B6 mice were inoculated with LLC tumor cells and when dermal tumors were well-established (250–300 mm³) mice were given a short course of direct (intratumoral) CDA treatment (100 µg on days 0, 2, 6). CDA treatment did not slow tumor growth immediately (day 3) but after the final CDA treatment tumor growth was reduced significantly (figure 1A). However, tumor control was transient as growth resumed in most mice, and most CDA-treated mice (75%) had to be sacrificed by 15 days after CDA treatment was initiated due to excessive tumor growth, extending survival by only ~5 days more than control mice treated with vehicle (figure 1B). CDA-treated mice that did not succumb to tumor growth (~25%) were challenged with a second bolus of LLC cells on the contralateral flank when primary lesions had healed, 60 days after CDA treatment was initiated. Secondary LLC tumors grew at the same rates as primary tumors in all mice (table 1, group B), indicating that direct CDA treatment did not incite stable protective antitumor immunity, even though primary tumors were eliminated by CDA monotherapy.

In a previous study, CDA treatment to activate STING protected 50% of mice implanted with immunogenic B16 melanomas until experimental end points and surviving mice all acquired stable protective antitumor immunity.11 As CDA treatment was initiated when B16 tumors were relatively small (~100 mm³) in this prior study, we next tested if CDA treatment was more effective in mice with smaller LLC tumors. Mice were treated with CDA as before, except that treatment was initiated when LLC tumors were first detected (90–120 mm³), 4–5 days earlier than in our initial study on mice with more established tumors. Earlier CDA treatment prevented progression of small LLC tumors, leading to rapid tumor regression and uniform survival (online supplementary figure S1A,B). In addition, earlier CDA treatment induced all mice to acquire stable protective immunity to LLC challenge (table 1, group G). Direct CDA treatment was necessary to control LLC tumor growth as intravenous CDA treatment when tumors first appeared had little impact on tumor growth and had no significant effect on survival (online supplementary figure S1C,D). Direct CDA treatment did not protect STING-deficient (STING-KO) mice with small LLC tumors (online supplementary figure S2), confirming that therapeutic responses were STING-dependent and revealing that STING expressed by LLC tumor cells was not the relevant therapeutic target of CDA.

As resistance to the antitumor effects of CDA developed rapidly as tumors became more established, we assessed if CDA treatment induced immune regulatory pathways that commonly contribute to ICPs and impede antitumor immunity, as potential causes of therapy resistance. Cells expressing programmed death ligand-1 (PD-L1) induce T cells expressing PD-1 to become functionally exhausted and/or undergo apoptosis. Like many patients with cancer, mice with LLC tumors are poorly responsive to blockade of PD-1/PD-L1 signaling.2–5 Direct CDA treatment induced rapid and significant increase in PD-L1 gene transcription in tumor lesions (figure 2A) and TDLNs (figure 2B), and elevated PD-L1 expression was sustained for at least 6 hours in both TME sites. COX2 expression increases in the TME during LLC tumor growth and COX2 inhibitors slow LLC tumor growth.13 14 Direct CDA treatment potentiated COX2 gene transcription rapidly in tumor lesions and TDLNs, and elevated COX2 transcription persisted for over 6 hours in both TME sites (figure 2C,D).

Next, we tested if direct CDA treatment induced IDO enzyme activity in tumor lesions and TDLNs by assessing production of kynurenine (Kyn), a tryptophan catabolite made by cells expressing IDO, in TME tissues from mice with LLC tumors. LLC tumor cells do not express IDO but IDO activity induced in TDLN DCs via STING/IFN-I signaling during LLC tumor growth suppresses antitumor immunity to promote optimal LLC tumor growth.12 STING agonists also induced DCs to express IDO, an attribute we exploited to attenuate autoimmune disease in mouse models of rheumatoid arthritis, multiple sclerosis and type I diabetes.15–19 Direct CDA treatment potentiated IDO activity rapidly and significantly in tumor lesions and TDLNs, relative to levels induced by LLC growth (figure 2E,F), Thus, intratumoral STING activation induced rapid increase in local expression of PD-L1, COX2 and IDO, which may attenuate the antitumor attributes of CDA.

To test if CDA-induced PD-L1 expression reinforces ICP potency mice were co-treated with CDA and anti-PD-1 monoclonal antibodies (mAbs) to block PD-L1/PD-1 interactions that inactivate effector T cells. Consistent with LLC tumors being refractory to PD-1 blockade, anti-PD-1 mAb monotherapy (150 µg/mouse, intraperitoneal, 15 min before each CDA injection) did not slow LLC tumor growth (figure 3A), and had minimal survival impact (figure 3B). As before (figure 1), direct CDA monotherapy did not prevent tumor growth in most mice (75%). Combining direct CDA treatment with PD-1 blockade led to rapid and uniform tumor regression and uniform survival until experimental end points (figure 3). No overt evidence of tumor relapse was observed in any mice co-treated with CDA and anti-PD-1 mAbs but some mice became moribund at day 60 prompting their sacrifice. On examination, all mice that became moribund had invasive and dispersed tumors originating at primary lesions and expanding into the peritoneal cavity (ie, they were not visible externally), and some mice also had distal (metastatic) tumors growing in lungs and/or liver (table 1, group C). These observations prompted sacrifice of remaining mice in this treatment group and on inspection all mice had invasive internal tumors and some had metastatic tumors. Thus, combining CDA and PD-1 blockade enhanced tumor control and survival but did not eliminate primary tumor burdens, indicating that induced antitumor immunity was transient and unstable.

Next, we tested if three different proprietary IDO inhibitor (IDOi) drugs with distinct pharmacological characteristics enhanced therapeutic responses to direct CDA treatment in the LLC tumor model. Indoximod (1-methyl-D-tryptophan (1MT)), navoximod (NLG-919) and lindrostat (BMS-986205) have all been tested in patients with cancer as monotherapies, or in combination with other anticancer treatments, with varying effects on clinical outcomes.10 Pharmacologically, lindrostat is the most effective IDO inhibitor, although many factors may impact the antitumor attributes of IDOi drugs.10 Mice with established LLC tumors were given CDA or IDOi monotherapy, or co-treated with CDA and an IDOi drug. As before, direct CDA monotherapy prevented tumor growth in only 20%–30% of mice (figure 4). Monotherapies with 1MT (2 mg/mL in drinking water days 1–10) or lindrostat (2 mg/kg, daily gavage, days 0–10) had no significant effects on tumor growth (figure 4A,E) or survival (figure 4B,F), relative to vehicle-treated mice. Navoximod (intraperitoneal, 5 mg/kg, daily from days 1–10) monotherapy slowed tumor growth and increased survival by 5–6 days (figure 4C,D).

Co-treatment with 1MT did not enhance CDA-mediated tumor control or increase survival significantly, relative to outcomes with CDA monotherapy (figure 4A,B). Navoximod synergized with CDA to further reduce tumor growth but did not increase survival significantly, relative to CDA monotherapy (figure 4C,D). No residual primary or metastatic tumors were found in mice surviving to experimental end points after co-treatments with CDA and 1MT or navoximod (figure 4B and D, table 1, groups D1, D2), as in mice given CDA monotherapy. In contrast, lindrostat co-treatment enhanced antitumor responses to CDA significantly, promoting rapid tumor regression and uniform survival to experimental end points (figure 4E,F). Nevertheless, on inspection at experimental end points (day 60) all mice in this treatment group had invasive tumors originating at primary lesions growing into the peritoneal cavity, and some mice also had metastatic tumors in lungs and liver (table 1, group D3). Thus, as in mice co-treated with CDA and PD-1 blockade, CDA and lindrostat co-treatment induced rapid tumor regression but did not eliminate primary tumor burdens, leading to tumor relapse at primary and distal sites. Collectively, our findings reveal that increased IDO activity in the TME due to STING activation attenuates antitumor responses to CDA, and lindrostat was the only one of three proprietary IDOi drugs tested that synergized with CDA to enhance antitumor responses substantively.

DCs expressing IDO in the TME suppress effector T cells and activate regulatory CD4 T cells (Tregs) to attenuate T cell immunity via PD-1-dependent mechanisms.20 To investigate potential links between these T cell regulatory pathways, we assessed if PD-1 blockade or IDOi drugs impacted IDO activity and PD-L1 expression induced by CDA treatment. PD-1 blockade had no effect on CDA-induced IDO activity in tumor lesions or TDLNs (online supplementary figure S3A,B). Lindrostat (BMS) co-treatment partially reduced CDA-induced PD-L1 expression in tumor lesions but had no significant impact on CDA-induced PD-L1 expression in TDLNs (online supplementary figure S3C,D). Thus, the IDO and PD-1 pathways make non-redundant contributions that promote resistance to STING activation independently, as disrupting either pathway enhanced antitumor responses to CDA. Since co-treatments to activated STING and block PD-1 or inhibit IDO led to eventual tumor relapse, we tested if simultaneous PD-1 and IDO blockade improved tumor control. Lindrostat plus PD-1 blockade co-treatment did not enhance tumor control or increase survival, relative to outcomes in vehicle-treated mice (online supplementary figure S3E,F), reflecting outcomes reported for the ECHO-301 clinical trial using epacadostat to inhibit IDO activity.9 As expected, combined lindrostat and PD-1 blockade enhanced tumor control and prolonged survival after CDA treatment without dermal tumor relapse (online supplementary figure S3E,F). However, on inspection at experimental end points all mice in this treatment group had invasive tumors originating from primary lesions growing into the peritoneal cavity, and some mice also had metastatic tumors (table 1, group E). Thus, simultaneous disruption of the PD-1 and IDO regulatory pathways did not improve antitumor responses to CDA, as tumor relapse still occurred in all mice given such treatments.

Next, we tested if a COX2-selective inhibitor, celecoxib, boosted antitumor responses to CDA. Mice with established LLC tumors were co-treated with CDA (dosing as before) and celecoxib (60 mg/kg, oral gavage, days 0–10). Celecoxib monotherapy had no effect on tumor growth or survival, while (as before) CDA monotherapy slowed tumor growth and improved survival outcomes but did not prevent most mice from succumbing to tumor growth (figure 5A,B). Co-treatment with celecoxib enhanced antitumor responses to CDA significantly, inciting rapid tumor regression and uniform survival without apparent tumor relapse until experimental end points (figure 5A,B). Moreover, no mice in this treatment group became moribund and all mice were challenged with LLC cells 60 days after initiating treatment to assess their immune status. No tumor growth was observed in mice surviving celecoxib and CDA treatment and no tumors were detected at primary lesions or at distal sites in other tissues on inspection 60 days after secondary LLC challenge (table 1, group F1). Thus, co-treatment with celecoxib and CDA induced durable protective antitumor immunity that eliminated primary tumor burdens and prevented tumor metastasis.

Since celecoxib and CDA co-treatments induced potent and robust antitumor immunity, we tested if this drug combination induced abscopal immunity that prevented growth of distal tumors not directly exposed to CDA. Accordingly, mice were inoculated to seed dermal LLC tumors (day 0) and a second bolus of LLC cells was administered 5 days later via intravenous injection to seed distal tumors in other tissues, including lungs. Celecoxib and CDA co-treatments were initiated when dermal tumors were established (as before) and 8/9 mice survived until experimental end points on day 30 (figure 5C). On inspection, no tumor growth was evident in lungs or other tissues of the eight surviving mice (table 1, group F2), indicating that celecoxib and CDA co-treatment stimulated abscopal immunity that targeted distal tumors not exposed directly to CDA, as well as primary dermal tumors exposed to CDA. Collectively, our findings reveal that CDA-induced COX2 poses a major barrier to eliciting protective antitumor responses after direct STING activation in the TME, as inhibiting COX2 led to fully protective antitumor responses to CDA that prevented growth of primary tumors and incited abscopal immunity to prevent growth of distal tumors not directly exposed to CDA.

To investigate how celecoxib impacts immune responses following STING activation, we analyzed expression of immune response genes in tumor lesions from CDA-treated mice with established LLC tumors. As expected, CDA treatment stimulated local transcription of genes encoding pro-inflammatory cytokines, IFN-β, IFN-γ, tumor necrosis factor (TNF)-α and interleukin (IL)-6 (figure 6A–D). Celecoxib co-treatment potentiated CDA-induced cytokine gene transcription, eliciting 10-fold to 40-fold increase in IFN-β, TNF-α and IL-6 transcription, relative to CDA monotherapy, while celecoxib monotherapy had no effect on basal cytokine gene transcription levels. Celecoxib also potentiated CDA-induced increase in matrix metallopeptidase 3 and 13 (MMP3, MMP13) transcription (figure 6E,F). Moreover, granzyme B (GZMB) transcription was elevated in tumor lesions given CDA and celecoxib co-treatments but was not elevated after CDA monotherapy (figure 6G), suggesting that activated effector T cells expressing GZMB accumulated in tumor lesions only if celecoxib was combined with late CDA treatment.

In summary, inhibiting COX2 activity was more effective in unleashing the antitumor potential of CDA than simultaneous disruption of the PD-1 and IDO pathways that commonly contribute to ICPs and promote therapy resistance in many patients with cancer and mouse tumor models. To assess if celecoxib impacts other STING-responsive immune regulatory pathways, we assessed PD-L1 transcription and IDO activity in tumor lesions and TDLNs after CDA treatment. Celecoxib co-treatment completely blocked CDA-induced IDO activity in tumor lesions and TDLNs (figure 7A,B), as IDO activity levels were comparable in untreated mice with tumors and in mice co-treated with celecoxib and CDA. Celecoxib did not block CDA-induced PD-L1 transcription after STING activation (figure 7C,D). Lindrostat treatment to inhibit IDO (figure 7E,F) and PD-1 blockade (figure 7G,H) had no significant impact on CDA-induced COX2 transcription in tumor lesions or TDLNs. Thus, the superior efficacy of celecoxib, relative to IDO inhibition and PD-1 blockade, may in part be due to effective disruption of the induced IDO pathway, although celecoxib may also disrupt other regulatory pathways responsive to CDA since lindrostat was not as effective as celecoxib in unleashing the full antitumor effects of CDA.

C57BL/6 (B6) mice were bred in a barrier facility at the Comparative Biology Centre, Newcastle University. STING-KO mice (fully backcrossed to B6 background) were described previously.38

LLC (ATCC) tumor cells were injected intradermally (2×10⁵ cells/mouse) into the right flank of female B6 mice and tumor growth was monitored. Tumor sizes were calculated using the formula V=(d1×d2)^3/2×(π/6), where d1 and d2 are perpendicular tumor diameters. Mice were treated with mixed linkage CDA ammonium salt (ML RR-S2 CDA or AKA ADU-S100; Insight Biotechnology, 100 µg/mouse, intratumorally or intravenously) or vehicle (deionized water), as described previously.11 Some mice were treated with celecoxib (Sigma-Aldrich, 60 mg/kg, oral gavage), 1MT (2 mg/mL, drinking water), navoximod (NLG-919, 5 mg/kg, intraperitoneal), kindly provided by NewLink Genetics, BMS-986205 (2 mg/kg, oral gavage), kindly provided by Bristol-Myers Squibb, anti-mPD1 (Clone RMP1-14, BioXcell), anti-IFNAR (MAR1-5A3), IgG2a isotype (clone 2A3, BioXcell) or IgG1 isotype control (MOPC-21, BioXcell) mAbs (150 µg/injection, 15 min before each CDA injection). Vehicles for BMS, navoximod and celecoxib were 5% dimethyl sulfoxide/10% Tween 20/30% polyethylene glycol/55% phosphate-buffered saline (PBS) (added in this order). 1MT was dissolved in 0.1 M NaOH, pH adjusted to 7.4 with 0.1 M HCl, 10 tablets/l of Hermesetas (Sodium Saccharin, Sucralose; Carrier: L-Leucine) sweetener added, and volume adjusted with water to give 2 mg/mL. To evaluate abscopal effects, mice were injected with 2×10⁵ cells/mouse (intradermally) and 5 days later with 5×10⁵ cells/mouse (intravenously).

IDO activity was measured in TDLN or tumor lesions as described previously.16 In brief, tissues were homogenized in PBS, added to IDO enzyme cocktails and kynurenine generated after 2 hours was measured by high-performance liquid chromatography.

RNA was extracted using Tri reagent (Sigma-Aldrich), using the manufacturer’s recommended procedure. RNA was then reverse-transcribed using a random hexamer cDNA RT kit (Clontech), and quantitative RT-PCR was performed using SsoFast EvaGreen supermix (Bio-Rad). PCR primers (murine) were as follows (5’−3’, forward and reverse);

β-actin: TACGGATGTCAACGTCACAC & AAGAGCTATGAGCTGCCTGA;

COX2: ATCATAAGCGAGGACCTGGG & CTGCAGGTTCTCAGGGATGT;

PD-L1: TGCGGACTACAAGCGAATCA & CTTCTCTTCCCACTCACGGG;

Granzyme B: GAAGCCAGGAGATGTGTGCT & GCACGTTTGGTCTTTGGGTC;

MMP3: CTATACGAGGGCACGAGGAG & CCACCCTTGAGTCAACACCT

MMP13: GGAGCCCTGATGTTTCCCAT & ATCAAGGGATAGGGCTGGGT;

TNF-α: TCGTAGCAAACCACCAAGTG & GGAGTAGACAAGGTACAACC

IL-6: AGACAAAGCCAAGT CCTTCAGAGA & GCCACTCCTTCTGTGACTCCAGC

IFN-γ: CCTTCTTCAGCAACAGCAAGGCG & CCCACCCCGAATCAGCAGCG

IFN-β1: GCAGCTGAATGGAAAGATCA & GTGGAGAGCAGTTGAGGACA

Threshold cycle (Ct) values were set in the early linear phase of amplification; relative expression of target genes were calculated as 2Ct(β-actin)−Ct(target gene).

Data were analyzed using GraphPad Prism. For tumor growth, two-way analysis of variance with Bonferroni post hoc multicomparison tests were performed. Unpaired two-tailed non-parametric Mann-Whitney U tests were performed for two group comparisons. Statistical significance for survival curves was evaluated using the log-rank test.