inhibitors were added to cultures prior to seeding of cells to final concentrations as indicated in the “Results” section. L-Arginine (Sigma; #A5006) stock solution was prepared at 5 g/L (5 × stock) in R10 medium [RPMI-1640 (Thermo; #11875119), 10% fetal bovine serum (Thermo; #16000044), 2 mmol/L L-glutamine (GlutaMAX; Thermo; #35050061), 100 μg/mL streptomycin, and 100 μg/mL penicillin (penicillin–streptomycin; Thermo; #15070063)], and the pH was adjusted to 7.0. L-Arginine was serially diluted in R10 medium and added to coculture experiments at final concentrations of 0.2 g to 1.0 g/L.

C57BL/6 (000664-BJ; B6), C57BL/6-Tg (TcraTcrb)1100Mjb/J (003831; OT-I), and B6.129P2-Nos2<tm1Lau>/J (002609; Nos2^−/−) were purchased from The Jackson Laboratories. Experiments in mice were performed in accordance with protocols approved by the Institutional Animal Care and Use Committee (IACUC) at Dana-Farber Cancer Institute.

CD8⁺ T cells were isolated from the spleens of OT-I mice by magnetic negative selection (Murine CD8⁺ negative selection kit; StemCell; #19853). All experiments (except where stated) were performed in R10 medium.

Bone marrow–derived DCs: Bone marrow cells (BMC) were flushed from the tibias and fibulas of C57BL/6 mice using sterile PBS. Red blood cells were lysed (1 minute at room temperature) using Red Cell Lysis Buffer (Thermo; #A1049201) and washed with R10 medium and resuspended at 2 × 10⁷ cells/mL. During washing, 10 mL R10 medium supplemented with 10 ng/mL GM-CSF (BioLegend; 576306) was added to 10-cm Petri dishes. Once BMCs were resuspended, 5 × 10⁶ cells were added per plate (dropwise into the center of the dish), incubated at room temperature for 15 minutes to allow to settle, then incubated at 37°C 5% CO₂ for 3 days. On day 3, 10 mL of GM-CSF supplemented prewarmed R10 medium was gently added followed by culture for another 3 days. On day 6, 90% of the medium was aspirated, replaced with R10 medium supplemented with 10 ng/mL GM-CSF, 1 μg/mL LPS (Escherichia coli O111:B4, Sigma-Aldrich; #LPS25), and 1 nmol/L SIINFEKL peptide (Invivogen; #vac-sin), and BM-DCs were then incubated overnight. On day 7, BM-DCs were gently lifted using a cell scraper, washed (300 × g, 5 minutes), and resuspended at a final density of 1 × 10⁵ cells/mL in R10 medium.

Bone marrow–derived MDSCs: BMCs were flushed from the tibias and fibulas of C57BL/6 mice using sterile PBS. Red blood cells were lysed (1 minute at RT) using Red Cell Lysis Buffer, washed with R10 medium, then resuspended at 2 × 10⁷ cells/mL. During washing, 10 mL R10 medium supplemented with 20 ng/mL GM-CSF and IL6 (BioLegend; #575706) was added to 10-cm tissue culture plates. Once BMCs were resuspended, 1 × 10⁷ cells were added per plate (dropwise into the center of the dish then gently swirled around the plate) and incubated at 37°C 5% CO₂ for 3 days. On day 3, 10 mL of GM-CSF/IL6 supplemented R10 medium was added and incubated overnight. On day 4, BM-MDSCs were gently lifted using a cell scraper, washed (300 × g, 5 minutes), and resuspended at a final density of 1 × 10⁶ cells/mL. Typical composition of cultures was ∼75% MDSC (of total viable) comprising ∼60% PMN-MDSCs and ∼30% M-MDSC (% of total MDSCs). For experiments using specific MDSC subsets, cells were stained using anti-CD11b-APC, anti-Gr-1-PE, anti-Ly6G-PE-Cy7, and anti-Ly6C-FITC. Defined cell populations were isolated by FACS as described in the “Flow cytometry and FACS” section.

KPC tumors were cultured, implanted, and harvested as described in the “Tumor cell lines and tumor models” section. After preparation of a single-cell suspension, samples were stained using a live/dead marker (1:1,000; Zombie Violet) and anti-murine CD16/32 (TruStain FcX; 1:100) as an Fc receptor block for 15 minutes at 4°C. Samples were then washed with fluorescence-activated cell sorting (FACS) buffer and stained with anti-CD45-PerCP-Cy5.5, anti-CD11b-APC, and anti-Gr-1-PE (all 1:100; 30 minutes at 4°C). After staining, cells were washed using FACS buffer and resuspended in FACS buffer supplemented with 0.5 mmol/L EDTA (final concentration) and 100 μg/mL DNase I (Sigma; #11284932001). Cells were then sorted into R10 medium supplemented with 200 ng/mL GM-CSF and IL6, washed using R10 medium (300 × g, 5 minutes), and resuspended at 1 × 10⁶ cells/mL for coculture assays. Sorted cells were assessed by flow cytometry to confirm purity of >90% for CD11b⁺Gr-1⁺ cells.

In vitro coculture assays were performed using 96-well U-bottom plates. OT-I T cells, SIINFEKL-pulsed DCs, and MDSCs were cocultured at a ratio of 1 (T cells) to 0.2 (DCs) to 2 (MDSCs), unless otherwise stated. DCs, OT-I T cells, and MDSCs were resuspended to 1 × 10⁶, 1 × 10⁵, and 1 × 10⁶ cells/mL, respectively, and 100, 50, and 100 μL of each cell type was plated in respective order, resulting in a total of 5 × 10⁴, 1 × 10⁴, and 1 × 10⁵ cells/well, respectively. In DC-only conditions, 100 μL R10 medium was added to maintain a total well volume of 250 μL. Where inhibitors/treatments were used, working concentrations (detailed in the “Small-molecule inhibitors and coculture supplements” section) or solvent controls were added to the wells prior to plating of cells. After addition of cells, cocultures were mixed gently using a multichannel pipette, centrifuged at 300 × g for 30 seconds and then incubated for indicated times at 37°C 5% CO₂.

T-cell proliferation was assessed by labeling of CD8⁺ T cells with CFSE (CellTrace); equal volumes of CFSE (2 μmol/L in PBS) and 1 × 10⁷ CD8⁺ T cells/mL in PBS were mixed, the reaction was incubated for 1 minute in a 37°C water bath, and cells were then washed twice with TCM. CFSE dilution by T cells was analyzed by flow cytometry at different time points (24–96 hours, as noted in the respective figure legends).

MDSC cultures were prepared on day 4 (d4), day 3, and day 2 before coculture so that d4 MDSCs could be added to cocultures at different time points. Coculture assays were prepared as described in the “MDSC suppression coculture assay” section, where conditions with MDSC spike-in at 24 or 48 hours were given 100 μL R10 in place of MDSCs. On the day of MDSC addition, 100 μL of supernatant was gently aspirated and MDSCs or R10 medium (to compensate for supernatant nutrient changes) were gently pipetted into the relevant wells. All cocultures were then harvested 96 hours from the start of initial cocultures for flow-cytometric analysis.

Cells for cocultures were prepared as described in the “MDSC suppression coculture assay” section for “DC-stimulated condition.” For the “Cytokine-stimulated condition,” OT-I CD8⁺ T cells and MDSCs were added to wells containing 100 μg/mL IL15 (BioLegend; #566304) and 5 μg/mL IL7 (BioLegend; #577806). Once plated, samples were mixed gently with a multichannel pipette, centrifuged at 300 × g for 30 seconds, and then incubated for the indicated times at 37°C 5% CO₂.

MDSCs were cultured as described in the “Isolation of cell populations for coculture experiments” section, and subsets were sorted as described in the “Flow cytometry and FACS” section. After sorting, MDSCs and M-MDSCs were plated into 12-well plates (300,000 cells/well) and stimulated with either 1 ng/mL IFNγ (BioLegend; 575306), 1 ng/mL TNFα (BioLegend; #575206), or both cytokines (1 ng/mL each) for 24 hours. Expression of Nos2 was assessed by qRT-PCR as described in the “Quantitative RT-PCR” section.

Apoptosis was assessed at the 72-hour time point for coculture assays. Samples were first washed with PBS (300 × g, 5 minutes) and then with Annexin Binding Buffer (Thermo; #V13246; 300 × g, 5 minutes). After washing, cells were resuspended in 100 μL Annexin Binding Buffer containing Zombie UV (1:1,000), Annexin V–Pacific Blue (1:250), and anti-CD8α-BV650 (1:250). Samples were incubated at 37°C for 15 minutes, followed by addition of 150 μL of Annexin Binding Buffer and analysis by flow cytometry. Early and late stages of apoptosis were determined using Zombie UV/Annexin V staining: early apoptosis (Zombie UV⁻Annexin V⁺) and late apoptosis (Zombie UV⁺Annexin V⁺).

Flow cytometry analysis was performed using a BD LSRFortessa instrument. All samples were washed (300 × g, 5 minutes) and stained using FACS buffer (4% FBS in PBS). All samples were stained at an antibody dilution of 1:150 for 45 minutes at 4°C, then washed (300 × g, 5 minutes) and resuspended in FACS buffer (4% FBS in PBS) for analysis/sorting. For intracellular staining, Cytofix/Cytoperm (BD) was used as per the manufacturer's instructions.

Cell sorting was performed using a BD Aria III instrument. Cells were sorted into complete RPMI medium in 15 mL falcon tubes maintained at 4°C. Immediately after sorting, cells were pelleted (300 × g, 5 minutes) and processed for downstream analysis as appropriate for each experiment. Cells were stained using a live/dead marker (1:1,000; Zombie Violet, BioLegend) and anti-murine CD16/32 (1:100; BioLegend) for 15 minutes at room temperature. Cells were then washed and stained using relevant antibodies for 30 minutes at 4°C. After staining, cells were washed using FACS buffer and resuspended in FACS buffer supplemented with 0.5 mmol/L EDTA and 0.1 mg/mL DNase I. All cells were sorted into R10 medium (supplemented as indicated). Once sorted, cells were washed with R10 medium (300 × g, 5 minutes) and prepared for downstream applications as described in the relevant sections. All antibodies used are detailed in the “Antibody and cell staining reagents” section.

CD8⁺ T cells were sorted from cocultures by FACS and then washed with PBS. RNA was isolated from CD8⁺ T cells using a Direct-zol RNA Miniprep Plus kit (Zymogen; #R2071) per the manufacturer's instructions and eluted into nuclease-free H₂O. RNA purity and concentration were determined using a DS-11 spectrophotometer (Denovix). For reverse transcription, 350 ng of RNA was used to transcribe cDNA using Superscript VILO (Thermo) per the manufacturer's instructions, and then cDNA was diluted 5- to 10-fold in ddH₂O. The following assay probes (conjugated to 6-carboxyfluorescein) were purchased from Thermo: Rps18 (housekeeping control; Mm02601777_g1), Ccng1 (Mm00438084_m1), Cdkn1a (Mm04205640_g1), Phlda3 (Mm00449846_m1), Bax (Mm00432051_m1), Mdm2 (Mm01233138_ m1), and Nos2 (Mm00440502_m1). Quantitative RT-PCR reactions were prepared using TaqMan Gene-Expression Master Mix (Thermo; #4369610) per the manufacturer's instructions with a final volume of 10 μL per reaction using 4.5 μL diluted cDNA. All RT-PCR reactions were performed using technical triplicates on biological triplicates. RT-PCR was performed using a QuantStudio 6 (Thermo). Mean Ct values were then used to calculate ΔΔCt values to compare gene expression.

Nitrite quantification was performed using a Griess Reagent Kit (Thermo; #G7921) per the manufacturer's instructions. After 72 hours of coculture, plates were centrifuged (300 × g, 5 minutes), and 200 μL supernatant was collected for this assay.

KPC tumor cells were a kind gift from Dr. Stephanie Dougan (Dana-Farber Cancer Institute). Cells were authenticated in 2017 by transcriptional sequencing. Cells were confirmed to be Mycoplasma-free by both Charles River Laboratories Mouse Essential CLEAR panel prior to gifting and by the Universal Mycoplasma Detection Kit (Fisher Scientific; #50189644FP) prior to expansion and storage in liquid nitrogen. KPC cells were cultured in DMEM (Thermo; #11054001) supplemented with 10% fetal bovine serum, 100 μg/mL streptomycin, and 100 μg/mL penicillin. For each experiment, KPC cells were grown to a confluency of 60% to 80% and passaged two times prior to implantation. For passaging, cells were dissociated using TrypLE Express Enzyme (Thermo; #12604039) per the manufacturer's instructions. Cell dissociation was stopped by addition of KPC culture medium, and cells were then washed three times with sterile PBS. KPC cells were resuspended in PBS to a final cell density of 6 × 10⁶ cells/mL. Mice were shaved and anesthetized with isoflurane prior to implantation of KPC cells (3 × 10⁵ cells in 50 μL per mouse) into the subcutaneous layer of the skin along the left flank. Tumors were harvested after 21 days.

Tumors were excised and placed into 10-cm tissue culture dishes containing 10 mL of ice-cold RMPI-1640. Tumors were then diced using scalpels to <2 mm pieces and washed with RPMI-1640 (300 × g, 5 minutes). Tumors were then resuspended in 9 mL of tumor dissociation solution [RPMI-1640 containing 1 mg/mL collagenase type IV from Clostridium histolyticum (Sigma, #C5138), 20 U/mL DNase type IV from bovine pancreas (Sigma; #D5025) and 0.1 mg/mL hyaluronidase type V from sheep testes (Sigma; #H6254)] and incubated at 37°C with shaking at 220 rpm for 45 minutes. Digested tumors were then passed through a 70-μm cell strainer, washed with RPMI-1640 (300 × g, 5 minutes), and cells were resuspended in 5 mL ACK lysis buffer followed by incubation at RT for 5 minutes. Cells were then washed twice with PBS (300 × g, 5 minutes) before downstream applications.

CD8⁺ T cells were isolated by FACS from cocultures and plated on 0.01% poly-L-lysine (Sigma; #A-005-C) coated 8-well chamber LabTek slides (Nunc); cells were incubated at 4°C for 20 minutes to allow them to adhere to the slides. Cells were then washed twice with PBS-Tween buffer [PBS supplemented with 0.05% Tween-20 (Sigma; #P1379); 200 μL per wash, removal by flicking] and then fixed at 4°C for 20 minutes in 200 μL Cytofix/Cytoperm buffer (BD; #554722). Cells were washed twice with PBS-Tween buffer and then permeabilized using PBS containing 0.05% Triton-X-100 (Sigma; #X100) and 1% BSA (Sigma; #A1933) for 20 minutes at room temperature. Following permeabilization, cells were washed twice with PBS supplemented with 1% BSA, 0.05% Tween-20, and slides were blocked for 30 minutes at room temperature using 10% normal goat serum (Thermo; #50062Z) supplemented with 5% FBS. Cells were then washed with PBS-Tween buffer and stained at 4°C overnight with anti-γH2AX-AF488 or anti-phospho-p53 (S15) diluted 1:200 in staining buffer (PBS containing 0.01% Tween 20 and 1% BSA). For detection of phospho-p53, cells were washed three times with PBS-Tween buffer, then stained with secondary Ab goat anti-rabbit IgG-AF488 at RT for 1 hour (1:200 in PBS containing 1% BSA 0.05% Tween-20). After staining, cells were washed twice with PBS and then stained using Hoechst-33342 (1:1,000 in PBS) for 5 minutes at room temperature, followed by two washes with PBS-Tween. After washing, cells were covered with 200 μL of PBS for imaging.

Imaging was performed using a Leica SP5X confocal microscope with a 63 × oil immersion lens (NA = 1.4) and 2–3 × digital zoom. Both γH2AX and phospho-p53 were imaged using an accumulation average of 4. Data were analyzed using ImageJ software. To determine mean intensity of fluorescence signal per cell, the integrated intensity of each cell was measured using the default threshold mask. For counting of γH2AX foci, a negative control sample of freshly isolated CD8⁺ T cells was stained and imaged as described above, and the maximum intensity value of negative samples was used to set a threshold minimum value. These thresholds were then applied to sample images and a threshold mask was applied to each image. Images were then converted to binary and particle measurement parameters which were set to >2 pixels and a circularity 0–1.00. Foci identified using these parameters were then recorded. For cells positive for γH2AX, cells with >1 foci were counted. For phospho-p53–positive cells, any cells with staining above background were counted.

FACS-purified CD8⁺ T cells were lysed (30 μL per 1 × 10⁶ cells) in Pierce RIPA buffer (Thermo; #89901) supplemented with Halt Protease and Phosphatase Inhibitor Cocktail (Thermo; #78444) at a 1 × final concentration. Lysates were incubated on ice for 30 minutes, sonicated for 20 seconds, incubated on ice for a further 10 minutes, then centrifuged at >20,000 × g for 15 minutes. Lysates were stored at × 80°C prior to use and defrosted by centrifugation (4°C, 20,000 × g, 15 minutes). Protein concentration was determined by BCA assay (Thermo; #A53225) and equal amounts of protein were prepared. Lysates were mixed with 4X SDS Loading Buffer (Thermo; #NP0007) supplemented with 60 mmol/L DMSO (Sigma; #472301) and heated at 95°C for 20 minutes. Samples were then pulsed in a centrifuge and loaded onto NuPAGE 8% Bis-Tris gels (Thermo; #WG1002BOX) which were run with 1 × MOPS running buffer (Sigma; M1254) at 150 V for 100 minutes. For molecular-weight approximation, 4 μL of SeeBlue2 ladder (Thermo; LC5925) was loaded. Protein was then transferred onto PVDF membranes in Towbin buffer (25 mmol/L Tris, 192 mmol/L glycine, pH 8.3) at 100 V for 100 minutes at 4°C. After transfer, membranes were blocked in 5% Blotting Grade Blocker (Bio-Rad; #1706404) for 30 minutes, washed with 1 × TBS-T buffer, and probed using relevant antibody and blocking solution (indicated in the “Antibody and cell staining reagents” section) at 4°C overnight. Membranes were then washed extensively (at least 3 washes for 5–10 minutes) with 1 × TBS-T and then probed with relevant HRP-conjugated secondary antibody in 5% Blotting Grade Blocker for 1 hour at room temperature. Membranes were then washed extensively (at least 3 washes for 5–10 minutes) with 1 × TBS-T and exposed using ∼800 μL Western Lightning Plus-ECL (PerkinElmer; #NEL105001EA) on a ChemiDoc MP Imaging System (Bio-Rad).

For scRNA-seq, naïve CD8⁺ T cells were isolated by negative selection as described in the “Cells and cell culture” section; CD8⁺ T cells from coculture assays were isolated by FACS at 24 and 72 hours as described in the “Flow cytometry and FACS” section. Immediately following isolation, cells were washed in PBS containing 0.04% RNase-free BSA. A total of 10,000 cells from each condition were targeted for 10X Genomics 3′ V2 single-cell assay (10X Genomics). Reverse transcription, cDNA amplification, and library preparation were performed per the manufacturer's instructions. All libraries were sequenced using an Illumina HiSeq 2500 on rapid-run mode, yielding ∼31,000 to 55,000 reads per cell.

Demultiplexing, alignment to the mm10 mouse genome, and UMI counting were performed using Cell Ranger (version 2.0.1, 10X Genomics). This included read alignment and barcode demultiplexing, followed by UMI and barcode filtering and correction. The resulting gene-expression matrix was then filtered to exclude genes with less than one UMI count in at least one cell. UMI counts were normalized by dividing the raw counts by the total counts in each cell. Subsequent clustering and visualization were conducted on a reduced expression matrix, retaining only the 1,000 genes exhibiting the highest dispersion values (defined as the variance over the mean expression). For each remaining gene, UMI counts were log-transformed and z-score was normalized. To filter single cells with low quality, we adopted the following steps: (i) building a median-centered median absolute deviation (MAD)–variance normal distribution based on the fraction of mitochondrial genes of all single cells; removal of single cells with a mitochondrial fraction larger than the median of this distribution (28); (ii) removal of single cells with fewer than 200 detected genes; and (iii) use of DoubletFinder (29) to filter potential doublets (set the doublet ratio as 7.5%). Following these steps, the data from different time points and conditions were integrated and analyzed together; this included 4,050 cells from 0 hours, 2,208 cells from 24-hour DC, 3,020 cells from 24-hour DC + MDSC, 2,260 cells from 72-hour DC, and 5,256 from 72-hour DC + MDSC cultures. Only genes that were expressed in more than 10 single cells were included. In order to focus on T-cell populations, additional filtering was performed to obtain cells that had at least five UMI counts of CD3E, CD8A, or CD8B1, after which 13,629 single cells remained. The Seurat (30, 31) package in R was then used for analyses. The data were then normalized and scaled using the “SCTransform” (32) function in Seurat and 2,000 highly variable genes were selected and top 30 principal components were used. Finally, shared nearest neighbor was applied for clustering, and uniform manifold approximation and projection (UMAP) was used to visualize the single-cell transcriptional profile in 2D space. Differentially expressed genes between different time points, conditions, or clusters were identified by “FindAllMarkers” function (use default parameter but set min.pct as 0.3) in Seurat. The subclustering analysis was performed using the same strategy described above.

The Gene Expression Omnibus accession number for the scRNA-seq data reported in this article is GSE145396.

All statistical analyses were performed using Prism 8 (GraphPad). P value threshold was set at 0.05. All corrections for multiple comparisons are detailed in the figure legends.

This study was designed to investigate the mechanisms by which MDSCs interfere with T-cell function in the presence of DCs presenting a cognate peptide. MDSCs were either generated in vitro from bone marrow (Supplementary Fig. S1A) or isolated from KPC tumors (Supplementary Fig. S1B), a model of pancreatic cancer. We directly compared cocultures of CD8⁺ T cells (OT-I T cells specific for Ova peptide) with DCs presenting the relevant high-affinity peptide (SIINFEKL) to cultures with DCs + MDSCs. ScRNA-seq was used to study the impact of MDSCs on the transcriptome of CD8⁺ T cells, and the identified pathways were further investigated by confocal microscopy and other techniques (Fig. 1A). Coculture of OT-I CD8⁺ T cells with peptide-pulsed DCs induced potent T-cell activation, as evidenced by proliferation and expression of the activation markers CD25 and CD69 (Fig. 1B–E; Supplementary Fig. S1C–S1E). The addition of MDSCs to these cultures potently inhibited T-cell proliferation (Fig. 1B and C; Supplementary Fig. S1C), but these T cells expressed a high level of the CD25 and CD69 activation markers at 24 hours (Fig. 1B, D, and E; Supplementary Fig. S1D and S1E), indicating that the MDSCs did not block essential early steps of T-cell activation in this system. In cocultures with DCs + MDSCs, T-cell numbers were greatly reduced at the 72-hour time point (∼25-fold; Fig. 1F) and labeling with the cell death marker Zombie UV demonstrated substantially increased cell death (Supplementary Fig. S1F and S1G).

MDSCs include populations of PMN-MDSCs and M-MDSCs. To determine the individual contribution of these subpopulations to inhibition of CD8⁺ T-cell proliferation and survival, we sorted PMN-MDSCs (Ly6G^hiLy6C^lo) and M-MDSCs (Ly6C^hiLy6G^lo; Supplementary Fig. S1A and S1H) and examined their inhibitory activity. Given that the relative frequencies of PMN-MDSCs to M-MDSCs in total MDSC cultures were 2:1, we adjusted the ratios of each subset accordingly for the relevant coculture condition. We found that M-MDSCs potently inhibited CD8⁺ T-cell proliferation (Fig. 1G–J) and survival (Fig. 1K and L). M-MDSCs reduced T-cell numbers at 72 hours to a similar extent as the total MDSC population (Fig. 1G). Labeling with Annexin V and Zombie UV showed that M-MDSCs induced substantial apoptosis of CD8⁺ T cells in cocultures (Fig. 1K and L). We also tested the suppressive capacity of MDSCs at lower T cell to MDSC ratios (Fig. 1M and N) and found that MDSC-mediated inhibition of CD8⁺ T-cell proliferation was directly correlated with MDSC numbers per culture. Similar findings were made with MDSCs directly isolated from KPC tumors (Supplementary Fig. S1B; Fig. 1O–R): CD8⁺ T-cell proliferation was greatly diminished in cocultures with DCs + MDSCs (Fig. 1O–Q), and CD8⁺ T-cell numbers were diminished at 72 hours of coculture (Fig. 1R). These data demonstrate that monocytic MDSCs potently inhibit T-cell proliferation and survival even in the presence of DCs presenting a high-affinity peptide.

To investigate whether MDSCs could inhibit T-cell proliferation following initial activation of T cells by DCs, we set up coculture experiments in which MDSCs were spiked into cultures of OT-I T cells and peptide-pulsed DCs after a delay of 24 or 48 hours. We found that T-cell proliferation and accumulation was attenuated even when MDSCs were added with a delay (Fig. 2A–C).

It has been proposed that MDSCs inhibit early signaling events downstream of the TCR and coreceptors (23, 33). We therefore tested whether MDSCs could inhibit CD8⁺ T-cell proliferation in response to the cytokines IL2 and IL15. MDSCs indeed inhibited cytokine-mediated T-cell proliferation and reduced T-cell numbers at 72 hours, although not to the same extent as cocultures with peptide-pulsed DCs + MDSCs (Fig. 2D–F). These data demonstrate that MDSCs inhibit T-cell proliferation and accumulation even following the initial activation by DCs.

We performed scRNA-seq analysis of CD8⁺ T cells sorted at 24 or 72 hours from cocultures with DCs or DCs + MDSCs to study how MDSCs affect the transcriptional programs of CD8⁺ T cells. CD8⁺ T cells from 24-hour cocultures with DCs and DCs + MDSCs tended to cocluster in a global analysis of the scRNA-seq data (Fig. 3A; Supplementary Fig. S2A); this 24-hour cluster was clearly distinct from the cluster representing input naïve CD8⁺ T cells. In contrast, at the 72-hour time point, CD8⁺ T cells from DCs versus DCs⁺ MDSCs were present in separate clusters.

Subclustering of CD8⁺ T cells from the 24-hour time point was performed to study potential differences of the transcriptional programs in T cells from DC versus DC + MDSC cocultures (Fig. 3B–D; Supplementary Fig. S2A). A total of seven subclusters were identified, and CD8⁺ T cells from DC + MDSC cocultures were underrepresented in two of these clusters (clusters 2 and 6). However, CD8⁺ T cells from cultures with DCs + MDSCs were well represented in other clusters, including cluster 0 (differential expression of cytotoxicity markers Gzmb and Eomes compared with other clusters), cluster 1 (differential expression of genes associated with DNA replication, including Mki67 and Cdk1), and cluster 3 (differential expression of chemokine and cytokine genes, including Ccl4, Il2, and Ifng; Fig. 3C–F). Consistent with these data, flow-cytometric analysis showed a similar percentage of CD8⁺ T cells expressing granzyme B and IFNγ at the 24-hour but not the 72-hour time point in DC + MDSC cocultures compared with DC cocultures (Fig. 3G; Supplementary Fig. S2B–S2D). Gene set enrichment analysis (GSEA) of KEGG pathways demonstrated activation and effector-related programs for all these clusters (Supplementary Fig. S2E). These results indicate that MDSCs did not inhibit all of the transcriptional changes associated with T-cell activation at this early time point.

We next investigated gene-expression programs in CD8⁺ T cells at the 72-hour time point. Comparison of highly variable genes highlighted transcriptional targets of p53 (Dcxr, Ccng1, Rps27l, Phlda3, Bax, Ldhb, Cdkn1a, H2afj, Pmm1, Aen, Perp, Zmat3, and Ei24) that were more highly expressed in CD8⁺ T cells from DC + MDSC versus DC cocultures (Fig. 4A and B). We therefore performed GSEA of highly variable genes expressed in T cells from DC + MDSC versus DC cocultures at the 72-hour time point. This analysis again highlighted that p53 pathway genes were more highly expressed by T cells from DC + MDSC versus DC-only cocultures (Fig. 4C and D).

Subclustering of CD8⁺ T cells identified four clusters from the 72-hour time point, with clusters 2 to 4 representing CD8⁺ T cells from DC + MDSC cocultures (Fig. 4E). GSEA showed enrichment of p53 pathway genes in all three T-cell clusters from 72-hour DC + MDSC cocultures (clusters 2–4); the p53 pathway ranked among the top two pathways for clusters 3 and 4 and was also identified for cluster 2 (Fig. 4F and G). T cells in clusters 3 and 4 may therefore represent cells with the highest level of transcriptional activation of the p53 pathway, which is consistent with the expression pattern of p53 pathway genes across T-cell clusters in UMAP plots (Fig. 4B). Using qPCR, we validated significantly increased expression of p53 pathway genes (Ccng1, Cdkn1a, Phlda3, Bax, and Mdm2) at the 72-hour time point in CD8⁺ T cells from DC + MDSC versus DC cocultures (Fig. 4H). These data demonstrate that the presence of MDSCs induced transcriptional activation of the p53 pathway in CD8⁺ T cells.

We used genetic and pharmacologic approaches to investigate the role of iNOS (Nos2 gene) on activation of the p53 pathway in T cells. First, we compared the capacity of MDSCs derived from the bone marrow of Nos2^+/+ and Nos2^−/− mice (Supplementary Fig. S3A–S3C) to inhibit proliferation and induce p53 pathway activation in CD8⁺ T cells. Second, we evaluated the effects of the iNOS inhibitor L-NMMA and the antioxidant NAC on this phenotype. MDSCs from Nos2^−/− bone marrow failed to inhibit T-cell proliferation, consistent with the reported inhibitory effect of NO on T cells (refs. 34–36; Fig. 5A–D). Addition of the iNOS inhibitor L-NMMA partially restored the ability of T cells to proliferate in DC + MDSC cocultures, whereas the antioxidant NAC was less effective. No additive or synergistic advantage to CD8⁺ T-cell proliferation in the presence of MDSCs was observed by combining NAC and L-NMMA, and L-NMMA at a concentration of 125 μmol/L remained the most effective treatment for blocking MDSC-mediated inhibition of CD8⁺ T cells (Supplementary Fig. S4A–S4D). Furthermore, addition of L-NMMA to DC + Nos2^−/− MDSC cocultures had no additional effects on OT-I T-cell activation (Supplementary Fig. S4E–S4I). To further investigate the importance of NO production by MDSCs as a major mechanism of CD8⁺ T-cell inhibition, we isolated MDSCs directly from KPC tumors implanted into Nos2^+/+ or Nos2^−/− mice. Nos2-deficient MDSCs were unable to inhibit CD8⁺ T-cell proliferation compared with Nos2-proficient MDSCs (Fig. 5A–D; Supplementary Fig. S5A–S5C). We also evaluated other proposed mechanisms by which MDSCs inhibit T-cell function, including the expression of PD-L1 and secretion of IL10, but found that blocking these pathways did not restore CD8⁺ T-cell proliferation in DC + MDSC cocultures (Supplementary Fig. S5D–S5H). One mechanism of MDSC-mediated suppression is the catabolism of L-arginine (18–22), but we found that supplementation of cocultures with L-arginine did not improve CD8⁺ T-cell survival or proliferation in the presence of MDSCs (Supplementary Fig. S5I and S5J).

Our data strongly suggest that iNOS-mediated NO production is the major mechanism of MDSC-mediated CD8⁺ T-cell suppression. We therefore proceeded to investigate whether expression of iNOS by MDSCs could contribute to p53 pathway activation in CD8⁺ T cells. RT-qPCR analysis of OT-I T cells sorted from cocultures demonstrated that p53 pathway genes were only overexpressed in CD8⁺ T cells from cocultures with DCs + Nos2^+/+ but not Nos2^−/− MDSCs (Fig. 5E). Also, addition of the iNOS inhibitor L-NMMA reduced mRNA levels of p53 pathway genes in DC + MDSC cocultures (Fig. 5F). Finally, Western blot analysis demonstrated a higher degree of phosphorylation of p53 at serine 15 (S15) in CD8⁺ T cells from cocultures of DCs + Nos2^+/+ MDSCs compared with DCs + Nos2^−/− MDSCs (Fig. 5G; Supplementary Fig. S5K–S5M). We also found that expression of p53 target genes was induced in CD8⁺ T cells cocultured with purified M-MDSCs (Fig. 5H), consistent with functional data demonstrating that the presence of M-MDSCs inhibited CD8⁺ T-cell proliferation and induced apoptosis (Fig. 1G–L).

We used confocal microscopy to further define the kinetics of p53 activation (Fig. 5I–K; Supplementary Fig. S6A–S6C). CD8⁺ T cells were isolated at three time points (24, 48, and 72 hours) from cocultures with DCs, DC + Nos2^+/+ MDSCs, or DC + Nos2^−/− MDSCs. At the 48-hour time point, we observed an increase in the fraction of phospho-p53–positive CD8⁺ T cells from DC + Nos2^+/+ MDSC cocultures (Fig. 5I–K); this increase was absent when MDSCs lacked expression of iNOS (DCs only: 8% ± 2% positive cells, DCs + Nos2^+/+ MDSCs: 58% ± 11% positive cells, DCs + Nos2^–/− MDSC: 13% ± 3% positive cells). The increased phospho-p53 signal in CD8⁺ T cells was maintained at the 72-hour time point in T cells from DC + Nos2^+/+ MDSC cocultures. These data demonstrate that p53 activation in CD8⁺ T cells is dependent on iNOS expression by MDSCs.

Our data showed few MDSC-mediated effects at 24 hours and accumulation of activated p53 in CD8⁺ T cells after 48 hours following initiation of cocultures, suggesting that there was a delay in the induction of MDSC-mediated inhibition. Previous studies show IFNγ/TNFα-mediated induction of the Nos2 gene in stromal cells (35). Therefore, we tested the hypothesis that release of these cytokines by activated CD8⁺ T cells could induce iNOS expression in MDSCs by treating total MDSCs or FACS-sorted M-MDSCs with IFNγ, TNFα, or both cytokines for 24 hours, quantifying Nos2 mRNA levels by qRT-PCR (Fig. 5L). Stimulation of MDSCs with IFNγ specifically and significantly induced Nos2 mRNA in both total MDSCs (IFNγ: 6 ± 1.9-fold increase) and FACS-enriched M-MDSCs (IFNγ: 3.8 ± 0.8-fold increase). Using the Griess reaction, we detected nitrites only in supernatants from Nos2^+/+ MDSCs following stimulation with IFNγ (Fig. 5M). We also detected nitrites in cocultures of CD8⁺ T cells with DCs + MDSCs; again, nitrites were only detected when MDSCs originated from Nos2^+/+ mice (Fig. 5N). MDSC numbers correlated with their capacity to inhibit CD8⁺ T-cell proliferation (Fig. 1N), which also directly correlated with the detected concentrations of nitrites in cocultures (Fig. 5N). These data show that IFNγ, a cytokine released by CD8⁺ T cells following activation, induced MDSC-intrinsic Nos2 mRNA levels, resulting in the release of NO. The induction of iNOS expression in MDSCs by T cell–derived IFNγ explains why MDSCs do not block the initial steps of T-cell activation but rather induce T-cell apoptosis following IFNγ release.

The observed activation of the p53 pathway (in addition to “DNA repair” and “Reactive oxygen species” pathways; Fig. 4C) suggested the presence of DNA damage in CD8⁺ T cells cocultured with MDSCs. The histone protein H2AX is rapidly phosphorylated at residue S139 (γH2AX) by the kinases ATM and ATR in response to double-strand DNA breaks; γH2AX, therefore, serves as a sensitive marker for DNA damage. Confocal microscopy of FACS-enriched CD8⁺ T cells labeled with a γH2AX-AF488-specific antibody detected multiple nuclear γH2AX foci at the 48-hour time point in a substantial fraction of CD8⁺ T cells from DC + Nos2^+/+ MDSC cocultures (39% ± 8% positive cells with 7±3 foci), but minimally in CD8⁺ T cells from cocultures with DCs + Nos2^–/− MDSCs (13% ± 1% positive cells with 2 ± 1 foci) or DCs (7% ± 6% positive cells with 2 ± 1 foci; Fig. 6A–F; Supplementary Fig. S6D–S6F). Similar findings were made for the 72-hour time point. We also performed this analysis for cocultures with purified M-MDSCs (Fig. 6G–I; Supplementary Fig. S7). Confocal imaging showed that both total MDSCs and M-MDSCs induced DNA damage in CD8⁺ T cells. Taken together, these data show that iNOS expression by M-MDSCs is required for the induction of DNA damage and activation of the p53 pathway in CD8⁺ T cells.

We next examined whether iNOS expression by host cells could induce DNA damage in tumor-infiltrating T cells. KPC tumor cells (Nos2^+/+ genotype) were implanted subcutaneously into Nos2^+/+ or Nos2^−/− mice on a C57BL/6 background. Nos2 deficiency reduced tumor burden and increased tumor infiltration by CD8⁺ T cells (Fig. 7A–D) and natural killer (NK) cells (Fig. 7E). A larger percentage of tumor-infiltrating T cells in Nos2^−/− compared with Nos2^+/+ mice expressed IFNγ, TNFα, and granzyme B (Fig. 7F–I).

Confocal microscopy of sorted CD8⁺ T cells from KPC tumors that had been implanted into Nos2^+/+ or Nos2^−/− mice showed that a larger fraction of cells from Nos2^+/+ compared with Nos2^−/− mice were positive for nuclear γH2AX foci (Nos2^+/+ 74.8% ± 11.8% of cells; Nos2^−/− 26.9% ± 9.7% of cells; Fig. 7J–M; Supplementary Fig. S8). This difference was also reflected by the number of γH2AX foci per positive cells (Nos2^+/+ 6 ± 2 foci/cell; Nos2^−/− 3 ± 1 foci/cell) and a decrease in the integrated fluorescence intensity/cell (Nos2^+/+ 1,762 ± 918 AU/cell; Nos2^−/− 777 ± 564 AU/cell). These data provide in vivo evidence for DNA damage in tumor-infiltrating CD8⁺ T cells that is dependent on iNOS expression by host cells (Fig. 7N).

These studies demonstrate that MDSCs potently inhibit CD8⁺ T cells, even in the presence of DCs presenting a high-affinity cognate peptide. ScRNA-seq analysis demonstrated that the presence of MDSCs did not prevent the expression of key activation-induced genes at the 24-hour time point. Rather, CD8⁺ T cells expressed several genes associated with p53 pathway activation at a later time point, including those associated with cell-cycle arrest (Cdk1na, Phlda3) and apoptosis (Bax). Confocal microscopy demonstrated induction of DNA damage in CD8⁺ T cells, which was dependent on expression of the Nos2 gene by MDSCs. IFNγ, a key cytokine released by activated CD8⁺ T cells, increased expression of Nos2 and release of NO by MDSCs, explaining why initial T-cell activation remained intact but T-cell proliferation and survival were profoundly impaired. This finding was relevant to the in vivo setting because tumor-infiltrating CD8⁺ T cells from Nos2-deficient compared with Nos2-proficient mice had significantly less DNA damage.

Several previous studies demonstrate that NO released by MDSCs is important for inhibition of T-cell proliferation, and our data are consistent with these reports (10, 20, 23, 33). We also show that T-cell proliferation is strongly dependent on the number of MDSCs and the concentration of released NO. However, previous studies proposed that MDSCs block initial T-cell activation by nitrosylation of the TCR, CD8 coreceptor, or signaling molecules (23, 33). NO produced by iNOS can freely diffuse across cellular membranes and react with cysteine residues (37); it can also react with superoxide produced by NADPH oxidase to form peroxynitrite, a highly reactive compound that can oxidize cysteine and nitrosylate tyrosine residues (38). Such protein modifications can be identified by Western blotting in T cells cocultured with MDSCs (23, 33). However, our data demonstrate that MDSCs are inhibitory even when added to cocultures with T cells + DCs after a delay of 24 or 48 hours. Furthermore, they inhibit T-cell proliferation induced by cytokines rather than TCR stimulation. The scRNA-seq data clearly demonstrate the induction of an activation-related transcriptional signature in T cells from cocultures with DCs + MDSCs, although a shift in the distribution of T cells across some of the subclusters was noted at the 24-hour time point. The presence of MDSCs could certainly modify early T-cell activation processes, but we did not observe a block in initial T-cell activation in our experiments based on assessment of early T-cell activation markers and cytokine production.

Several suppressive mechanisms have been described for MDSCs, including release of IL10 and catabolism of arginine (see Supplementary Table S1 for details; refs. 18–22). However, we found that inhibition of NO release alone was sufficient to block MDSC-mediated inhibition of CD8⁺ T-cell proliferation. Furthermore, induction of DNA damage and activation of the p53 pathway in T cells was observed only with MDSCs from Nos2^+/+ but not Nos2^−/− mice, indicating a dependence on NO synthesis. Consistent with this, we observed increased T-cell apoptosis in the presence of MDSCs concurrent with a decline in T-cell numbers at later time points. These findings are consistent with the hypothesis that DNA damage and p53 pathway activation are relevant for the loss of T cells in these cultures. NO and peroxynitrite can induce deamination of DNA bases (including guanine and cytosine) and DNA breaks (39, 47). However, these highly reactive compounds can also modify proteins and membrane lipids, and we therefore do not propose that the induction of DNA damage is the sole mechanism for the observed inhibition of T-cell proliferation and the drastic decline in T-cell numbers. Other mechanisms, including depletion of key amino acids such as L-arginine, release of inhibitory cytokines, and expression of inhibitory surface molecules, are likely to be important contributors to the in vivo function of MDSCs (17, 24, 25). The relative importance of these mechanisms may depend on features of the tumor microenvironment.

We observed reduced tumor weight when KPC cells were implanted into Nos2^−/− compared with Nos2^+/+ mice. Also, KPC tumors from Nos2^−/− mice showed greater infiltration by CD8⁺ T cells, including IFNγ⁺ and Gzmb⁺ CD8⁺ T cells. Confocal imaging demonstrated that a substantially smaller percentage of tumor-infiltrating CD8⁺ T cells from Nos2^−/− hosts had evidence of DNA damage based on γH2AX staining, consistent with our conclusion that MDSCs can induce DNA damage and apoptosis in activated T cells. Reduced tumor growth in Nos2^−/− compared Nos2^+/+ mice may also reflect the dynamic relationship between T cell–mediated immunity and MDSC-mediated suppression: a reduced tumor mass may also decrease secretion of growth factors that promote differentiation of myeloid cells to immunosuppressive MDSCs, thus favoring T cell–mediated tumor immunity.

Tumors with substantial infiltration by MDSCs are characterized by impaired T cell–mediated tumor immunity (40–46). The profound block in proliferation and induction of apoptosis in T cells by MDSCs is therefore a significant barrier to tumor immunity. Strategies that inhibit the release of NO by MDSCs, block its detrimental effects on T cells, or reduce MDSC accumulation into tumors could therefore help to promote protective T cell–mediated tumor immunity.

K.W. Wucherpfennig reports grants from NIH and other from AACR (fellowship) and Cancer Research Institute (fellowship) during the conduct of the study, as well as grants from Novartis and personal fees and other from TCR2 Therapeutics (stock options), Immunitas Therapeutics (stock options), T-Scan Therapeutics (stock options), SQZ Biotech (stock options), and Nextech Invest (stock options) outside the submitted work. No disclosures were reported by the other authors.

The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH/NCI.

A.N.R. Cartwright: Conceptualization, data curation, formal analysis, funding acquisition, validation, investigation, methodology, writing–original draft, writing–review and editing. S. Suo: Software, formal analysis. S. Badrinath: Investigation. S. Kumar: Methodology. J. Melms: Methodology. A. Luoma: Data curation, methodology. A. Bagati: Resources. A. Saadatpour: Software, formal analysis. B. Izar: Methodology. G.-C. Yuan: Methodology. K.W. Wucherpfennig: Conceptualization, formal analysis, supervision, funding acquisition, writing–original draft, project administration, writing–review and editing.

The authors thank Dr. Stephanie Dougan for providing the murine KPC tumor cell line. This work was supported by NIH grants R01 CA238039, R01 CA251599, P01 CA163222 (to K.W. Wucherpfennig), and T32 CA207021 (to A. Luoma, A. Bagati, and K.W. Wucherpfennig); the 2018 AACR-AstraZeneca Lung Cancer Research Fellowship, Grant Number 18-40-12-KUMA (to S. Kumar); and a fellowship from the Cancer Research Institute (CRI; to A.N.R. Cartwright).

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.