Targeting Innate Immune Checkpoint TREX1 Is a Safe and Effective Immunotherapeutic Strategy in Cancer

Abstract

Three-prime repair exonuclease 1 (TREX1) is the major DNase in mammalian cells that degrades cytosolic DNA to prevent activation of the cyclic GMP-AMP synthase (cGAS)-stimulator of interferon genes (STING) pathway. Genotoxic stress, DNA damage, and radiotherapy induce TREX1 expression in cancer cells, allowing them to evade innate immune activation of type I IFN-mediated antitumor response. Therefore, targeting TREX1 could represent a potential approach to stimulate antitumor immunity. In this study, we conducted a high-throughput small-molecule inhibitor screen of TREX1 using a cell-free DNase assay. Compound 296 specifically inhibited TREX1 DNase activity at low micromolar concentrations, induced type I IFN signaling in cancer cells, and inhibited tumor growth in mice in an inteferon alpha/beta receptor (IFNAR)-dependent manner. Treatment with compound 296 also stimulated T-cell infiltration into tumors and synergized with immune checkpoint blockade. Trex1 knockout cancer cells elicited robust systemic antitumor immunity through tumor-intrinsic cGAS-STING activation and functioned as autologous cancer vaccines that protected against tumor challenge and metastasis. An inducible whole-body Trex1 knockout mouse model was established to simulate "on-demand" systemic TREX1 inactivation in adult mice. Sustained TREX1 loss suppressed a broad range of solid and metastatic tumors in adult mice without incurring severe immune toxicity, even when combined with immune checkpoint blockade, demonstrating the feasibility of an immune-safe therapeutic window. Together, these data demonstrate the antitumor efficacy and immune safety of multiple therapeutic modalities targeting TREX1, including targeting small-molecule inhibitors of TREX1 and employing TREX1 knockout tumor cells as an autologous cancer vaccine. These approaches should pave the way for developing TREX1-targeted cancer immunotherapies.

Significance: Therapeutic modalities targeting TREX1 can activate cGAS-STING signaling and can be incorporated into autologous cancer vaccine designs to improve cancer treatment, supporting the potential of inactivating TREX1 to harness innate immunity. See related commentary by Hanks, p. 2778.

Figure 1.. High-throughput small-molecule inhibitor screen of TREX1.

(A) A diagram of the cell-free TREX1 DNase assay. Recombinant human TREX1 (DNase domain only, same throughout) is incubated with SYBR (S)-labeled single-stranded DNA (ssDNA). Fluorescent signal is only detectable when SYBR complexes with DNA. DNA degradation by TREX1 reduces fluorescent signal. (B) Fluorescent measurement of DNase activity. Recombinant human TREX1 and DNase I was incubated with ssDNA/SYBR for indicated time (x-axis). ssDNA/SYBR fluorescence intensity at 0 minutes (without enzyme) was normalized to 100. Data represent mean ± SD of n = 2 biological replicates per group. (C) Fluorescent measurement of DNase activity. Recombinant human TREX1 at decreasing concentrations (1 μM, 100 nM, 10 nM, and 1 nM) was incubated with ssDNA/SYBR for indicated time (x-axis). EDTA was added as negative control. Data represent mean ± SD of n = 2 biological replicates per group. (D) A summary of the screen workflow. (E) Confirmation screen of the 1,920 primary hits. RZ scores of compounds at 5 μM are shown. Compound I (CmpI) is a positive control used at 25 μM. Dash lines indicate the RZ score threshold > 3. (F) Cytotoxicity screen of the 41 potent TREX1 inhibitors identified in the confirmation screen. Mouse embryonic fibroblasts (MEFs) were treated with each compound at 10 μM for 30 h, then cell confluency was measured by IncuCyte. (G) IC₅₀ of top TREX1 inhibitors using the cell-free DNase assay. The dose-dependent inhibitory curve was fit to calculate the IC₅₀. Nonlinear regression with variable slope model was used. Data represent mean ± SD of n = 2 (Cmp I) or n = 3 (#157, #296 and #289) biological replicates. (H, I) qRT-PCR analysis of IFN and inflammatory genes in MC38 cells treated with TREX1 inhibitors. MC38 cells were incubated with the indicated compounds at increasing concentrations (10, 50, 100 μM) or at a single dose (20 μM) for 24, 48, or 72 hours. Each immune gene mRNA fold change was normalized to the DMSO group (set to 1). “Immune gene score” is the sum of all five mRNA fold change values (e.g., DMSO value is 5). Data represent mean of n = 3 biological replicates per group. ND, not detected because of toxicity. Summary of data is shown in (H). Individual data points from each mRNA for compound #296 or #289 is shown in (I). Data represent mean ± SD of n = 3 biological replicates per group. One-way analysis of variance (ANOVA) with Dunnett’s multiple comparisons testing was used to generate P values. (A, D, Created in BioRender. Xing, C. (2025) https://BioRender.com/lptdizk, https://BioRender.com/kwmz4sb)

Figure 2.. Compound #296 induces antitumor immunity and synergizes with immune checkpoint blockade.

(A) Cell-free DNase assay of compound #296 with rhTREX1 and DNase I. The dose-dependent inhibitory curve was fit to calculate the IC₅₀ against rhTREX1. Nonlinear regression with variable slope model was used. Data represent mean ± SD of n = 3 biological replicates. (B) Intermolecular contacts of compound #296 bound to hTREX1 DNase domain (amino acids 1 to 242) in molecular docking structure (left). #296 (orange), contacted TREX1 amino acids (cyan), and key enzymatic active sites (green) are shown as a stick model (right). Hydrogen-bounds are labeled in yellow. (C) Tumor growth study. Wild-type mice were subcutaneously implanted with MC38 tumors on the flank. When average tumor size reached 40 mm³, mice were peritumorally injected with vehicle or 3 μmol of TREX1 inhibitor #296 for 2 cycles spaced 1 week apart (arrows). Blinded caliper measurements of tumor size are shown. Data represent mean + SD of n ≥ 7 mice per group. Two-way ANOVA with Dunnett’s multiple comparisons testing was used to generate the P value. (D-M) Flow cytometric analysis of tumor infiltrating immune cells isolated from vehicle- or #296- treated MC38 tumors in (C). Quantifications of frequency or number of cells are shown for CD45⁺ immune cells (D), CD11b⁺ myeloid cells (E), CD3⁺ T cells (F), CD8⁺ T cells (G), and CD4⁺ T cells (H). Ratios of CD8⁺ to CD4⁺ T cells are shown in (I). Quantifications of percentage or mean fluorescence intensity (MFI) of PD-1, GZMB, or SCA-1 are shown for CD8⁺ T cells (J-L) and CD11b⁺ myeloid cells (M). Data represent mean ± SD of n = 4 mice per group. Unpaired two-tailed Student’s t test was used to generate P values. (N) Compound #296 and cGAMP combination therapy tumor study. Wild-type mice were subcutaneously implanted with MC38 tumors on the flank. On days 9, 12, 15, mice were peritumorally injected with vehicle or 3 μmol of compound #296 (red arrows). On days 10, 13, 16, mice were peritumorally injected with 10 μg cGAMP prepared in mixture with Lipfectamine 2000 (blue arrows). Blinded caliper measurements of tumor size are shown. Data represent mean + SD of n ≥ 6 mice per group. Two-way ANOVA with Dunnett’s multiple comparisons testing was used to generate the P values. (O) Compound #296 and anti-PD-1 combination therapy tumor study. Wild-type mice were subcutaneously implanted with B16-F10 tumors on the flank. On days 9, 12, 14, mice were peritumorally injected with vehicle or 3 μmol of compound #296 (red arrows). On days 10, 13, 15, 17, mice were intraperitoneally injected with 200 μg isotype control or anti-PD-1 antibody (blue arrows). Blinded caliper measurements of tumor size are shown. Data represent mean + SD of n ≥ 7 mice per group. Two-way ANOVA with Dunnett’s multiple comparisons testing was used to generate the P values.

Figure 3.. TREX1 deficient cancer cells induce robust antitumor immunity.

(A) qRT-PCR analysis of immune gene expression. Resting-state mRNA levels of IFN-β (Ifnb1), ISGs (Ifi27l2a, Cxcl10, Ccl5), and inflammatory genes (Tnf) were measured by qRT-PCR in wild-type (Trex1^WT) and Trex1^KO MC38 cells. Data represent mean ± SD of n = 4 biological replicates per group. Unpaired two-tailed Student’s t test was used to generate P values. (B) Cell proliferation analysis of Trex1^WT and Trex1^KO MC38 cells. Confluence was measured by continuous live-cell imaging using IncuCyte. Data represent mean ± SD of n = 4 biological replicates per group. Two-way ANOVA was used to generate the P value. (C-L) Tumor growth and survival studies. Trex1^WT or Trex1^KO MC38 cells were implanted subcutaneously on the flank of wild-type (WT, C and D), Sting1^−/− (E and F), Rag1^−/− (G and H), NSG (I and J), or Ifnar1^−/− (K and L) mice. Blinded caliper measurements of tumor size are shown in C, E, G, I, and K. Data represent mean + SD of n = 10 (C, E and I) or n = 7 (G and K) mice per group. Two-way ANOVA with Šídák’s multiple comparisons testing was used to generate P values. Kaplan-Meier survival curves are shown in D, F, H, J, and L. n = 10 (D, F and J) or n = 7 (H and L) mice per group. P values were generated by the log-rank (Mantel-Cox) test. ns, not significant.

Figure 4.. Tumor-intrinsic cGAS-STING signaling drives systemic antitumor immunity.

(A-C) Tumor growth and survival study. Wild-type mice were subcutaneously implanted with Trex1^WT, Trex1^KO, Trex1^KOCgas^KO, Trex1^KOSting1^KO, Trex1^KOTbk1^KO, and Trex1^KOIrf3^KO MC38 cells. Blinded caliper measurements of tumor size are shown in (A). Growth curves of individual tumors are shown in (B). Data represent mean + SD of n ≥ 8 mice per group. Two-way ANOVA with Dunnett’s multiple comparisons testing was used to generate P values. Kaplan-Meier survival curves are shown in (C). n ≥ 8 mice per group. P values were generated by the log-rank (Mantel-Cox) test. (D-M) Flow cytometric analysis of tumor infiltrating immune cells isolated from Trex1^WT and Trex1^KO MC38 tumors grown in wild-type C57BL/6J mice. Quantifications of frequency or number of cells are shown for CD45⁺ immune cells (D), CD11b⁺ myeloid cells (E), CD3⁺ T cells (F), CD8⁺ T cells (G), and CD4⁺ T cells (H). Ratios of CD8⁺ to CD4⁺ T cells are shown in (I). Quantifications of percentage or MFI of PD-1, GZMB, or SCA-1 are shown for CD8⁺ T cells (J-L) and CD11b⁺ myeloid cells (M). Data represent mean ± SD of n ≥ 4 mice per group. Unpaired two-tailed Student’s t test was used to generate P values. (N-P) Autologous cancer vaccine solid tumor study. A schematic of the experiment is shown in (N). Wild-type mice were mock-treated (Naïve) or subcutaneously vaccinated on the left flank with Trex1^KO MC38 cells on day −7, day 0, or day +3 (scheme 1 to 3). On day 0, mice were challenged with Trex1^WT MC38 cells subcutaneously injected into the right flank. Tumor burden was measured by blinded caliper measurements from day 0 to day 30. Blinded caliper measurements of tumor size are shown in (O). Data represent mean + SD of n ≥ 6 mice per group. Two-way ANOVA (mixed-effects model) with Šídák’s multiple comparisons testing was used to generate P values. Kaplan-Meier survival curves are shown in (P). n ≥ 6 mice per group. P values were generated by the log-rank (Mantel-Cox) test. (Q-R) Autologous cancer vaccine metastatic tumor study. A schematic of the experiment is shown in (Q). Wild-type mice were mock-treated (Naïve) or subcutaneously vaccinated on the left flank with Trex1^KO B16-F10 cells on days −7, −4, and −1. On day 0, mice were challenged with Trex1^WT B16-F10 cells intravenously (i.v.) injected via the tail vein. On day 15, mice were sacrificed for lung metastasis analysis. Quantifications of lung metastatic burden are shown in (R). (Left) Number of metastatic foci on the surface of lung. (Right) Metastasis area as percentage of the area of lung on the section. Data represent mean + SD of n ≥ 7 mice per group. P values were given by two-tailed unpaired Mann-Whitney U test. (N, Q, Created in BioRender. Xing, C. (2025) https://BioRender.com/6byqcra, https://BioRender.com/az7zn7d)

Figure 5.. Inducible whole-body Trex1 loss in adult mice inhibits tumor growth.

(A) Study design for early treatment of solid tumor. Trex1^fl/flUBC^CreERT2 and Trex1^fl/fl mice were subcutaneously implanted with cancer cells (as indicated on top of subsequent panels), followed by 5 days of intraperitoneal (i.p.) administration of tamoxifen to induce whole-body Trex1 knockout only in Trex1^fl/flUBC^CreERT2 mice (iTrex1^KO). Tumor size was measured from day 6 to day 30. (B) Immunoblot analysis of mouse tissue lysates from Trex1^fl/fl and iTrex1^KO mice 4 weeks after tamoxifen treatment. (C, D) MC38 colon carcinoma study. Mice were treated as in (A). Blinded caliper measurements of tumor size are shown in (C). Data represent mean + SD of n ≥ 8 mice per group. Two-way ANOVA with Šídák’s multiple comparisons testing was used to generate the P values. Kaplan-Meier survival curves are shown in (D). P value was generated by the log-rank (Mantel-Cox) test. (E) E0771 breast carcinoma study. Mice were treated as in (A). Blinded caliper measurements of tumor size are shown. Data represent mean + SD of n ≥ 5 mice per group. Two-way ANOVA with Šídák’s multiple comparisons testing was used to generate the P value. (F, G) Lewis lung carcinoma (LLC) study. Mice were treated as in (A). Blinded caliper measurements of tumor size are shown in (F). Data represent mean + SD of n ≥ 8 mice per group. Two-way ANOVA with Šídák’s multiple comparisons testing was used to generate the P value. Inset below shows blinded caliper measurements of tumor size on day 24 and 27. Two-tailed unpaired Mann-Whitney U test was used. Kaplan-Meier survival curves are shown in (G). P value was generated by the log-rank (Mantel-Cox) test. (H) Study design for early treatment of metastatic tumor. On day 0, Trex1^fl/fl and Trex1^fl/flUBC^CreERT2 mice were intravenously injected with B16-F10 melanoma cells, followed by 5 days of intraperitoneal administration of tamoxifen. Mice were sacrificed on day 22 and lung metastasis was analyzed. (I-K) B16-F10 metastatic melanoma study. Mice were treated as in (H). Number of metastatic foci on the surface of lung is shown in (I). Data represent mean + SD of n ≥ 5 mice per group. P value was given by two-tailed unpaired Mann-Whitney U test. Metastasis area as percentage of the area of lung on the section is shown in (J). P value was given by two-tailed unpaired Mann-Whitney U test. Representative hematoxylin and eosin (H&E) -stained lung sections from Trex1^fl/fl and iTrex1^KO mice are shown in (K). Scale bars, 2 mm. (L) Study design for late treatment of solid tumor. Trex1^fl/fl and Trex1^fl/flUBC^CreERT2 mice were subcutaneously injected with MC38 cells on day 0. Tamoxifen was administered intraperitoneally from day 12 to 16. (M) Blinded caliper measurements of tumor growth. Data represent mean + SD of n = 7 mice per group. Two-way ANOVA with Šídák’s multiple comparisons testing was used to generate the P value. Day 22 blinded caliper measurements of MC38 tumors are shown on the right. Data represent mean ± SD of n = 7 mice per group. P value was given by two-tailed unpaired Mann-Whitney U test. (A, H, L, Created in BioRender. Xing, C. (2025) https://BioRender.com/xy3mzd5)

Figure 6.. Trex1 loss in adult mice induces low levels of systemic immune activation.

(A) qRT-PCR analysis of immune gene expression. Splenocytes isolated from Trex1^fl/fl and iTrex^KO mice 3 months after tamoxifen treatment were analyzed. Data represent mean ± SD of n ≥ 3 mice per group. Unpaired two-tailed Student’s t test was used to generate P values. (B-G) Flow cytometric analysis of splenic T cells isolated from Trex1^fl/fl and iTrex^KO mice 3 months after tamoxifen treatment. Representative FACS plots and quantifications of Naïve and memory cell phenotypes are shown for CD4⁺ (B) and CD8⁺ (C) T cells. CM, central memory; EM, effector memory. Data represent mean ± SD of n ≥ 4 mice per group. Unpaired two-tailed Student’s t test was used to generate P values. CD69 and CD25 FACS analysis showing activation status of CD4⁺ (D, E) and CD8⁺ (F, G) T cells. Data represent mean ± SD of n ≥ 4 mice per group. Unpaired two-tailed Student’s t test was used to generate P values.

Figure 7.. Trex1 loss in adult mice does not cause severe immune toxicity.

(A) A diagram of study design. Trex1^fl/flUBC^CreERT2 and Trex1^fl/fl mice were intraperitoneally (i.p.) administered with tamoxifen for 5 days. Serum samples were collected at 2, 4, and 26 weeks after tamoxifen treatment for Luminex cytokine analysis. Tissues were collected at 26 weeks for histology and immunoblot analyses. (B) Immunoblot analysis of mouse tissue lysates from Trex1^fl/fl and iTrex1^KO mice 26 weeks after tamoxifen treatment. (C-D) Luminex cytokine analysis of sera from Trex1^fl/fl and iTrex1^KO mice collected at 2, 4 and 26 weeks after tamoxifen treatment. Serum from a 5.4-week-old Trex1^−/− mouse is included for comparison. The heatmap of all 23 cytokines and chemokines assessed is shown in (C). No value means value below detection limit. Bar graph of IL-12p40 is shown in (D). Data represent mean ± SD of n ≥ 5 mice per group. Unpaired two-tailed Student’s t test was used to generate the P value. (E) Serum CRP levels of Trex1^fl/fl and iTrex1^KO mice 26 weeks after tamoxifen treatment. Data represent mean ± SD of n ≥ 5 mice per group. Unpaired two-tailed Student’s t test was used to generate the P value. (F, G) Body weight (F) and spleen weight (G) of Trex1^fl/fl and iTrex1^KO mice at 26 weeks after tamoxifen treatment. Data represent mean ± SD of n = 6 mice per group. Unpaired two-tailed Student’s t test was used to generate the P value. (H) Representative H&E-stained tissue sections from Trex1^fl/fl and iTrex1^KO mice 26 weeks after tamoxifen treatment. Tissues from a 5.4-week-old Trex1^−/− mouse are shown as positive controls. Arrows indicate immune infiltration. Scale bars, 100 μm for heart; 200 μm for skeletal muscle and skin; 800 μm for liver, spleen and lung. (I-M) Tissue IHC analysis of Trex1^fl/fl and iTrex1^KO mice 26 weeks after tamoxifen treatment. Quantifications of the percentage of CD45⁺ area are shown for heart (I), skeletal muscle (J), skin (K), liver (L), and lung (M). Tissues from a 5.4-week-old Trex1^−/− mouse are included as positive controls. Data represent mean ± SD of n = 4 mice per group. Unpaired two-tailed Student’s t test was used to generate the P values. (A, Created in BioRender. Xing, C. (2025) https://BioRender.com/xy3mzd5)