DNA-PK inhibition enhances neoantigen diversity and increases T cell responses to immunoresistant tumors

Abstract

Effective antitumor T cell activity relies on the expression and MHC presentation of tumor neoantigens. Tumor cells can evade T cell detection by silencing the transcription of antigens or by altering MHC machinery, resulting in inadequate neoantigen-specific T cell activation. We identified the DNA-protein kinase inhibitor (DNA-PKi) NU7441 as a promising immunomodulator that reduced immunosuppressive proteins, while increasing MHC-I expression in a panel of human melanoma cell lines. In tumor-bearing mice, combination therapy using NU7441 and the immune adjuvants stimulator of IFN genes (STING) ligand and the CD40 agonist NU-SL40 substantially increased and diversified the neoantigen landscape, antigen-presenting machinery, and, consequently, substantially increased both the number and repertoire of neoantigen-reactive, tumor-infiltrating lymphocytes (TILs). DNA-PK inhibition or KO promoted transcription and protein expression of various neoantigens in human and mouse melanomas and induced sensitivity to immune checkpoint blockade (ICB) in resistant tumors. In patients, protein kinase, DNA-activated catalytic subunit (PRKDC) transcript levels were inversely correlated with MHC-I expression and CD8+ TILs but positively correlated with increased neoantigen loads and improved responses to ICB. These studies suggest that inhibition of DNA-PK activity can restore tumor immunogenicity by increasing neoantigen expression and presentation and broadening the neoantigen-reactive T cell population.

Keywords: Adaptive immunity; Antigen presentation; Cancer immunotherapy; Immunology; Oncology.

Figure 1. Combination immunotherapy with DNA-PK inhibition demonstrates potent antitumor CD8⁺ T cell response and is associated with a favorable antigen processing and inflammatory gene expression profile.

(A) Schema of the treatment protocol. C57BL/6 mice with established (25 mm²) B16-F10 tumors underwent the treatment plans. (A with B) Tumor growth and (C) survival were monitored for 40 days. (D and E) Mice with established tumors were treated as described in A, and tumors were collected 7–9 days after initiation of treatment. (D) Volcano plots display the log₂(fold change) in total mRNA transcript expression levels in B16-F10 tumors, comparing treatment versus no treatment, and the associated log₁₀(P values) generated by NanoString gene expression analysis of 3 tumors treated with NU or SL40 or no drug and 4 tumors treated with NU-SL40. Genes in D are colored according to their pathway association. (E) Bubble plots depict the fold change in gene expression in NU-SL40–treated tumors from pathways highlighted in volcano plots as being upregulated or downregulated compared with the untreated group. Bubble size represents the average mRNA transcript counts in NU-SL40 replicates. The P value (compared with untreated tumors) is depicted by a color scale. ****P < 0.0001, by a mixed-effects model with Geisser-Greenhouse correction and Tukey’s multiple-comparisons test (B) and Kaplan-Meier survival curve with log-rank (Mantel-Cox) (C). Avg, average.

Figure 2. NU-SL40 treatment promotes the infiltration of activated CD8⁺ TILs and alters the tumor myeloid cell compartment.

(A) Mice with established tumors were treated as described in Figure 1A. The indicated tumor lymphoid cell populations of single-cell, viable CD3⁺ or CD3^–CD45⁺ cells normalized to 50,000 CD45⁺ cells were determined by flow cytometry. (no drug [ND]: n = 6; SL40: n = 5; NU: n = 4; NU-SL40: n = 5). **P < 0.01, by 2-way ANOVA. (B) Representative flow plots with adjunct MFI histograms and (C) pie charts representing the percentage of CD8⁺ TILs expressing PD-1 and/or 4-1BB across treatment groups (no drug: n = 5; SL40: n = 5; NU: n = 4; NU-SL40: n = 4). (D) Ratio of CD8⁺/CD4⁺ TILs (no drug: n = 20; SL40: n = 14; NU: n = 9; NU-SL40: n = 13). ****P < 0.0001, by 2-way ANOVA. (E) UMAP analysis of the pooled single-cell, viable CD45⁺ TIL populations (top panel) described in A–C (CD4, CD8, NK1.1, B cells) and (bottom panel) M1- or M2-like macrophages identified as CD45⁺F4/80⁺CD11c⁺CD206^– or CD45⁺F4/80⁺CD11c^–CD206^–; F4/80⁺CD45⁺CD11c^–CD206⁺; MDSCs: CD11b⁺Gr1⁺; DC, CD45⁺CD11c⁺MHC-II⁺ (no drug: n = 6; SL40: n = 5; NU: n = 4; NU-SL40: n = 5).

Figure 3. DNA-PK inhibition drives TCRvβ diversity of highly functional tumor-reactive CD8⁺ T cells.

(A) Schematic of drug treatment and tissue processing with representative flow cytometric analysis of TCRvβ on CD8⁺ TILs. SSC-A, side scatter area. (B) UMAP distribution of the absolute number of CD8⁺ TILs clustered by TCRvβ chain (number labels and color scale for differentiation) (untreated: n = 15; SL40: n = 9; NU: n = 5; NU-SL40: n = 8). (C) Number of CD8⁺ TILs by TCRvβ chain per 1 million single-cell events. A rout outlier test was performed. Blue and red bars represent significant decreases or increases in TCRvβ counts in treatment conditions compared with no treatment. Each dot represents a single mouse (untreated: n = 15; SL40: n = 9; NU: n = 5; NU-SL40: n = 8). (D) Schematic of C57BL/6 B16-F10 tumor model and tumor collection for TIL isolation via magnetic bead–positive selection followed by ex vivo culturing with or without IFN-γ–pretreated (100 U/mL for 24 hours) B16-F10 melanoma cells. (E) Representative flow plot with adjunct MFI histograms representing the number of isolated CD8⁺ TILs expressing GzmB and PD-1 obtained from control and NU-SL40–treated mice (16-hour coculture). (F) Heatmap of TCRvβ distribution of CD8⁺ TILs that expressed PD-1 and produced GzmB. TILs were pooled from tumors (untreated: n = 4; NU-SL40: n = 5), and counts were normalized to 2 × 10⁵ CD3⁺ cells. The sum of the TCRvβ chain in each condition is represented above the columns, the sum of total TCRvβ in each condition is indicated to the right of each row. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001, by multiple unpaired, 2-tailed t test.

Figure 4. DNA-PK inhibition increases tumor-associated antigen expression levels, induces a unique neoantigen expression profile in melanoma, and represents better targets for human TILs.

(A and B) B16-F10 melanoma cells were treated with 2 μM NU7441 or DMSO control for 72 hours, at which point gradual darkening was observed and the OD₄₀₅ recorded. (C) Bar graph comparing the levels of RNA per FPKM of known melanogenesis-associated antigens 48 hours after treatment with 2 μM NU7441 or DMSO control. The fold change between DMSO and NU7441 treatments is noted above the bars. (D) B16-F10 melanoma cells were treated with 2 μM NU7441 or DMSO control for 48 hours, and the levels of the indicated proteins were determined by Western blotting. The fold change between groups is shown to the right of each band. (E and F) B16-F10 melanoma cells were treated as described in A, and the neoantigens and FPKM were determined as described in Methods. (E) Venn diagram representing the number of uniquely expressed or shared B16-F10 neoantigens present in control-treated melanomas and those induced by NU7441. (F) The gene name and amino acid mutation expressed following DNA-PKi treatment are shown on the left. The matched bar graph shows the levels of RNA per FPKM of neoantigen-producing genes exposed to NU7441 treatment, and the binding affinity of these epitopes for H2-Db and H2-Kb was determined using the MHC binding prediction algorithms from the Immune Epitope Database and Tools (IEDB) (iedb.org) site. (G) Schematic showing the generation of melanoma cell lines and TILs from a patient melanoma tumors and the experiments performed in H–J. (H) The MB3429 melanoma cell line was treated with 2 μM NU7441 or DMSO control for 48 hours, and the levels of the indicated transcripts were determined by RT-PCR and are shown as ΔCt. (I) MB3429 melanoma cells were treated with 2 μM NU7441 or DMSO control for 48 hours, and the levels of the indicated proteins were determined by Western blotting (fold change between groups is indicated to the right of each band). (J) Matched TILs and tumors were derived from the same tumor fragment. The tumor cell line was treated with DMSO or DNA-PKi (2 μM NU7441) for 48 hours, at which point the drug was washed off prior to coculturing with TILs at a 1:1 ratio for 18 hours. Cytotoxicity was determined by annexin V staining with flow cytometric gating on tumor cells (based on light scatter and CD3^–) and viability dye.

Figure 5. DNA-PKi plus an immune adjuvant drive the generation and expansion of a unique panel of neoantigen-reactive TILs with enhanced effector function ex vivo.

(A) Schematic of the experimental design. Mice were treated as described in Figure 3D. TILs were isolated from NU-SL40 or untreated tumors using a positive magnetic selection for CD4⁺ and CD8⁺ T cells. Twelve plasmids were generated to contain TMGs of the 10 neoantigens identified in Figure 4. (B and C) TMGs were transfected into the murine DC2.4 line and cocultured with CD4⁺ and CD8⁺ TILs collected from control- or NU-SL40–treated mice (pooled from 10 mice/group) at a 1:10 TIL/DC ratio. After 48 hours, IFN-γ production by TCRvβ-specific responses to DC-presented neoantigens was determined by ELISA. Bar graphs depict IFN-γ production by TILs stimulated with TMG-DCs compared from 2 independent experiments. Values were normalized to production after stimulation with a TMG-GFP control. (D and E) The ability for CD8⁺ TILs to produce IFN-γ or GzmB was determined by intracellular staining and flow cytometry. TCRvβ usage in response to stimulation with each TMG-expressing DC was also investigated. Heatmaps represent the number of CD8⁺ TIL per 3,000 total TIL expressing different TCRvβ chains and producing (D) IFN-γ or (E) GzmB in response to stimulation from each TMG.

Figure 6. PRKDC levels inversely correlated with TILs, MHC-I, and the response to ICB therapy in patients with melanoma and are mirrored by B16-F10^{PRKDC KO} tumors.

(A) Scatter plot of z scores for HLA-A and CD8α expression versus PRKDC expression obtained from TCGA. (B) Associations between CD8α and PRKDC mRNA expression by z score, with overall survival in months indicated by the color scale in patients who were responders (large circles) or nonresponders (small circles). (C) Graph distinguishing the percentage of CD8^lo, CD8^hi, PRKDC^lo, and PRKDC^hi cells in melanoma tumors that responded or not to ICB. (D) Percentage of patients with melanoma expressing WT or altered PRKDC, who responded or not to ICB. (E) Violin plots depicting differences in tumor mutation burden (left, P < 0.0001) and neoantigen load (right, P = 0.0002) in patients with normal (WT, n = 172) versus altered (n = 40) PRKDC expression. Statistical significance was determined by unpaired Mann-Whitney U test. (F) Staining for total DNA-PK and p–DNA-PK (Ser2056) in samples from patients with melanoma. (G) C57BL/6 mice with established (25 mm²) B16-F10 tumors remained untreated or were treated with anti–PD-1/–CTLA-4 blockade, NU-SL40, or NU-SL40 in conjunction with anti–PD-1/–CTLA-4 blockade (n = 8/group). Tumor growth was monitored over time. (H and I) WT B16-F10 cells (orange) and melanoma cells engineered to KO PRKDC (teal) were injected into mice. When tumors were established, mice were left untreated or treated with anti–PD-1/–CTLA-4 with or without anti-CD40 therapy. (H) Tumor growth and (I) survival were monitored over time (n = 8 mice/group). (J) Mice treated with combination anti–PD1/–CTLA-4 and anti-CD40 that showed controlled tumor growth were rechallenged with DNA-PK–KO cells after 300 days (naive, n = 4; rechallenge; n = 5). Tumor growth and survival were monitored among rechallenged and naive, challenged mice using 2-way ANOVA. *P < 0.05, ***P < 0.001, and ****P < 0.0001.