key: cord-0295622-c6gftlmd
authors: Jneid, Bakhos; Bochnakian, Aurore; Delisle, Fabien; Djacoto, Emeline; Denizeau, Jordan; Sedlik, Christine; Fiore, Frédéric; Kramer, Robert; Walters, Ian; Carlioz, Sylvain; Malissen, Bernard; Piaggio, Eliane; Manel, Nicolas
title: Cellular selectivity of STING stimulation determines priming of anti-tumor T cell responses
date: 2021-12-01
journal: bioRxiv
DOI: 10.1101/2021.12.01.469893
sha: d9a39a8c6eb01e813efe867e3866f056c1b057dc
doc_id: 295622
cord_uid: c6gftlmd

T cells that recognize tumor antigens are crucial for anti-tumor immune responses. Induction of anti-tumor T cells in immunogenic tumors depends on STING, the intracellular innate immune receptor for cyclic guanosine monophosphate-adenosine monophosphate (cGAMP) and related cyclic dinucleotides (CDNs). However, the optimal way to leverage STING activation in non-immunogenic tumors is still unclear. Here, we show that cGAMP delivery by intra-tumoral injection of virus-like particles (cGAMP-VLP) leads to differentiation of tumor-specific T cells, decrease in tumor regulatory T cells (Tregs) and anti-tumoral responses that synergize with PD1 blockade. By contrast, intra-tumoral injection of synthetic CDN leads to tumor necrosis and systemic T cell activation but no differentiation of tumor-specific T cells, and a demise of immune cells in injected tumors. Analyses of cytokine responses and genetic models revealed that cGAMP-VLP preferentially targets STING in dendritic cells at a 1000-fold less dose than synthetic CDN. Sub-cutaneous administration of cGAMP-VLP showed synergy when combined with a tumor Treg-depleting antibody to elicit systemic tumor-specific T cells, leading to complete and lasting tumor eradication. These finding show that cell targeting of STING stimulation shapes the anti-tumor T cell response and reveal a therapeutic strategy with T cell modulators.

Antigen-presenting cells are activated by innate immune signals emanating from germline-44 encoded pattern recognition receptors that recognize non-self or altered-self molecules. STING is 45 an intracellular pattern recognition receptor for cyclic dinucleotides (CDNs) implicated in the 46 response to bacteria and to intracellular DNA of foreign and altered-self origins. In mouse 47 7 We next measured tumor growth. ADU-S100 and cGAMP-VLP were tested with or 123 without anti-PD1 to assess the impact of immune checkpoint inhibition on the response. cGAMP-124 VLP induced a delay in tumor growth (Figure 1C, 1D) . Adding anti-PD1 enhanced this delay 125 and led to complete responses in a subset of mice ( Figure 1C) . In comparison, ADU-S100 126 induced a delay in tumor progression and some complete responses, but there was no additive 127 effect of anti-PD1. 2'3'-cGAMP alone or co-injected with VLP induced a smaller tumor growth 128 delay and no complete responses were observed. Empty VLP had no effect. Similar trends were 129 observed on mouse survival (defined in this study as the time until the ethical endpoint of 2000 130 mm 3 tumor size is reached) ( Figure 1E) . Specifically, anti-PD1 enhanced the survival of mice 131 treated with cGAMP-VLP, while it had no impact when combined with ADU-S100. Furthermore, 132

we observed that the anti-tumor effect of ADU-S100 was characterized by necrosis of all the 133 injected tumors, while necrosis was rarely observed with cGAMP-VLP ( Figure S2A) . 134

These results suggested potential differences in T cell responses induced by cGAMP-VLP 135 or ADU-S100. We measured the frequency of OVA-specific CD4 + and CD8 + T cell responses in 136 blood 10 days after treatment initiation. cGAMP-VLP induced significant responses and the 137 majority of mice showed detectable responses ( Figure 1F ). In contrast, ADU-S100 did not 138 induce detectable T cell responses in most mice. In few mice, a T cell response was detected, but 139 its magnitude did not reach the average response observed with cGAMP-VLP. Overall, the 140 induction of OVA-specific T cell responses by ADU-S100 was not significant. It has been 141 proposed that ADU-S100 ablates the T cell responses, and that at lower doses it may induce 142 with cell lines. We treated the tumor cell line B16-OVA, the endothelial cell line MS1, the 242 dendritic cell line MutuDC and the macrophage cell line RAW. cGAMP-VLP induced the 243 highest levels of IFN-β in RAW cells, followed by MutuDC and MS1, in a dose-dependent 244 manner (Figure S4A, S4B) . The IFN-β induction in B16-OVA cells was the lowest. ADU-S100 245 also induced dose-dependent IFN-β, but this was less cell-type selective than cGAMP-VLP. 246

Soluble cGAMP induced detectable IFN-β only at the highest tested dose. To gain further 247 insights in the induction of interferons by antigen-presenting cells, we treated bone marrow 248 derived macrophage (BMDM) and dendritic cells (BMDC), the latter obtained either with GM-249 CSF (which generates mainly inflammatory dendritic cells) or with FLT3L (which generates a 250 mixed population of cDC1, cDC2 and pDCs). cGAMP-VLP and ADU-S100 induced similar 251 levels of IFN-α and IFN-β in BMDM and BMDC (with GM-CSF) ( Figure S4C ). In contrast, 252 cGAMP-VLP induced significantly higher levels of both cytokines in BMDC (with FLT3L) 253 ( Figure S4D ). Synthetic cGAMP induced detectable cytokines only at the highest tested dose, 254 despite 1000-fold higher amounts than in cGAMP-VLP. These results suggested a preferential 255 activation of STING in antigen-presenting cells by cGAMP-VLP, in particular in FLT3L-derived 256 cells. To determine if this was associated with preferential uptake of the particles, we attempted 257 to detect cGAMP-VLP in vivo in samples stained for p24, but the antibody-based detection was 258 not sensitive enough. As a surrogate, we treated splenocytes with cGAMP-VLP and stained for 259 p24 ( Figure S5A) . The highest levels of uptake were detected in macrophages, cDC1 and cDC2 260 ( Figure S5B, S5C, S5D) . The particles were also detected in some lymphocytes, but only in a 261 fraction of cells within each population. Altogether these results indicate that cGAMP-VLP 262 targets preferentially antigen-presenting cells. 263

To decipher the contribution of STING within antigen-presenting cells, we generated 266 STING-OST fl mice in which the first coding exon of Sting1 was flanked by LoxP sites. We also 267 introduced a Twin-Strep-tag (OST) at the N-terminus of STING protein. We crossed the mice to 268

LysM-cre or Itgax-cre and confirmed preferential deletion of STING in macrophages or dendritic 269 cells, respectively, using Strep-Tactin staining, and thus referred to these mice as STING-OST macrophages, the induction of IFN-α and IL-6 in serum by cGAMP-VLP and ADU-S100 was 272 reduced (Figure 5A, 5B) . However, the induction of OVA-specific T cells by cGAMP-VLP 273 ( Figure 5C ) and the anti-tumoral effect ( Figure 5D ) were maintained. In comparison, the anti-274 tumor effect of ADU-S100 was partially reduced. Following STING deletion in dendritic cells, 275 the induction of IFN-α and IL-6 by cGAMP-VLP was reduced, but not for ADU-S100 ( Figure  276 5E). The induction of OVA-specific T cells by cGAMP-VLP was reduced, but not completely 277 lost ( Figure 5F ) and the anti-tumor effect of cGAMP-VLP was essentially abrogated in these 278 mice ( Figure 5G ). In contrast, the anti-tumor effect of ADU-S100 was reduced but maintained. 279

These results indicate that STING is specifically required in dendritic cells for the anti-tumor 280 effect of cGAMP-VLP, while the anti-tumor effect of ADU-S100 depends partially on STING in 281 macrophages and dendritic cells. 282 283

The activation of STING in dendritic cells by cGAMP-VLP raised the possibility that it 285 could induce anti-tumor T cell responses even after injection outside of the tumor mass. We first 286 tested the B16-OVA model combined with anti-PD1 ( Figure 6A ). Sub-cutaneous (s.c.) injection 287 of cGAMP-VLP induced detectable levels of IFN-α, IFN-β, IL-6 and TNF-α, albeit to lower 288 levels than following intra-tumoral (i.t.) injection ( Figure 6B ). Tumor growth was delayed after 289 14 s.c. injection of cGAMP-VLP (Figure 6C ), leading to significantly smaller tumors ( Figure 6D) . 290 cGAMP-VLP s.c. also induced anti-OVA T cell responses ( Figure 6E ) and increased the survival 291 of tumor-bearing mice ( Figure 6F) . 292

In these experiments, the i.t. route remained more effective than the s.c. route at inducing 293 T cell responses and anti-tumor effects. This suggested that a negative regulator of the immune 294 response might be eliminated locally by i.t. activation of STING. We previously noted that 295 cGAMP-VLP induced a reduction of Tregs in the injected tumor, but not in the distal tumors 296 ( Figure 4B ). This raised the possibility that intra-tumor Tregs might limit the anti-tumor effect of 297 systemic STING activation by cGAMP. In order to test this hypothesis, we used an IgG2a isotype 298 antibody against CTLA4 (anti-CTLA4-m2a), which has been shown to selectively deplete Tregs 299 in tumors (Arce Vargas et al., 2017; Selby et al., 2013) , and we confirmed this effect in the 300 MCA-OVA tumor model (Figure 6G, 6H) . Treatment with anti-CTLA4-m2a had no effect on 301 the induction IFN-α, IL-6 and TNF-α by cGAMP-VLP ( Figure S7A ). In monotherapy, cGAMP-302 VLP s.c. or anti-CTLA4-m2a increased the frequency of OVA-specific CD8 + and CD4 + T cells, 303 but no significant response to the endogenous tumor antigen p15 (Figure 6I, S7B) . In contrast, 304 combining cGAMP-VLP s.c. with anti-CTLA4-m2a synergized to significantly increase the 305 levels of T cells against p15, and further increased the levels of T cells against OVA. 306 Accordingly, combination therapy induced a near-complete reduction in tumor size ( Figure 6J , 307 S7C). Similarly, monotherapies induced an increase in survival, but only the combination therapy 308 induced long-term survival of treated mice ( Figure 6K ). Completely responding mice were also 309 protected from a secondary tumor challenge ( Figure S7D) These results highlight the crucial importance of targeting STING activation in particular 316 cell types, namely dendritic cells, to optimize the antigen-specific anti-tumor responses. STING 317 was previously shown to be required in dendritic cells in vitro to induce an interferon response to 318 immunogenic tumor cells or tumor DNA (Deng et al., 2014; Woo et al., 2014) . In vivo, it was 319 previously noted that dendritic cells are a major source of IFN-β in tumors that induce STING-320 dependent immunogenic responses (Andzinski et al., 2016) . Intriguingly, STING in CD11c + cells 321 is also implicated in the negative regulation of allogeneic responses (Wu et al., 2021) . Altogether, 322 STING in dendritic cells emerges as a linchpin for the induction of antigen-specific T cell 323

In contrast to cGAMP-VLP, the anti-tumor responses induced by ADU-S100 were not 325 associated with the induction of tumor-specific T cells. It was previously proposed that the 326 induction of antigen-specific T cells by ADU-S100 was dose-dependent (Sivick et al., 2018) . We 327 did not observe such bimodal behavior in the tumor model we tested. We noted that ADU-S100 328 induced some level of tumor-specific T cells in experiments with in-house bred mice (Figures 329 5C, 5F), but not with mice obtained from an external source (Figures 1F, 2C, S2C ). This raises 330 the intriguing possibility that housing parameters such as the composition of the microbiota, or 331 genetic background, might affect the immunogenic properties of synthetic CDNs. We also noted 332 that synthetic CDNs induced necrosis at the intra-tumoral injection site which was rarely seen 333 with cGAMP-VLP. This is consistent with a role of STING activation in endothelial cells caused The cGAMP-VLPs were serially diluted in PBS at room temperature and acquired on a 469

NanoSight as previously described (Liao et al., 2019) . 

All mice from the STING agonist-treated group (cGAMP-VLP and ADU-S100) and vehicle-609 treated group were sacrificed 24 hours after the last intratumoral injection. Spleen, draining/non-610 draining lymph nodes and tumors were excised. Splenocytes were isolated by pressing the spleen 611 through a 40-µm cell strainer, axillary or inguinal LNs were dissected, pierced once with fine tip 612 forceps, and collected into RPMI on ice. For the splenocytes, RPMI was replaced with 2 mL 613 enzymatic solution of CO2-independent medium containing 1 mg/mL liberase (Sigma) and 20 614 μ g/mL DnaseI (Roche), and incubated for 30 minutes in a 37°C incubator with gentle agitation. 615

After 30 minutes, red blood cells were lysed using an ammonium chloride lysis buffer as 616 described above. Cells were pelleted (300 x g, 10 minutes, 4°C) and resuspended in ice cold 617 PBS ADU-S100 cGAMP-VLP PBS ADU-S100 cGAMP-VLP CD8 + CD4 + PBS ADU-S100 cGAMP-VLP PBS ADU-S100 cGAMP-VLP CD8 + CD4 + PBS ADU-S100 cGAMP-VLP PBS ADU-S100 cGAMP-VLP CD8 + CD4 + PBS ADU-S100 cGAMP-VLP PBS ADU-S100 cGAMP-VLP CD8 + CD4 + PBS ADU-S100 cGAMP-VLP PBS ADU-S100 cGAMP-VLP CD8 + CD4 + PBS ADU-S100 cGAMP-VLP PBS ADU-S100 cGAMP-VLP PBS ADU-S100 cGAMP-VLP PBS ADU-S100 cGAMP-VLP CD8 + CM CD4 + CM CD8 + EM CD4 + EM PBS ADU-S100 cGAMP-VLP PBS ADU-S100 cGAMP-VLP PBS ADU-S100 cGAMP-VLP PBS ADU-S100 cGAMP-VLP ADU-S100 0.05 µg ADU-S100 0.5 µg ADU-S100 5 µg ADU-S100 25 µg ADU-S100 50 µg 

Evaluation of Cross-655 presentation in Bone Marrow-derived Dendritic Cells in vitro and Splenic Dendritic Cells ex vivo 656 Using Antigen-coated Beads

Growing tumors induce a local STING dependent 659

Type I IFN response in dendritic cells: STING Dependent Type I IFN Response in Dendritic 660 Cells

Depletes Tumor-Infiltrating Regulatory T Cells and Synergizes with PD-1 Blockade to Eradicate 664 Established Tumors

Viruses transfer the antiviral second messenger 667 cGAMP between cells

Intrinsic antiproliferative activity of the 670 innate sensor STING in T lymphocytes

Inclusion of cGAMP within 673 virus like particle vaccines enhances their immunogenicity

Direct Activation of STING in the 676

Tumor Microenvironment Leads to Potent and Systemic Tumor Regression and Immunity

Overcoming resistance to checkpoint blockade 680 therapy by targeting PI3Kγ in myeloid cells

STING activation of tumor endothelial cells 683 initiates spontaneous and therapeutic antitumor immunity

STING-Dependent Cytosolic DNA Sensing Promotes Radiation-687 Induced Type I Interferon-Dependent Antitumor Immunity in Immunogenic Tumors

Radioresistant Stromal Cells Are Essential for Therapeutic Efficacy of Cyclic Dinucleotide 692 STING Agonists in Nonimmunogenic Tumors

Transmission of innate immune signaling by packaging of 695 cGAMP in viral particles

The Kinase IKKβ Regulates a STING-and NF-κB-Dependent 698

Antiviral Response Pathway in Drosophila

Signalling strength determines proapoptotic functions of STING

ExoSTING, an extracellular vesicle loaded with STING 704 agonists, promotes tumor immune surveillance

Refractoriness of STING therapy 706 is relieved by AKT inhibitor through effective vascular disruption in tumour

Single-step Strep 709 -tag® purification for the isolation and identification of protein complexes from mammalian 710 cells

Small Molecule 713 Enhancers of Endosome-to-Cytosol Import Augment Anti-tumor Immunity

Acetylcholinesterase is not a generic marker of 717 extracellular vesicles

Activated STING in a 720 vascular and pulmonary syndrome

Engineered PLGA microparticles for long-term, pulsatile release 723 of STING agonist for cancer immunotherapy

Effective delivery of 725 STING agonist using exosomes suppresses tumor growth and enhances antitumor immunity

STING cyclic dinucleotide sensing originated in bacteria

CD4 cells can be more efficient at tumor rejection than CD8 cells

Agouti C57BL/6N embryonic stem cells for mouse 735 genetic resources

High-efficiency deleter mice show that FLPe is an alternative to Cre-738 loxP

Anti-CTLA-4 Antibodies of IgG2a Isotype Enhance Antitumor Activity through 741 Reduction of Intratumoral Regulatory T Cells

Retroviruses use CD169-mediated 744 trans-infection of permissive lymphocytes to establish infection

Magnitude of Therapeutic STING 747

Activation Determines CD8(+) T Cell-Mediated Anti-tumor Immunity

The PDGF alpha receptor is required for neural crest cell development and for 750 normal patterning of the somites

PD-1 blockade induces 753 responses by inhibiting adaptive immune resistance

Biodegradable STING agonist nanoparticles for enhanced cancer immunotherapy

STING-dependent cytosolic DNA 759 sensing mediates innate immune recognition of immunogenic tumors

STING negatively regulates allogeneic T-cell responses by 762 constraining antigen-presenting cell function