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

Mice

Wild-type C57BL/6J, Rag1^−/−, Ifnar1^−/−, NSG, and UBC^CreERT2 mice were purchased from the Jackson Laboratory (#000664, #002216, #028288, #005557, and #007001). Sting1^−/− mice were obtained from Glen Barber (University of Miami, USA). Trex1^fl/fl mice were obtained from Rayk Behrendt (University of Bonn, Germany) and Axel Roers (University of Heidelberg, Germany). Trex1^fl/fl mice were bred to UBC^CreERT2 mice to generate tamoxifen-inducible (Trex1^fl/flUBC^CreERT2) knockout mice. Both male and female mice were used in the study. All mice were housed in standard pathogen free barrier facilities under 12-hour light-dark cycles. Animal work was approved by the Institutional Animal Care and Use Committee (IACUC) at University of Texas Southwestern Medical Center.

Cell lines

HEK293T, E0771, and B16-F10 cells were obtained from ATCC. MC38 cells were obtained from Yang-Xin Fu (University of Texas Southwestern Medical Center). LLC cells were obtained from Cheng Cheng Zhang (University of Texas Southwestern Medical Center). Primary MEFs were isolated and cultured as described before¹⁸. Cells were cultured in DMEM with 10% (vol/vol) FBS, 10 mM HEPES, 2 mM L-glutamine, and 1 mM sodium pyruvate with the addition of 100 U/ml penicillin and 100 mg/ml streptomycin at 37°C with 5% CO₂. Mycoplasma tests were conducted monthly and confirmed to be negative. A maximum of 5 passages or 3 months in culture were allowed for each cell line. Authentication of the cell line were performed when they were initially acquired according to morphological or genetic properties described by the source or vendor. The biological sex of the cells are unknown.

Purification of recombinant human TREX1 protein

The DNase domain (amino acids 1-242) of human TREX1 protein was cloned to 6×His pQE30 (T7 promoter) vector and subsequently transformed into BL21(DE3) competent E. coli (C2527H, NEB). A single bacterial colony with the plasmid was transferred into LB media with 30 μg/ml kanamycin and 33 μg/ml chloramphenicol. The bacteria culture was incubated at 37°C overnight in a shaking incubator. A fresh 1 liter of LB media with 30 μg/ml kanamycin was inoculated with the 10 ml primary culture and incubated at 37°C for 5 hours while shaking. 0.5 mM IPTG was added when absorbance at 600 nm reached 0.6, and the 1-liter culture was incubated at 18°C overnight in a shaking incubator. Bacteria were pelleted and resuspended in 100 ml of lysis buffer (20 mM PBS pH 7.0, 500 mM NaCl, 10% glycerol, 1% Triton-X 100, 1× protease inhibitor), followed by sonication until the solution was clear. The solution was further centrifuged at 17,000 rpm for 1 hour at 4°C and the supernatant was transferred to new 50 ml conical tubes. 500 μl of Ni-NTA agarose (R90115, Invitrogen) and 10 mM imidazole were added to the supernatant. The solution mix was incubated for 30 minutes at 4°C and then passed through the column at 4°C. The Ni-NTA agarose was washed with 1 litter lysis buffer supplemented with 25 mM imidazole. 2 ml of lysis buffer supplemented with 350 mM imidazole was added to the column. Elution was collected after 5 minutes of incubation.

Cell-free DNase assay

The cell-free DNase assay was carried out with a 30-mer ssDNA substrate (5’-TTTTTGGTTTGGTTTTGGCGGAGTTTTCGG-3’). Recombinant human TREX1 and DNase I (18047019, Invitrogen) was diluted in lysis buffer (20 mM Tris-HCl pH 7.4, 5 mM MgCl₂, 2 mM DTT, 100 μg/ml BSA, and 1% NP-40). Diluted human TREX1 or DNase I were added to the plate. ssDNA/SYBR Gold mix was added in 30 μl of assay buffer (20 mM Tris-HCl pH 7.4, 5 mM MgCl₂, 2 mM DTT, 100 μg/ml BSA, 31.6 nM 30-mer ssDNA and 0.72x SYBR Gold (S11494, Invitrogen)). For specificity assessment of compound #296 (aurintricarboxylic acid, ab141027, Abcam) against TREX1 and DNase I, ssDNA was replaced by dsDNA with 3’ overhang (purified pcDNA3.1 plasmid digested with ApaI (R0114S, NEB) and KpnI-HF (R3142S, NEB)) to prepare the assay buffer. 15 nM TREX1 or 0.5 U/μl DNase I were diluted in 10 μl lysis buffer, mixed with 10 μl diluted compound for 30 minutes at room temperature, followed by addition of 80 μl of the dsDNA/SYBR Gold mix-containing assay buffer. Fluorescence intensity was measured on a plate reader (Synergy HT, BioTek) using FITC 485/535 filters every 2 minutes. Fluorescence intensity was normalized to T = 0 minutes.

High-throughput compound screen and analysis

TREX1 DNase assay was optimized for the high-throughput compound screen. Multiple rounds (> 5) of mock screen were performed prior to each high-throughput screening. For the primary screening, 50 nl of 5 mM compound (or DMSO) was dry-spotted to a black 384-well plate using Echo acoustic liquid handler (E650, Beckman) with a robotic control system. 3 nM recombinant human TREX1 was transferred to the plate in 20 μl of lysis buffer (20 mM Tris-HCl pH 7.4, 5 mM MgCl₂, 2 mM DTT, 100 μg/ml BSA, and 1% NP-40) by a liquid handler (MultiFlo FX, BioTek). Enzyme and compounds were incubated for 30 minutes at room temperature. ssDNA/SYBR Gold mix was added in 30 μl of assay buffer (20 mM Tris-HCl pH 7.4, 5 mM MgCl₂, 2 mM DTT, 100 μg/ml BSA, 31.6 nM 30-mer ssDNA and 0.72x SYBR Gold). The plate was centrifuged at 1000 rpm for 1 minute to remove bubbles. Fluorescence intensity was measured on a plate reader (EnVision, PerkinElmer) using FITC 485/535 filters at T = 0, T = 10, T= 20, and T = 30 minutes. Primary hits were selected for the confirmation screening. The confirmation screening was performed similarly to the primary screening except that compounds in the confirmation screen were assessed at 4 concentrations (0.25, 1, 5, 10 μM) in triplicates.

Z’ factor was calculated for each plate to ensure the quality and robustness of the assay. Z’ factor describes how well the positive and negative controls are separated. Z’ factor is calculated as follows: ${\mathrm{Z}}^{\mathrm{\text{'}}}factor=1-\frac{3\times ({\sigma }_p+{\sigma }_n)}{\left|{\mu }_p-{\mu }_n\right|}$, where ${\sigma }_p$ and ${\sigma }_n$ are the standard deviations of the positive and negative controls, respectively. ${\mu }_p$ and ${\mu }_n$ are the means of the positive and negative controls, respectively. Each screen was analyzed based on the slope, which was calculated by a “Least Squares Fit” algorithm built within Genedata Screener from the first three time points. Slope values were normalized to the median values of compounds in the same plate. A robust Z (RZ) score was calculated as follows: $\mathrm{R}\mathrm{Z}\mathrm{s}\mathrm{c}\mathrm{o}\mathrm{r}\mathrm{e}=\frac{condensedactivityoftriplicates-medianofcompounds}{relativestandarddeviationofcompounds}$, where condensed activity is the most representative single value of the triplicates using a strategy called Robust Condensing. The RZ score threshold of an effective inhibitor or activator is > 3 or < −3, respectively. A relative RZ score was calculated by normalization of the RZ score from each compound to the highest RZ score in the screening. The half-maximal activity concentration (qAC₅₀) was calculated after curve fitting for two-point normalized activity in triplicates at 4 doses. A cytotoxicity screen was performed using the cell proliferation assay described below with the addition of 10 μM of selected compounds (or equal volume of DMSO).

CRISPR-Cas9 and lentiviral transduction

LentiCRISPRv2 vector was used to generate knockout cell lines. Target guide sequences are: Trex1 #1: AGTGCCCAGACCGCCCCGTG, #2: TCAACGCTTCGATGACAACC; Cgas: GCGGACGGCTTCTTAGCGCG; Sting1: GTCCAAGTTCGTGCGAGGCT; Tbk1: CATAAGCTTCCTTCGCCCAG; Irf3: GGCTGGACGAGAGCCGAACG. Lentiviruses carrying the sgRNA and Cas9 were generated in HEK293T cells with the packaging plasmid psPAX2 and the envelope plasmid pMD2.G. Cells were selected with antibiotics after transduction for several days. Single cell clones were selected and verified by immunoblot for CRISPR-Cas9 knockout cell lines.

Immunoblot

The following primary antibodies were used for immunoblot analysis: anti-TREX1 (sc-133112, Santa Cruz, 1:50), anti-cGAS (83623, CST, 1:1000), anti-STING (13647, CST, 1:1000), anti-TBK1 (3504, CST, 1:1000), anti-IRF3 (4302, CST, 1:1000), anti-α-Tubulin (T5168, Sigma, 1:20,000), and anti-GAPDH (5174, CST, 1:1000). Secondary antibodies used included goat anti-Rabbit IgG-HRP conjugate (1706515, Bio-Rad, 1:3000) and goat anti-Mouse IgG-HRP conjugate (1706516, Bio-Rad, 1:3000). Cell pellets were lysed with 1× RIPA lysis buffer supplemented with 1× protease inhibitor cocktail (Roche, 11836170001). For mouse tissues, all collected tissues were cut into small pieces and homogenized with Zirconia/Silica beads (11079110z, BioSpec) in 1× RIPA lysis buffer using the Precellys 24 homogenizer (Bertin). Protein concentration was determined by BCA assay (23227, ThermoFisher). Protein samples were prepared in 5× Protein Loading Buffer (EC-887, National Diagnostics) diluted in 1× RIPA lysis buffer. Equal amounts of protein were separated by SDS-PAGE and transferred to a nitrocellulose membrane. Membranes were blocked in 5% milk in TBST buffer for 1 hour at room temperature and then incubated in primary antibody diluted in 3% BSA in TBST at 4 °C overnight. After washing, the membranes were incubated with HRP-conjugated secondary antibody for 1 hour at room temperature. Membranes were developed with SuperSignal^™ West Pico PLUS Chemiluminescent Substrate (34580, ThermoFisher) and imaged on BioRad ChemiDoc.

RNA isolation and qRT-PCR analysis

Total RNA was extracted with TRI reagent according to the manufacturer’s protocol (T9424, Sigma-Aldrich). cDNA was generated using iScript Reverse Transcription Supermix (1708840, Bio-Rad). qRT-PCR was performed using iTaq Universal SYBR Green Supermix (1725120, Bio-Rad), CFX Connect Real-Time PCR Detection System (1855201, Bio-Rad) and Bio-Rad CFX Maestro software (12013758, Bio-Rad).

cGAMP ELISA

cGAMP in the cell culture supernatant (with the addition of 50 μM ENPP1 inhibitor STF-1084, a gift from Dr. Lingyin Li, Standford University ⁴⁰ and cell lysate was measured by a 2’3’-cGAMP ELISA Kit (501700, Cayman Chemical) according to manufacturer’s instructions. The absorbance of the ELISA plate was measured by a plate reader (Synergy HT, BioTek) using Gen 5 software (BioTek).

Cell proliferation

Cell proliferation was quantified by confluence measurement using the IncuCyte S3 Live-Cell Analysis System (Sartorius). Cells were seeded in 96-well plates with or without compound treatment. Each well was imaged continuously at 2-hour intervals and the confluence in each well was quantified by the IncuCyte Base Analysis Software.

Molecular docking of TREX1 inhibitor

The crystal structure of human TREX1 DNase domain (PDB: 7TQN) was taken from a Protein Data Bank entry and used for protein-ligand blind docking with compound #296 using the CB-Dock2 server ^41,42.

In vivo tumor study

For tumor studies with TREX1 inhibitors, 6- to 10-week-old wild-type or Ifnar1^−/− C57BL/6J mice were used. MC38 colon adenocarcinoma cells were grown as detailed above, freshly harvested when approximately 50% confluent, and resuspended in PBS. 5×10⁵ cells in 100 μl PBS per mouse were injected subcutaneously into the right flank on day 0. TREX1 inhibitor treatment was initiated when average tumor size reached 40 mm³ (~ day 9). Before the treatment, mice with no tumor or particularly larger tumors (>100 mm³) were excluded. Compounds #296 and #289 were prepared as 30 mM working solution in 10% DMSO, 40% PEG 400 (PX1286B, Sigma-Aldrich), 20% absolute ethanol, and 30% 0.9% NaCl. Mice were peritumorally injected with 100 μl vehicle or 30 mM inhibitors (equivalent to 3 μmol per mouse) once daily for 3 consecutive days, or 2 cycles of 2 injections spaced 1 week apart (a total of 4 injections). For IFN and ISG analysis, mice were peritumorally injected with 100 μl vehicle or 10 mM inhibitors (equivalent to 1 μmol inhibitors per mouse) once daily for 3 consecutive days. Tumors were dissected 1 day after the last inhibitor treatment for qRT-PCR analysis. For cGAMP combination therapy on MC38 tumors, mice were peritumorally injected with vehicle or #296 on days 9, 12, 15, and Lipofectamine 2000 (11668027, Invitrogen) or 10 mg cGAMP/Lipofectamine 2000 prepared in 100 μl Opti-MEM (31985070, Gibco) on days 10, 13, and 16. For anti-PD-1 combination therapy on B16-F10 melanomas, mice were inoculated with 4×10⁵ B16-F10 cells the same way as described above on day 0. Mice were peritumorally injected with vehicle or #296 on days 9, 12, 14, and 200 μg isotype control (BE0089, BioXCell) or anti-mouse PD-1 (BE0146, BioXCell) prepared in 100 μl PBS on days 10, 13, 15, and 17. For all tumor studies, tumor size was measured blindly using a caliper and calculated as length × width × height/2. Mice were sacrificed when tumor volume reached 2000 mm³, or sever ulceration was observed.

For MC38 tumor experiments in wild-type C57BL/6J, Sting1^−/−, Rag1^−/−, Ifnar1^−/−, and NSG mice, 6- to 10-week-old littermates of the same sex were randomly assigned to experimental groups. 5×10⁵ MC38 or derivative knockout cells were prepared in 100 μl PBS as described above and injected subcutaneously into the right flank on day 0. For MC38 autologous cancer vaccine experiments, Trex1^KO MC38 cells were prepared the same way as above but 5×10⁵ Trex1^KO MC38 cells in 100 μl per mouse were injected subcutaneously into the left flank on day −7, day 0, or day 3. Mice were subcutaneously challenged with 5×10⁵ wild-type MC38 cells in 100 μl PBS per mouse on the right flank on day 0. For B16-F10 autologous cancer vaccine experiments, Trex1^KO B16-F10 cells were prepared the same way as described above but 4×10⁵ Trex1^KO B16-F10 cells in 100 μl per mouse were injected subcutaneously into the left flank on days −7, −4, and −1. Mice were subcutaneously challenged with 2×10⁵ wild-type B16-F10 cells in 200 μl PBS per mouse intravenously on day 0. Mice were sacrificed on day 15 for analysis of lung metastasis. Lung metastasis area on H&E-stained sections were quantified using QuPath version 0.4.3.

For tumor experiments with inducible Trex1 knockout mice, 6- to 10-week-old Trex1^fl/fl and Trex1^fl/flUBC^CreERT2 littermates of the same sex were used. For subcutaneous tumor experiments, 5×10⁵ MC38, E0771, or LLC cells in 100 μl PBS per mice were injected subcutaneously into the right flank of Trex1^fl/fl or Trex1^fl/flUBC^CreERT2 littermates on day 0, followed by intraperitoneal injection of 75 mg/kg body weight tamoxifen (T5648, Sigma-Aldrich) prepared in corn oil (20 mg/ml stock) once daily for 5 consecutive days. For treatment of established MC38 tumors, mice were treated as detailed above except that tamoxifen was administered on days 12 through day 16. For B16-F10 metastatic melanoma experiments, 2×10⁵ B16-F10 cells in 200 μl of PBS were delivered intravenously on day 0, followed by 5 consecutive days of tamoxifen treatment as described above. Mice were sacrificed on day 22 for analysis of lung metastasis. Lung metastasis area on H&E-stained sections were quantified using QuPath version 0.4.3.

Flow cytometry

Mouse spleens were crushed and grinded on a 70 μM cell strainer with a 5 ml syringe plunger and washed with 5 ml RPMI + 10% FBS. The strainer was washed with an additional 10 ml of RPMI media. Mouse tumors were harvested, minced, and filtered through a 70 μm cell strainer followed by digestion in HBSS (with calcium and magnesium) containing 1 mg/ml Collagenase IV (C5138, Sigma) and 0.1 mg/ml DNase I (11284932001, Roche) for 40 minutes at 37°C with rotation. The digestion reaction was quenched with the addition of 10 mM EDTA. Red blood cells (RBCs) were removed using ACK (ammonium-chloride-potassium) RBC lysis buffer. Cells were resuspended in FACS buffer (2% FBS in PBS, 2 mM EDTA) and strained through a 70 μm cell strainer. Cell counts were quantified using an automated cell counter. Equal number of cells were stained with live-dead Zombie Aqua dye (423102, Biolegend) followed by blocking Fc receptors using TruStain FcX antibody (101320, Biolegend) prior to staining with cell surface antibodies. The cell surface antibodies used were as follows: Alexa Fluor^® 700 anti-mouse CD3 antibody (100216, Biolegend, 1:200), Brilliant Violet 421^™ anti-mouse CD4 antibody (100443, BioLegend, 1:400), FITC anti-mouse CD4 antibody (100510, BioLegend, 1:400), APC/Cyanine7 anti-mouse CD8a antibody (100714, BioLegend, 1:200), PE anti-mouse/human CD11b antibody (101208, BioLegend, 1:200), PE anti-mouse CD25 antibody (101904, BioLegend, 1:200), APC anti-mouse/human CD44 antibody (103012, BioLegend,1:600), APC anti-mouse CD45 antibody (103112, BioLegend, 1:400), PE/Cyanine7 anti-mouse CD62L antibody (104418, BioLegend,1:600), PerCP/Cyanine5.5 anti-mouse CD69 antibody (104522, BioLegend, 1:400), Brilliant Violet 421^™ anti-mouse CD279 (PD-1) antibody (135221, BioLegend, 1:400), PE/Cyanine7 anti-human/mouse Granzyme B Recombinant Antibody (372213, BioLegend, 1:200), PerCP/Cyanine5.5 anti-mouse Ly-6A/E (Sca-1) antibody (108124, BioLegend, 1:200). Cells were stained in FACS buffer with indicated antibodies for 30 minutes at 4°C and washed with FACS buffer. Data was collected on Beckman-Coulter CytoFLEX and analyzed using FlowJo software (version 10.10.0).

Serum cytokine array and CRP ELISA

Mouse serum was isolated by centrifuging clotted blood at 10,000×g for 10 minutes. Serum was immediately frozen at −80°C until further analysis. Mouse serum was subjected to Bio-Plex Pro Mouse Cytokine 23-plex assay (M60009RDPD, Bio-Rad) for serum cytokine analysis at the UTSW Microarray & Immune Phenotyping Core. Serum CRP levels were measured by a Mouse C-Reactive Protein/CRP ELISA Kit (MCRP00, R&D Systems) according to manufacturer’s instructions. The absorbance of the ELISA plate was measured by a plate reader (Synergy HT, BioTek) using Gen 5 software (BioTek).

Tissue histology and immunohistochemistry

Mouse tissues collected for histopathology staining were taken from mice after whole body perfusion with PBS by cardiac puncture. Heart, skeletal muscle, skin, liver, and spleen were directly fixed for 24 hours at 4°C in 4% paraformaldehyde (PFA) and then switched to PBS. Lungs were flushed of red blood cells by injection of 3-5 ml PBS into the right cardiac ventricle, followed by intratracheal inflation with 1 ml 4% PFA. Lungs were transferred to a 50 ml conical containing 4% PFA for 1 hour at room temperature and then switched to PBS at 4°C. Paraffin embedding, sectioning, and hematoxylin and eosin (H&E) staining of mouse tissues were performed at the UTSW Tissue Management Shared Resource Core and UTSW Histo Pathology Core. Briefly, paraffin sections were deparaffinized in xylene and serial ethanol followed by H&E staining. CD45 immunohistochemistry (IHC) of tissues and tumors were performed at the UTSW Tissue Management Shared Resource Core with anti-CD45 Rabbit mAb (70257, CST, 1:200). Images were captured using a Hamamatsu Nanozoomer 2.0HT slide scanner. Tumor CD3, CD4, CD8, F4/80, and Ly6G IHC analyses were performed by HistoWiz (Brooklyn, New York).

Quantification and statistical analysis

Statistical tests and exact values of n are noted in figure legends. All analyses were performed using GraphPad Prism 10 software. Statistical significance was identified with P < 0.05. Tissue H&E and IHC images were analyzed using QuPath version 0.4.3. Flow cytometry data were analyzed using FlowJo version 10.10.0. Schematic diagrams were created with BioRender.com.

Key resources

Key resources (antibodies, chemicals and recombinant proteins, commercial assays, experimental models, oligonucleotides, recombinant DNA, and software and algorithms) are listed in Supplementary Table S1.

Data availability

The data generated in this study are available upon request from the corresponding author.

High-throughput small-molecule inhibitor screen of TREX1.

To discover small-molecule inhibitors of TREX1 catalytic activity, we performed a high-throughput compound screen using a cell-free DNase assay. We labeled a 30-mer ssDNA with SYBR Gold, which only emits fluorescent signal when stably incorporated into ssDNA. Therefore, ssDNA degradation will lead to gradual loss of fluorescence signal (Fig. 1A). Incubation of ssDNA/SYBR with recombinant human (rh) TREX1 (DNase domain only) led to a gradual reduction of fluorescence compared to mock-treated ssDNA/SYBR, reaching a plateau at around 40 minutes (Fig. 1B). As a comparison, DNase I (prefers dsDNA) did not cause substantial degradation of ssDNA up to 60 minutes. We also observed a dose-dependent decrease of DNA degradation kinetics with decreasing amounts of rhTREX1 enzyme (Fig. 1C). TREX1 belongs to the DnaQ-like exonuclease family and binds to metal cation cofactor Mg²⁺ which assists catalysis. Chelation of Mg²⁺ by ethylene diamine tetra-acetic acid (EDTA) completely inhibited the enzymatic reaction (Fig. 1C).

We next used this assay to perform several mock screens with small libraries of compounds, then a full library screen containing 295,680 compounds in 384-well format (Fig. 1D). The UT Southwestern compound library contains small molecules from commercial sources, such as ChemDiv, ChemBridge, ComGenex, Prestwick Chemical, and Timtec. These compounds passed 48 structure-based filters that identified undesirable characteristics as well as satisfied a relaxed set of Lipinski’s rules for good bioavailability. The library also contains 2,500 compounds from UT Southwestern chemists and about 7,100 partially purified natural product fractions.

We performed the primary screen of compounds at 5 μM and identified 3,314 primary hits with an absolute RZ score > 3 (Fig. 1D). After removing auto-fluorescent compounds and pan-assay interference compounds (PAINS), which are small molecules frequently identified as false positive hits, 1,920 compounds remained as potential TREX1 inhibitors or activators. We next performed a confirmation screen of these primary hits at 4 concentrations (0.25, 1, 5, 10 μM) in triplicates (Fig. 1E). 180 potential TREX1 inhibitors were confirmed (RZ score > 3 at 5 μM), of which 41 compounds showed robust inhibition across the tested concentration range (qAC₅₀ < 10 μM). Finally, using a cell proliferation assay, we identified 35 potent TREX1 inhibitors with low cytotoxicity at 10 μM (Fig. 1F and Supplementary Table S2). The top three candidates are: #157 (IC₅₀=0.12 μM), #296 (IC₅₀=1.81 μM), and #289 (IC₅₀=2.85 μM) (Fig. 1G).

Next, 19 candidates with known chemical structures were chemically synthesized and examined in a secondary functional screen to test their ability to induce IFN-I response in MC38 mouse colorectal carcinoma cells. We also included two TREX1 inhibitors found in the public domain, Compound I (Cmp I) and C16 ^43,44, as comparisons. We treated MC38 cells with each compound at 3 concentrations (10, 50, 100 μM) or at a single concentration (20 μM) if cytotoxic at higher concentrations, then measured expression of 5 representative immune genes induced by the cGAS-STING pathway (Ifnb1, Cxcl10, Ccl5, Tnf, Il6) at 24, 48, and 72 hours. Compound #296 and #289 induced robust expression of cGAS-STING target genes as soon as 24 hours in a dose-dependent manner (Fig. 1H and 1I), while Compound I only induced a mild IFN-I response after 48 hours. Compound #181, #222, #530, #534, and #736 induced expression of Il6, but not IFN genes or interferon-stimulated genes (ISGs). Further, compound #296 and #289 showed very low cytotoxicity even at high concentrations, with less than 5% inhibition of cell proliferation at 100 μM (Supplementary Fig. S1A and B). Therefore, we select #296 and #289 as lead compounds that not only inhibit TREX1 DNase activity at low μM concentrations but also induce robust IFN-I response in cancer cells.

Compound #296 induces antitumor immunity and synergizes with immune checkpoint therapy.

To determine the in vivo antitumor activity, we treated established MC38 tumors with compound #296 and #289 by peritumoral injection (once per day for 3 consecutive days) and then dissected tumors 24 hours after the last injection to analyze IFN gene expression in tumors. Both compounds #296 and #289 induced expression of Ifnb1 and ISGs in treated tumors (Supplementary Fig. S1C and D). Compound #296 showed a stronger efficacy than #289, consistent with in vitro studies. We next implanted MC38 tumors subcutaneously into wild-type C57BL/6J mice; after tumor grew to a palpable size (day 9), we randomized mice into three groups and treated them with vehicle, compound #296 or #289 by peritumoral injection (once per day for 3 consecutive days). Both compounds significantly reduced tumor growth and extended overall survival of mice compared to the vehicle-treated group, although compound #296 exhibited higher antitumor efficacy (Supplementary Fig. S1E and F). To determine whether the compound-induced antitumor response depends on IFN-I signaling, we also injected compound #296 and #289 in MC38 tumors established in Ifnar1^−/− mice. Neither compound inhibited tumor growth nor improved survival in Ifnar1^−/− mice (Supplementary Fig. S1G and H).

We next characterized compound #296 further. We compared rhTREX1 to DNase I in a cell-free DNA degradation assay. We used a dsDNA substrate with 3’ overhang that can be efficiently degraded by both rhTREX1 and DNase I. Compound #296 inhibited TREX1 activity at IC₅₀ = 0.4630 μM and has no inhibitory activity on DNase I up to 2 μM (Fig. 2A). We also performed molecular docking to determine the interaction between human TREX1 DNase domain (PDB: 7TQN) and compound #296. The docking structure indicated that compound #296 was anchored to the catalytic pocket of human TREX1 by three hydrogen bonds (Fig. 2B). These data suggest that our lead compound #296 is a selective inhibitor of TREX1.

To understand how compound #296 inhibits tumor growth, we increased the dosing regimen to two treatment cycles spaced one week apart, which resulted in slightly better control of MC38 tumors than single cycle treatment (Fig. 2C, Supplementary Fig. S1E). Flow cytometric analysis of the tumor-infiltrating immune cells (TILs) revealed a higher percentage of CD45⁺ cells, along with a greater number of CD45⁺ immune cells and CD11b⁺ myeloid cells per gram of tumor in #296-treated tumors compared to vehicle-treated tumors (Fig. 2D–E; Supplementary Fig. S2A). Although no significant change in CD3⁺ T cell numbers was observed, #296-treated tumors showed a significant increase in CD8⁺ T cells (% of CD3⁺), a decrease in CD4⁺ T cells (% of CD3⁺), and a higher CD8⁺ to CD4⁺ ratio (Fig. 2F–I; Supplementary Fig. S2A and B). Expression levels of PD-1, granzyme B (GZMB), and interferon-inducible stem-cell antigen-1 (SCA-1) in CD8⁺ T cells did not significantly change, although we observed a trend toward increased SCA-1 expression in both CD8⁺ T cells and CD11b⁺ myeloid cells, suggesting a more activated and stem-like phenotype (Fig. 2J–M; Supplementary Fig. S2B and C). Immunohistochemistry (IHC) analysis further supported these findings, showing that #296-treated tumors had higher infiltration of CD8⁺ T cells not only in the tumor core but also in invasive margins and necrotic areas compared to vehicle-treated tumors (Supplementary Fig. S3A–C). In contrast, macrophage (F4/80⁺) infiltration was reduced in both the tumor core and invasive margins, and macrophage (F4/80⁺) and neutrophil (Ly6G⁺) signals were decreased in necrotic areas. Collectively, these results indicate that compound #296 enhances antitumor immunity by promoting immune cell infiltration, recruiting cytotoxic T cells, and remodeling the spatial organization of tumor-associated macrophages and neutrophils.

We also evaluated the potential of combining compound #296 with other therapies. Combination of #296 and cGAMP in MC38 colon adenocarcinoma did not further enhance therapeutic efficacy (Fig. 2N). This is likely because TREX1 acts upstream of cGAS-STING activation, rendering additional STING stimulation less impactful. To test whether TREX1 inhibition can synergize with immune checkpoint inhibitors (ICIs), we conducted combination therapy on B16-F10 melanoma, which is known to be resistant to ICIs ^45,46. Combination of #296 and anti-PD-1 antibody significantly inhibited B16-F10 tumor growth whereas #296 or anti-PD-1 monotherapy and control treatment did not (Fig. 2O). These findings suggest that TREX1 inhibitors, such as compound #296, are most effective on immunologically cold tumors when used in combination with existing ICIs.

We next evaluate immune toxicity of compound #296 as a monotherapy and combination therapy with anti-PD-1. Serum cytokine and chemokine analysis of B16-F10 tumor-bearing mice treated with vehicle or #296 showed that the majority of the cytokines and chemokines in the 23-plex panel remained unchanged. Only IL-12p40 was significantly elevated in #296-treated mice, which was further increased when combined with anti-PD-1 therapy (Supplementary Fig. S4A–D). MIP-1α (CCL3), MIP-1β (CCL4), RANTES (CCL5), and MCP-1 (CCL2) were induced by #296, but were not furthered increased by anti-PD-1 combination (Supplementary Fig. S4C and D). Importantly, none of the treatment groups increased serum C-reactive protein (CRP) levels as a clinically relevant maker of inflammation (Supplementary Fig. S4E and F). In fact, #296 treatment decreased serum CRP levels, likely due to reduced tumor burden. Furthermore, IHC analysis of CD45⁺ immune cell infiltration in lung and colon tissues showed no increase in either monotherapy or combination therapy groups (Supplementary Fig. S4G and H), consistent with the absence of systemic inflammation. Together, these data suggest compound #296 is a selective TREX1 SMI that induces antitumor immunity and synergizes with immune checkpoint therapy.

TREX1 deficient cancer cells as an autologous cancer vaccine.

In vivo stability and pharmacology of compound #296 require further optimization in the future. To explore other modalities of targeting TREX1 as cancer immunotherapy, we next used genetic models to delete the Trex1 gene either in cancer cells or conditionally in adult mice. We first generated Trex1-knockout (Trex1^KO) MC38 cells and confirmed the mechanism of immune activation (Supplementary Fig. S5A). Trex1^KO MC38 cells showed significantly elevated expression of IFN genes (Ifnb1), interferon-stimulated genes (ISGs, Ifi27l2a, Cxcl10), and inflammatory genes (Tnf, Ccl5) at resting state compared to non-targeted wild-type (Trex1^WT) MC38 cells (Fig. 3A). The induction of tonic IFN signaling was driven by cGAS-STING activation, as further knockout of Cgas or Sting1 abolished IFN and ISG expression (Supplementary Fig. S5B). No differences in proliferation between Trex1^WT and Trex1^KO MC38 cells were observed (Fig. 3B). After subcutaneous injection into wild-type C57BL/6J mice, Trex1^KO MC38 tumors grew significantly slower than Trex1^WT tumors (Fig. 3C). Five out of ten Trex1^KO tumors underwent rapid regression after initial growth and then the mice remained tumor-free throughout the 100-day duration of experiment (Fig. 3C and D). These data are consistent with recently published studies^38,47,48.

To explore the mechanism of antitumor immunity induced by Trex1^KO MC38 cells, we first considered a possibility that Trex1^KO MC38 cells may activate tumor cGAS and then produce and secrete cGAMP to activate STING in surrounding host immune cells^40,49. We observed significantly higher cGAMP level in both culture supernatant and lysate of Trex1^KO MC38 cells compared to Trex1^WT MC38 cells (Supplementary Fig. S5C and D). Surprisingly, Trex1^KO MC38 tumors still grew significantly slower compared to Trex1^WT tumors in Sting1^−/− mice with higher rate of survival (Fig. 3E and F). In comparison, we did not observe any difference in tumor growth or overall survival when Trex1^WT or Trex1^KO MC38 cells were implanted in Rag1^−/− mice, NOD-scid IL2Rgamma^null (NSG) mice, or Ifnar1^−/− mice (Fig. 3G–L). These data suggest that Trex1^KO MC38-induced antitumor immunity signals through host IFN-I receptors and adaptive immunity but does not require host STING.

We next generated Trex1^KOCgas^KO, Trex1^KOSting1^KO, Trex1^KOTbk1^KO, and Trex1^KOIrf3^KO MC38 cells to investigate the cell-intrinsic pathway that initiates antitumor immunity (Supplementary Fig. S5E). When implanted into wild-type C57BL/6J mice, 9/10 Trex1^KO tumors were rejected as expected (Fig. 4A–C). However, additional knockout of Cgas, Sting1, Tbk1, or Irf3 increased tumor growth and decreased overall survival to levels comparable to Trex1^WT tumors (Fig. 4A–C). These results suggest that tumor-intrinsic cGAS-STING signaling is necessary for mounting an effective host antitumor immune response to TREX1-deficient tumors.

To understand the impact of Trex1 knockout cancer cells on the tumor microenvironment, we assessed the population and functional status of tumor-infiltrating immune cells in Trex1^KO MC38 tumors by flow cytometry. Compared to Trex1^WT tumors, Trex1^KO tumors displayed higher percentages of CD45⁺ immune cells and increased numbers of CD45⁺ cells, CD11b⁺ myeloid cells, and CD3⁺ T cells per gram of tumor (Fig. 4D–F; Supplementary Fig. S2A). Additionally, Trex1^KO tumors exhibited an elevated proportion of CD8⁺ T cells (% of CD3), reduced CD4⁺ T cells (% of CD3), a higher CD8⁺ to CD4⁺ ratio, and increased expression of cytotoxic and activation markers on CD8⁺ T cells, including GZMB and PD-1 (as a percentage and MFI; Fig. 4G–K; Supplementary Fig. S2B). Furthermore, Trex1^KO tumors showed elevated SCA-1 expression in both CD8⁺ T cells and CD11b⁺ myeloid cells (as a percentage and MFI), indicating enhanced activation and interferon exposure (Fig. 4L and M; Supplementary Fig. S2B and C). These findings suggest that TREX1 loss in tumors elicits antitumor immunity by increasing immune cell infiltration, promoting T cell recruitment and activation, boosting cytotoxicity, and inducing immune-modulatory changes in both T cells and myeloid cells.

To explore other modalities of TREX1-based cancer immunotherapy, we next examined whether Trex1^KO tumor cells can be used as a cancer vaccine to induce immune protection against autologous wild-type tumor challenge. We inoculated wild-type C57BL/6J mice subcutaneously with 5×10⁵ Trex1^KO MC38 cells (“vaccine”) on the left flank and challenged the right flank with 5×10⁵ wild-type MC38 cells (“tumor”). We tested three different schemes: “vaccine” administered 7 days prior to “tumor” challenge, or at the same day, or 3 days after “tumor” challenge (Fig. 4N). Trex1^KO cells administration 7 days prior protected wild-type tumor challenge at a distal site, with regression of 7/9 tumors (Fig. 4O and P). However, Trex1^KO MC38 cells administration on the same day or 3 days after wild-type tumor inoculation did not confer significant protection (Fig. 4O and P). The complete responders remained tumor-free for the entire 100-day duration of the experiment (Fig. 4P). We further tested whether Trex1^KO cancer cells can be used as a vaccine to prevent tumor metastasis using the syngeneic B16-F10 melanoma model (Fig. 4Q). Compared to naïve non-vaccinated mice, three subcutaneous injections of 4x10⁵ Trex1^KO B16-F10 cells administered over one week prior to a wild-type B16-F10 tumor challenge via tail vein (a route that results in rapid lung metastasize) significantly reduced the metastatic burden in the lungs (Fig. 4R). Taken together, these data provide a proof-of-concept for a cell-based cancer vaccine design that harnesses TREX1 inactivation to confer protection against future challenges with autologous tumors and tumor metastases.

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

We next established an inducible whole-body Trex1 knockout mouse model to simulate “on-demand” systemic TREX1 inactivation in adult mice that resembles systemic pharmacological TREX1 inhibition. We bred Trex1-flox (Trex1^fl/fl) mice with UBC^CreERT2 to generate Trex1^fl/flUBC^CreERT2 mice (we call iTrex1^KO). Tamoxifen injection substantially reduced TREX1 protein level 4 weeks later in various tissues, such as bone marrow, spleen, and heart, demonstrating efficient gene deletion (Fig. 5A and B). We first characterized the impact of inducible whole-body Trex1 deletion in mice as a treatment using three experimental schemes.

First, early treatment of solid tumors. iTrex1^KO mice and Trex1^fl/fl littermates were subcutaneously implanted with wild-type MC38 tumors, immediately followed by tamoxifen injections to induce Trex1 knockout (Fig. 5A). Remarkably, MC38 tumor grew quickly in Trex1^fl/fl mice but tumor growth was nearly completely suppressed in iTrex1^KO mice, which rejected 8/10 tumors (Fig. 5C and D). We also observed antitumor activity against two other solid tumor models: E0771 breast carcinoma and Lewis lung carcinoma (LLC; Fig. 5E–G), although the inhibitory effect on immunologically “cold” LLC tumor was weaker than that on immunologically “hot” MC38 and E0771 tumors.

Second, early treatment of metastatic tumors. We intravenously injected mice with B16-F10 melanoma cells, immediately followed by tamoxifen injections to induce Trex1 knockout (Fig. 5H). iTrex1^KO mice had reduced lung metastasis compared to littermate Trex1^fl/fl mice as measured by significantly reduced number of metastatic foci on the surface of lung (metastasis area on hematoxylin and eosin (H&E) -stained lung sections was also reduced but not statistically significant Fig. 5I–K).

Third, late treatment of solid tumors. We implanted MC38 tumors, waited 12 days to allow tumors to reach an average size of 170 mm³, then injected tamoxifen to induce Trex1 knockout (Fig. 5L). The tumor burdens of iTrex1^KO mice were significantly lower than those of Trex1^fl/fl littermates, albeit less effective compared to early treatment (Fig. 5M). Together, these data demonstrate that inducible whole-body Trex1 deletion in adult mice results in potent antitumor response that can effectively control hot and cold solid tumors as well as metastatic tumors.

To understand the antitumor mechanism of Trex1 inactivation in adult mice, we next performed immune cell profiling of iTrex1^KO mice and Trex1^fl/fl littermates. We first measured tonic IFN-I signaling in splenocytes 3 months after tamoxifen injection. We observed significant upregulation of IFN gene (Ifnb1) and ISG (Cxcl10) in splenocytes isolated from iTrex1^KO mice compared to those from Trex1^fl/fl littermates (Fig. 6A). CD4⁺ T cells shifted from naïve to effector memory cells in iTrex1^KO mice compared to Trex1^fl/fl littermates, and CD8⁺ T cells shifted from naïve to central memory cells (Fig. 6B and C; Supplementary Fig. S6A–C). We also observed increased activation markers CD69 and CD25 on CD4⁺ T cells in iTrex1^KO mice compared to Trex1^fl/fl littermates (Fig. 6D–G). We did not observe any tissue abnormalities (more below). The increased IFN-I signaling and T cell activation phenotypes are consistent with mild innate immune activation. Therefore, collectively these data suggest that low level systemic innate immune activation due to TREX1 loss is highly effective at tumor control.

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

One of the main side effects of cancer immunotherapy is immune toxicity or immune-related adverse events (irAEs)⁵⁰. Congenital loss-of-function mutations in TREX1 are associated with several autoinflammatory and autoimmune disorders including Aicardi-Goutières syndrome (AGS) due to strong activation of cGAS-STING-IFN signaling²⁶. Trex1^−/− mice also develop systemic inflammatory disease that can be rescued by Cgas^−/− or Sting1^{−/− 31,32}. However, it is important to consider that postnatal short-term TREX1 inactivation in wild-type organisms is unlikely to elicit the same biological response observed in germ-line Trex1^−/−. Therefore, we used iTrex1^KO mice as the “worst case scenario” for future TREX1 inhibitor use and characterized possible immune toxicity associated with TREX1 inactivation in adult mice.

We injected tamoxifen into 6-week-old iTrex1^KO mice and Trex1^fl/fl littermates to induce Trex1 deletion in adult mice (Fig. 7A). Then, we analyzed inflammatory cytokines and chemokines in serum 2, 4, and 26 weeks after tamoxifen-induced gene deletion. Tamoxifen-induced Trex1 knockout resulted in sustained depletion of TREX1 protein up to 26 weeks after injection (Fig. 7B). We did not observe any changes in serum inflammatory cytokines and chemokines comparing iTrex1^KO to Trex1^fl/fl littermates at early (2 and 4 weeks) time points (Fig. 7C). At the end point (26 weeks), iTrex1^KO mice only showed increased IL-12p40 (out of the 23-plex panel; Fig. 7D). No significant change in serum CRP level or weight loss was observed in iTrex1^KO mice (Fig. 7E and F). We observed a modest increase in spleen weight in iTrex1^KO mice compared to Trex1^fl/fl littermates (Fig. 7G), consistent with increased IFN-I signaling and T cell activation at 12 weeks after tamoxifen treatment (Fig. 6A and D–G). H&E stainings of major organs of 32-week-old iTrex1^KO mice (26 weeks after tamoxifen injection) were indistinguishable from those of Trex1^fl/fl littermates (Fig. 7H). CD45 IHC staining showed a slight increase in CD45⁺ area in the heart, skeletal muscle, skin, liver and lung of the iTrex1^KO mice compared to Trex1^fl/fl littermates (Fig. 7I–M). As a positive control, a 5.4-week-old germline Trex1^−/− mouse exhibited severe immune infiltration in multiple tissues (Fig. 7H–M).

We also evaluated the toxicity of combination therapy by giving iTrex1^KO mice anti-PD-1 treatment (10 mg/kg) on days 7, 10, 13, 16 after the first tamoxifen injection and analyzed serum and tissue inflammation markers on day 33. Only IL-12p40 was increased in iTrex1^KO mice, which was further increased by anti-PD-1 treatment (Supplementary Fig. S7A and B). IL-13 and Eotaxin were significantly elevated only in anti-PD-1-treated iTrex1^KO mice, but not in isotype-treated iTrex1^KO mice or anti-PD-1-treated Trex1^fl/fl control mice (Supplementary Fig. S7A and B). Anti-PD-1-treated iTrex1^KO mice showed mild increases in IL-1α, IL-2, IL-4, IL-12p70, and GM-CSF levels. Serum CRP levels did not change in any treatment groups (Supplementary Fig. S7C). Tissue IHC results demonstrated a mild increase of CD45⁺ immune cell infiltration in the lung, colon, and liver tissues of iTrex1^KO mice compared to Trex1^fl/fl control mice, which were not further increased by anti-PD-1 treatment (Supplementary Fig. S7D–F). Overall, these data suggest that TREX1 inactivation in adult mice does not induce severe immune toxicity either alone or in combination with ICI over an extended period of time. Together with the broad antitumor activity observed in the same iTrex1^KO mouse model, our data demonstrate the exciting feasibility for TREX1 antagonism to achieve antitumor efficacy with favorable immune safety.