Micellar formulation provides potent and tolerable TLR7 agonist treatment for anti-cancer immunotherapy

Abstract

Immunostimulants can be highly effective anti-cancer therapeutics; however, their systemic use is often limited by adverse reactions (AEs). Formulating immunostimulants into nanoparticle systems can potentially alleviate these, but nanoparticle design is key. In previous studies, we encountered anti-nanoparticle reactions with systemically administered PEGylated liposomes containing Toll-like receptor (TLR) agonists. In this work, we hypothesized that using a micellar drug delivery platform, rather than a liposomal platform, could retain the benefits of nanoparticle delivery systems while avoiding PEG recognition and generation of anti-PEG antibodies. Indeed, micellar formulation of the TLR7 agonist 1V270 induced far lower anti-PEG antibody levels and was well tolerated while retaining a similar circulation profile across multiple dosing. Furthermore, 1V270-micelles showed strong efficacy as monotherapy in murine syngeneic cancer models and showed combinatorial efficacy with anti-PD1 treatment. Following intravenous administration, tumors developed an inflammatory reaction and macroscopic hemorrhage 6 h post treatment followed by significant cell death 24 h post treatment, which was not observed in spleens and livers. Tumors displayed strong innate signaling within 24 h, which was accompanied by persistent massive infiltration of neutrophils and antigen-specific cytotoxic T cells, reduction in cancer cells and broad upregulation of immune-related genes. 1V270-micelles were well tolerated by non-human primates at doses equivalent to those displaying therapeutic activity in murine cancer models. Overall, the study provides novel insights into the mode of action of TLR7 agonists and demonstrates good and sustained tolerability of 1V270-micelles across animal models and excellent efficacy in murine cancer models by bridging innate and adaptive immunity.

Introduction

Toll-like receptor (TLR) 7 agonists can bridge the innate and adaptive immune systems by polarizing the immunosuppressive tumor microenvironment (TME) to support potent anti-cancer immune phenotypes and ultimately therapeutic activity.^1–3 TLR7 agonists are recognized by TLR7 in endosomal compartments of several immune and non-immune cell types including neutrophils, macrophages (MΦ), dendritic cells (DCs) and T cells.^4–7 Through a series of effects, including an influx of innate cells, improved antigen presentation, and the generation of antigen-specific CD8⁺ T cells, TLR7 agonists can repolarize the TME, ultimately providing an environment capable of supporting an effective anti-cancer immune response.^1–3 Despite many impressive preclinical results, the clinical use of TLR7 agonists is limited to topical application of Imiquimod.^1,8–11 However, local application is not suitable for treatment of many solid tumors and metastases, making systemic administration an attractive strategy for targeting solid tumors. Systemic administration of TLR7 agonists as a free drug has so far been hampered by dose-limiting toxicity due to systemic immune activation resulting in lack of efficacy at tolerated doses.^10,12 Therefore, reduction of toxicity of systemically administered TLR7 agonists is central for their use in cancer immunotherapy.

Lipid-based drug delivery systems provide a platform that is highly flexible in terms of drug incorporation and have the potential to improve efficacy and safety of drugs by altering the circulatory properties, cellular uptake, and biodistribution of the drug.¹³ PEGylated liposomes are widely used as drug delivery systems to reduce adverse effects, prolong circulation time, and possibly increase accumulation in cancerous tissue.^14,15 However, we and others have demonstrated that PEGylated liposomes containing TLR agonists are not suitable for repeated dosing due to induction of anti-PEG antibodies resulting in opsonized particles, accelerated blood clearance, and hypersensitivity reactions.^16–18 Recently, we have demonstrated that the recognition of PEGylated nanoparticles by anti-PEG antibodies is highly dependent on the liposomal surface.¹⁹ Based on this, we hypothesized that a micellar drug delivery platform could reduce challenges associated with humoral detection upon repeated dosing. To this end, we developed micelles containing a TLR7 agonist and showed that these were well tolerated in mice, rats and non-human primates in repeated dosing without being subjected to antibody-mediated clearance. The micelles showed efficacy in multiple murine syngeneic cancer models and demonstrated a novel and potent effect on the TME. Thus, micellar formulation of a TLR7 agonist provides a promising approach for systemic immunotherapy.

Materials and methods

Study design

This study aimed at improving the therapeutic index for TLR7 agonists by formulating the agonist in nanoparticles. Initially, we investigated the optimal nanoparticle formulation with regards to anti-PEG antibody response and accelerated blood clearance in vivo. Following this, we determined biodistribution as well as anti-tumor activity of 1V270-micelles against a commonly used TLR7 agonist using syngeneic cancer models. The immunological mechanism of action was characterized for immediate and adaptive immunological changes on a single cell level using flow cytometry and on mRNA level using Nanostring. Anti-tumor activity was evaluated in additional subcutaneous syngeneic cancer models in combination with checkpoint blockade. The immediate effects were then confirmed in cancer models having shown differential efficacy. Finally, tolerability was evaluated in cynomolgus monkeys in a dose escalating study.

All experiments were based on protocols with specific research questions, with a clear design and plan for analysis. Treatment of tumor-bearing animals was only conducted on established tumors (mean volume of ≈100 mm³), which for most models occurred between 7 and 14 d post inoculation, and after being subjected to a stratified block randomization based on tumor volume. Group sizes were determined based on prior experience. Experimental groups with mice were typically composed of 5 to 10 animals per group to ensure statistical power. Lower number of animals were used with larger animals for ethical reasons. In non-tumor bearing mouse studies, mice were subjected to a stratified block randomization based on weight before treatments. No exclusion criteria were defined. Animals were kept under controlled circumstances (light conditions and air temperature and humidity) and acclimatized for at least 5 d prior to study conduct. Investigators were blinded during some of the efficacy studies and treatment and measurements order was random in some experiments. Laboratory animals’ care was in accordance with institutional guidelines. No outliers were removed from data.

Preparation of micelles

Micelles were prepared from DSPE-mPEG2k (Lipoid) and 1V270 (Chimete or Corden Pharma). Lipids were dissolved in organic solvent and combined in a molar ratio of 90:10 (DSPE-PEG2k: 1V270), vortexed and ultrasonicated. Solvent was removed by lyophilization (Scanvac Coolsafe lyophilizer, Labogene) and micelles were re-hydrated in PBS (150 mM NaCl and 10 mM phosphate, pH 7.1 to 7.4) with gentle vortexing followed by ultrasonication. Micelle dispersions were sterile filtered through a nylon 0.22 µm syringe filters (Frisenette).

Preparation of [¹⁴C]1V270-micelles was done using [¹⁴C]1V270 (ViTrax, USA) with specific activity of 2.11 MBq/mg. The radiochemical purity of [¹⁴C]1V270 was 99.7% based on HPLC and chemical purity of 98.6% based on HPLC. Formulation was done by mixing [¹⁴C]1V270 with non-radioactive 1V270 and DSPE-mPEG2k dissolved in DMSO. PBS was injected into the highly concentrated mixture at 30°C with stirring to form the [¹⁴C]1V270-micelles followed by further dilution with PBS.

Micelles were stored at 2–8°C until use. The hydrodynamic diameter of was measured by dynamic light scattering (DLS) using a Zetasizer (Malvern Instruments), in the same buffer as prepared in, and reported as distribution according to particle number. Zeta potential was measured in 5% (w/w) glucose, 10 mM HEPES, 1 mM CaCl₂ in milliQ water, pH 7.4 using a Dip cell Kit (Malvern Instruments), and Zetasizer.

Preparation of liposomes

Anionic liposomes were prepared from (POPC, Lipoid), (POPG, Avanti Polar lipids), Cholesterol (Chol, Sigma Aldrich), (DOPE-PEG2k, Avanti Polar lipids) and 1V270 in a molar ratio of 44.25:30:20:5:0.75. Neutral liposomes were prepared in a molar ratio of 64.25:30:5:0.75 (POPC: Chol: DOPE-mPEG2k : 1V270). Liposomes were prepared by mixing the indicated components in tert-butanol: water. Lipid mixtures were lyophilized, re-hydrated in PBS and heated to 65°C for 1 h under stirring. Solutions were then extruded once through 3 stacked Whatman filters (GE Healthcare) with pore sizes of 400, 200, and 100 nm followed by 7 extrusions through 2 stacked 100 nm Whatman filters. Extrusion was performed at 10–30 bar nitrogen pressure using a 10 ml LIPEX thermobarrel pressure extruder on a heating block at 65°C. Liposomes were characterized similarly to micelles and stored at 2–8°C until use.

CryoTEM

1V270-micelle or anionic or neutral 1V270-liposome solutions were placed on freshly glow discharged 300 mesh copper grids coated with lacey carbon (Ted Pella). Solutions were allowed to sit for 10 s before excess was blotted away and plunge frozen in liquid ethane using a Thermo Fisher Vitrobot Mark IV (Thermo Fisher Scientific). Samples were imaged using a FEI Tecnai G2 20 TWIN transmission electron microscope (Thermo Fisher Scientific) operated at 200 keV in low dose mode with a FEI High-Sensitive (HS) 4k × 4k Eagle camera.

Animal models

Studies involving mice were conducted at University of Copenhagen (Denmark) or at Crown Bioscience (China) for which all procedures were approved by the local ethical committees. Studies conducted at University of Copenhagen were also approved by the Danish National Animal Experiment Inspectorate. Experimental work with mice was done on female or male mice aged 6–13 weeks BALB/cJRj and C57BL/6JRj obtained from Janvier Labs or Shanghai Lingchang Biotechnology and conducted at University of Copenhagen (Denmark) and at Crown Bioscience (China). Mice were kept in cages of up to 10 animals per cage. Male mice were only used for evaluation in the RM-1 prostate cancer model. Treatments with micelles, liposomes. or vehicle were given intravenously in a tail vein. Anti-PD1 antibody (InvivoMab clone RMP1-14, BioXcell) was administered intraperitoneally.

Studies involving rats were conducted at Charles River (United Kingdom) under license from the UK Home Office for control of procedures and at ITR Laboratories Canada Inc (Canada) where the Animal Care Committee reviewed, assessed, and approved the study plans prior to initiation of any work. Experimental work with rats was done on 8–9 weeks old male or female Sprague Dawley rats (Charles River, USA, and United Kingdom) and conducted at Charles River (United Kingdom) and ITR Laboratories Canada Inc (Canada). Treatments with 1V270-micelles were administered by intravenous infusion through an indwelling catheter in a tail vein, over a duration of 1 h.

Studies involving non-human primates were conducted at Cynbiose (France) which was approved by the local ethical committees and the Ethics Committee of VetAgro Sup. Experimental work with cynomolgus monkeys was done on ∼2 yr old males imported from Vietnam via BioPRIM to Cynbiose (France) where the study was conducted. Animals were dosed by intravenous infusion over 20 min and assessed for mortality, clinical signs, body weight, food consumption, body temperature, heart rate and arterial, systolic and diastolic blood pressure following infusions.

Mouse cancer models

The murine cancer cell lines H22, Panc02, CT26, A20, B16–BL6, and RM-1 were cultured in RPMI 1640 supplemented with 10% fetal bovine serum (FBS) and 2 mM Glutamax, 100 units/ml penicillin and 100 µg/ml streptomycin. EMT6, MC38, B16-F10, LL/2, Hepa 1–6, and Renca were cultured in DMEM containing 4.5 g/l glucose supplemented with 10% FBS and 2 mM Glutamax, 100 units/ml penicillin and 100 µg/ml streptomycin. All cells were maintained at 37°C and 5% CO₂ in a humidified atmosphere. For the studies conducted by Crown Biosciences, all cell lines were cultured without penicillin and streptomycin.

Mice were inoculated in the right flank with 0.1 ml PBS or cold RPMI 1640 (without supplements) with ≥90% viable cancer cells based on trypan blue staining. Inoculums contained 3–5 × 10⁵ CT26, 5 × 10⁵ EMT6, 1 × 10⁶ H22, 5 × 10⁶ Hepa 1–6, 1 × 10⁶ MC38, 1 × 10⁶ Renca, 3 × 10⁶ Panc02, 5 × 10⁵ A20, 2 × 10⁵ B16-F10, 2 × 10⁵ B16-BL6, 3 × 10⁵ LL/2 or 1 × 10⁶ RM-1 cancer cells. Tumor volumes were measured with calibers as length×width²/2, and block randomized based on tumor volume. For the efficacy study in MuScreen models, mice were euthanized when tumor volumes reached 3,000 mm³, group tumor volume of at least 2,000 mm³, weight loss of >20% from baseline, >15% for consecutive 72 h or failure to thrive. For remaining studies, mice were euthanized when tumors reached 1,200 mm³, weight loss >20% or failure to thrive. Tumor volume curves are reported with last observation carried forward for euthanized mice. Tumor growth inhibition (TGI) was calculated as %TGI=(1-Treatment_{tumor_volume}/Control_{tumor_volume})×100 on the day when the first mouse in the control group was euthanized based on tumor volume. TGI was calculated (the last day where all animals were alive) which was day 15 for CT26, 16 for EMT6, 12 for H22, 21 for Hepa 1–6, 14 for MC38, 16 for Renca, 28 for Pan02, 14 for A20, 9 for B16-F10, 9 for B16-BL6, 16 for LL/2 and 11 for RM-1. Complete responders were defined as tumor free mice on the last day of measurement in each experiment. For rechallenge, mice were rechallenged with 3 × 10⁵ CT26 cells on the opposite flank and considered to having rejected the rechallenge when no tumor had grown 60 d after the rechallenge.

Biodistribution and excretion

Following administration of [¹⁴C]1V270-micelles, BALB/cJRj mice were anesthetized with Hypnorm and Midazolam and perfused with 10 ml saline through the left heart ventricle. Organs were dissolved in Solene-350 overnight at 60°C. Hydrogen peroxide was added to reduce color from erythrocytes. Dissolved tissues were counted by ¹⁴C liquid scintillation (Hidex Liquid Scintillation Counter 300SL) in at least 9-fold excess Ultima Gold MV (Perkin Elmer).

Urine and faeces were collected into containers from male Sprague Dawley rats following administration of [¹⁴C]1V270-micelles. Urine samples were collected and measured in duplicates after mixing with AquaSafe 500 Plus scintillation fluid. Feces samples were collected, homogenized in water in duplicate portions and combusted using a PerkinElmer Model 307 Automatic Sample Oxidiser. ¹⁴CO₂ was collected by absorption in Carbosorb^® to which Permafluor^® E⁺ scintillation fluid was added. Samples were analyzed using a Liquid Scintillation Analyzer with automatic quench correction by an external method.

Tissue distribution of radioactivity in rats was analyzed in whole-body 30 µm-thick sagittal sections using quantitative whole-body autoradiography.

ELISA and protein multiplex

Mouse plasma was prepared from EDTA-treated blood by centrifugation at 2,000×g, 4°C for 15 min. Quantification of anti-PEG IgM and IgG by ELISA was done using 400-fold diluted plasma by ELISA as described previously.¹⁸

Tumors and spleens from mice were processed as described previously.²⁰ Tumor lysates were analyzed using DuoSet ELISA (R&D systems). Plasma was analyzed using Mouse IFN Alpha All ELISA kit (PBL Assay Science) and the V-PLEX Plus Proinflammatory Panel 1 Mouse kit (MSD Mesoscale) which was read on a QuickPlex SQ 129 reader.

Plasma from monkeys was analyzed using a U-plex panel (MSD Mesoscale) and read on a QuickPlex SQ 129 scanner. C-reactive protein (CRP) was measured by enzyme-linked immunosorbent assay (ELISA). All measurements were done in technical duplicates.

Single liposome opsonization measurement

Mouse plasma were used in a Single Liposome Opsonization Measurement (SLOM) competition assay as previously described.²¹ Briefly, PEGylated liposomes functionalized with Biotin and fluorescently labeled with DiD (DSPC: Cholesterol: DSPE-PEG: DSPE-PEG-Biotin: DiD molar ratio 56.35:38.2:5.2:0.05:0.2) were immobilized on glass coverslips for microscopy (Ibidi) previously passivated with a 1:10 mixture of BSA-Biotin: BSA. Plasma was diluted 1:40 in HEPES buffer and incubated for 10 min at room temperature with either DSPE-PEG micelles (1 µM or 50 µM DSPE-PEG) or PEGylated liposomes (DSPC: Cholesterol: DSPE-PEG 56.6:38:2:5.2 molar ratio) (40 µM or 2 mM total lipid, corresponding to 1 µM and 50 µM surface-displayed DSPE-PEG). As a control, no nanoparticle was added to the plasma. Pre-incubated plasma was added to the wells with immobilized liposomes and incubated for 10 min at RT. After HEPES buffer washes, immobilized liposomes were incubated for 10 min at RT with 10 µg/ml Alexa Fluor 488 labeled anti-mouse IgG or IgM (Thermo Fisher Scientific). Following incubation, microscopy slides were washed in HEPES buffer before imaged using a Leica TCS SP5 inverted confocal microscope (Leica Microsystems). Images analysis and data treatment for quantitative determination of co-localization between liposomes and IgG/IgM was performed using the ComDet plugin in FiJi (ImageJ).

Gene expression

CT26 tumors were snap frozen and RNA extracted. Tumors were pulverized (CP02 CryoPrep automated dry pulverizer, Covaris), 10–30 mg tissue were bead beated for 20 s (CKmix bead beating tubes, Precellys) in RLT buffer (Qiagen) supplemented with 1% β-mercaptoethanol, centrifuged at 12,000 × g for 3 min, and RNA was extracted from supernatant using an RNeasy kit (Qiagen) in a QIAcube instrument (Qiagen). Gene expression analysis was performed using the Mouse Pan Cancer Immune Panel (Nanostring) at the SCIBLU genomic core facility at Lund University (Sweden). Gene expression data were analyzed using the advanced analysis module in Nanostring software.

Flow cytometry

Tumors were excised, weighed, cut into pieces and digested in tumor dissociation enzyme mix for murine tumors (Miltenyi Biotec) for 40 min at 37°C in a heated shaking water bath. Resulting suspensions were passed through 70 μm cell strainers and erythrolysed in VersaLyse (Beckman Coulter) for 10 min and passed through another 70 μm cell strainer. Samples used for erythrocyte determination were not subjected to VersaLyse treatment. Cell yields were determined using a Muse^® Cell Analyzer (Merck Millipore). Processed cells (0.5–10 × 10⁶) were treated with an Fc-blocking antibody (clone 2.4G2, BD Biosciences) in FACS buffer (PBS + 0.5% BSA + 0.1% NaN₃ + 2 mM EDTA). Fc-blocked samples were stained with an amine-reactive dye and antibodies (listed in Table S1) in FACS buffer and brilliant stain buffer (BD Biosciences) for 30 min on ice. Samples stained intracellularly were further processed using eBioscience™ FoxP3/Transcription Factor Staining Buffer Set (Thermo Fisher Scientific). Samples were acquired on a BD LSRFortessa X-20 (BD Biosciences) and analyzed with FlowJo software. For t-distributed stochastic neighbor embedding (t-SNE) visualization, four samples from each group were downsampled to 25,000 viable cells per sample using FlowJos Downsample plugin and concatenated. t-SNE was run on the concatenated sample using the t-SNE plugin built into FlowJo.

Pharmacokinetics

Pharmacokinetics of 1V270-micelles in mice, rats and cynomolgus monkeys were evaluated on plasma samples. Quantification of 1V270 in plasma samples was done using an LC-MS/MS procedure employing 1V270-D5 as an internal reference standard. Pharmacokinetic parameters were calculated from the combined bioanalytical data by DGr Pharma (The Netherlands) using a non-compartmental analysis with an intravenous model in Phoenix^® WinNonlin^® version 8.1 (Certara). The analysis for mice was performed by Ardena, for rats by ITR Laboratories Canada Inc. and for cynomolgus monkeys by Ardena and DGr Pharma.

Statistical analyses

Data are presented as means ± standard error of the mean (SEM) unless otherwise noted. Statistical analyses were performed using GraphPad Prism version 9 and 10 and are written in figure legends.

The number of replicates within each group (n) and specific statistical methods are indicated in figure legends. Appropriate post hoc tests were applied.

Data from tumor-naïve mice were confirmed to be parametric and comparisons performed using t tests or 1-way ANOVAs. Data from tumor-bearing animals were evaluated for being parametric. Non-parametric tests were used when appropriate. Statistical analyses of data from tumor-bearing animals included Kruskal-Wallis, 2-way ANOVAs and 3-way ANOVAs. Survival analyses were performed using log-rank (Mantel-Cox) tests. Correlation analyses were done using Pearson correlation.

Significance is presented as: not significant (ns), P ≥ 0.05; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

Results

Formulating 1V270 into micelles evades antibody-mediated accelerated drug clearance

To evaluate whether a micellar drug delivery platform could alleviate the anti-PEG antibody challenges observed for TLR-containing liposomes, we formulated the lipid anchored TLR7 agonist 1V270^22,23 (Fig. 1A) into ∼15 nm DSPE-PEG2k micelles (1V270-micelles; Fig. 1B). Formulated 1V270-micelles had a slightly negative zeta potential and a critical micelle concentration of approximately 10 µM lipid (Fig. S1). For comparison to a more traditional nanoparticle system, 1V270-micelles were compared to ∼100 nm anionic and neutral 1V270-liposome formulations containing DSPE-PEG2k (Fig. 1A and Table S2).

Considering the detrimental effect of anti-PEG antibodies for accelerated blood clearance,^16,18,24 we initially investigated the induction of anti-PEG antibodies after a single administration of 1V270-liposomes and 1V270-micelles in mice at a TLR7 agonist dose level of 11 mg/kg. The micellar formulation induced 57% lower anti-PEG IgM levels and 93% lower anti-PEG IgG levels compared to the liposomal formulation 5 d after injection (P = 0.003 and 0.03, respectively; Fig. 1C). Induction of accelerated blood clearance was investigated after repeated dosing of 1V270-micelles in rats, since accelerated blood clearance has been reported to be stronger in rats compared to mice.²⁵ Rats received weekly infusions of 3 or 10 mg/kg 1V270-micelles. Plasma was evaluated for 1V270 over time and plasma half-life of elimination (t_½) was found comparable between the first and fourth infusion (P = 0.56, Fig. 1D). The plasma half-life was found to be similar in mice (Fig. S2). Together, these data indicate that 1V270-micelles are not subject to accelerated blood clearance.

Since the lower levels of antibodies in 1V270-micelle treated mice did not induce accelerated blood clearance, we sought to investigate whether the induced antibodies were able to bind micelles using a SLOM assay.²¹ Here we investigated antibody binding from plasma to fixated PEGylated liposomes in an ex vivo setting. Free PEGylated liposomes or micelles were mixed with plasma samples from 1V270-micelle treated mice as competitive antibody binders and thereby indirectly showing ex vivo binding of antibodies to the competing nanoparticles when binding to the fixated PEGylated liposome is reduced. Competition with micelles did not interfere with IgM binding to liposomes while competition with liposomes (control with equimolar PEG concentration) did. Binding to liposomes by IgG was not inhibited with low concentration of micelles and only partially by addition of high micelle concentrations (Fig. 1E). This indicates that the anti-PEG antibodies induced by 1V270-micelles could bind liposomes but not micelles ex vivo.

As tolerability and therapeutic activity are greatly influenced by biodistribution properties of a drug, tissue distribution of 1V270-micelles was investigated in rats and tumor bearing mice using ¹⁴C-labeled 1V270 for formulation ([¹⁴C]1V270-micelles). In both mice and rats, highly perfused organs (kidney, liver, lungs and spleen) contained only low activity (1%–4% ID/g at 6 h for mice) and activity followed the circulatory half-life with no signs of tissue accumulation (Fig. 1F and Figs S3 and S4). Tumor and tumor-draining lymph node displayed similar activity as highly perfused organs in mice. As evident by activity only increasing from the 2 h to the 6 h timepoint in tumors, 1V270-micelles are subject to the enhanced permeability and retention effect (Fig. 1E and S4). Negligible activity was observed in skin and heart, while signal in brain, muscle and eyes was at the background level (Fig. S3 and S4). To further investigate clearance of 1V270-micelles, urine and feces excretion was investigated in rats infused with 5.6 mg/kg [¹⁴C]1V270-micelles (the interspecies equivalent rat dose to 11 mg/kg in mice, Table S3). After 24 h, 30%–82% had been excreted and 95%–99.9% after 48 h, with feces being the primary excretion route (Fig. S5).

1V270-micelles are efficacious in the murine CT26-cancer model

To investigate whether the circulatory properties and biodistribution of 1V270-micelles translated to a therapeutic efficacy, we evaluated the anti-tumor efficacy and compared it to the benchmark TLR7 agonist Resiquimod (R848). R848, which does not contain a lipid anchor and therefore cannot be formulated into lipid nanoparticles, has demonstrated therapeutic efficacy in the CT26 cancer model.^1,26 1V270-micelles were evaluated at doses of 5.5, 11, and 16.5 mg/kg (100, 200 and 300 nmol, respectively) and R848 was evaluated at an equimolar dose to 11 mg/kg 1V270-micelles (200 nmol) in mice bearing established CT26 tumors. Treatments were administered intravenously every fourth day for a total of 5 injections (Fig. 2A). All treatments were well tolerated across the investigated dose levels, and only transient weight loss after administration was observed (Fig. S6A). Vehicle-micelles caused no TGI while R848 provided 25% TGI and 1V270-micelles provided 66% to 86% TGI in a dose-dependent manner (Fig. 2B and S6B). Treatment with 200 nmol 1V270-micelles provided 8/9 complete responders (Fig. 2C), while R848 provided 3/9 complete responders at an equimolar dose (Fig. 2C). Complete responders were rechallenged with CT26 cancer cells on the contralateral flank on day 87 after the first treatment. All mice having formerly received immunotherapy rejected this rechallenge while naïve animals did not. This indicates that a potent anti-cancer immune memory had been established (Fig. 2D).

To elucidate the effect of a single dose and repeated drug administrations, CT26-bearing mice were treated with 1V270-micelles using the previous dosing schedule (5 doses given every fourth day) and compared to a single dose as well as 2 or 5 doses given every seventh day. Multiple treatments showed an improved, but statistically insignificant, effect (6/10 complete responders for a single administration compared to 10/10 complete responders for 5 administrations given every fourth day, P = 0.12). This supports the importance of repeated dosing and thus restimulation of the TME for a durable anti-cancer response. Extended intervals between injections showed a tendency towards slightly lower efficacy (8/10 complete responders for 2 and 5 treatments given every seventh day compared to 10/10 complete responders for five treatments given every fourth day, P = 0.60; Fig. S7A, B).

1V270-micelles induce an immediate inflammatory state in tumors

To examine the mechanism behind the strong anti-tumor activity of 1V270-micelles, we compared gene expression profiles in CT26 tumors from treated vs untreated mice. Several pro-inflammatory genes were induced one day after a single injection of 1V270-micelles (Fig. 3A). The rapid accumulation of 1V270 in tumors translated to an intratumoral immune activation which was evidenced by increased levels of the potent pro-inflammatory cytokines TNF-α and IL-1α in tumor 6 h after administration. Spleens also showed an increased level of TNF-α, although lower than tumors, while level of IL-1α in spleens was slightly higher than in tumors. No significant increase in IFN-α was observed in tumors despite type I interferons commonly being associated with TLR7 responses (Fig. 3B, P > 0.05). The potent pro-inflammatory stimulation by the 1V270-micelles was accompanied by tumor hemorrhage evident by accumulation of erythrocytes in the tumor (Fig. 3C) and visual reddening of tumors (Fig. 3D). Overall cellular viability in tumors decreased by 76% 24 h after treatment (P = 0.006) while viability of spleens and livers were only slightly affected (Fig. 3E). Further analysis of the TME by flow cytometry revealed that most cell populations were depleted from the TME (Fig. 3F). While cancer cells in untreated tumors represented 57% of all viable cells this was reduced to only 11% of all viable cells 24 h after 1V270-micelle treatment (P = 0.002; Fig. 3F, G). In contrast, immune cells went from constituting 33% of the TME in untreated tumors to 71% 24 h after treatment (P = 0.013; Fig. 3F–H). In line with the upregulation of neutrophil chemoattractants (Fig. 3A), the early immune infiltrate was highly dominated by neutrophils (P = 0.0002; Fig. 3F–I). Additionally, a broad activation (based on CD69) of the immune infiltrate, occurred within 24 h of treatment (Fig. 3J).

Strong enrichment of antigen specific response

To further elucidate the immune response to 1V270-micelle treatment we measured the expression of 622 immune-oncology related genes in tumors at selected time points in the treatment schedule used for efficacy studies. Treatment caused an increase in the expression level of a variety of immune-related genes including genes associated with cytokines, chemokines, innate and adaptive response. For the later time points, this also included upregulation of genes associated with antigen processing and T cell function in responding tumors but not in poorly responding tumors which were evaluated four days after the third treatment (Fig. 4A). To further examine the treatment effect at a single cell level we investigated the TME and tdLNs by flow cytometry after the first and third treatments, where we quantified immune cell types and cancer cells. The heatmap in Fig. 4B, showing tumor-infiltrating cells, demonstrates induction of a response with activation of both the innate and adaptive immune system. Here, all cell types, except neutrophils, were depleted from the TME one day after the first treatment. Following this, re-infiltration of the TME after one and three treatments included a large increase in CD8⁺ T cells and DC subsets amongst other cell types. Intriguingly, we found cancer cells and tumor associated MΦs (TAMs) to remain depleted while infiltration of CD8⁺ T cells remained high in all responding tumors (Fig. 4B). Overall cellular viability in tumors was significantly reduced one day after treatment and remained low only in responding tumors (Fig. 4C). The increased immune infiltration was sustained for at least nine days after a single treatment (P < 0.001) and remained increased only in responding tumors (Fig. 4D). Cancer-antigen specific CD8⁺ T cells were evaluated by determining reactivity towards the immunodominant cancer-antigen on CT26 cells, gp70_423–431 (AH-1).²⁷ Notably, the proportion of intratumoral CD8⁺ T cells reactive towards AH-1 showed a non-significant enrichment already one day after treatment and further increased on later timepoints in responding tumors (Fig. 4E). This was accompanied by a more favorable ratio of CD8⁺ T cells to regulatory T cells (T_reg) (Fig. 4F). Neutrophils in tumors showed elevated expression of migration-associated CX3CR1 for at least nine days after treatment in responding tumors (Fig. S11A). In further support of adaptive immunity induction, we observed a large increase in the expression of the co-stimulatory molecule CD86 in tdLNs one day after treatment across immune cell subsets capable of cross-presentation (Fig. 4G). The number of immune cells in tdLNs showed a trend towards an increase following treatment (Fig. 4H), but no clear tendencies in cellular composition of tdLNs were identified (Fig. S11B). Similar to observations in tumors, AH-1 reactivity amongst CD8⁺ T cells was enriched in tdLNs after treatment and remained elevated in mice with responding tumors (Fig. 4I).

Therapeutic efficacy is observed across cancer models

Having shown efficacy of 1V270-micelles in the CT26 cancer model and demonstrated favorable modulation of the TME, we further investigated the drug efficacy across 12 murine syngeneic cancer models. Combination with anti-PD-1 therapy was included to examine potential combinatorial effects of stimulating multiple parts of the cancer immunity cycle. Treatment started in all models at a similar mean tumor volume (≈100 mm³). Mice were given five treatments of 1V270-micelles every fourth day and anti-PD1 every fourth day for a total of 6 treatments. In all tested models, 1V270-micelles induced significant TGI and were superior to anti-PD-1 monotherapy (Fig. 5A and Figs S16, S17). 1V270-micelles were highly potent in CT26, EMT-6, A20, MC38 and H22 with TGI ranging 85% to 97%, while monotherapy with anti-PD-1 provided only 11% to 70% TGI. In these models, the benefit of combination treatment was not possible to evaluate due to a strong anti-tumor effect of 1V270-micelles.

Several models were resistant to anti-PD1 monotherapy whereas 1V270-micelles monotherapy showed efficacy. Combination of the 2 drugs resulted in combinatorial efficacy by converting anti-PD-1 non-responsive tumor models into PD-1-responsive. This included Pan02, Renca, RM-1 and LL/2. In these models only, combining anti-PD1 and 1V270-micelles showed an increase of the anti-tumor response compared to 1V270-micelles monotherapy (Fig. 5A and S16). Meanwhile, improvement of efficacy was seen in all investigated models receiving combination treatment when compared to anti-PD1 monotherapy.

The panel of tumor models included 3 models (LL/2, B16-F10, and B16-BL6) that are commonly regarded as insensitive to chemo-, immuno-, and radiotherapy where 1V270-micelles provided limited, but notable, TGI (24% to 41%).

To identify the mechanism underlying this differential responsiveness, we evaluated changes in the TME 6 and 24 h after 1V270-micelle injection in CT26, EMT6, MC38, Renca, and B16-F10 and correlated these results to the TGI. The immune composition in untreated tumors showed no correlation with TGI (P ≥ 0.39, Fig. S17A). Despite the differential responsiveness, the overall cellular viability was also similarly low across models after 24 h (4% to 12%, Fig. 5B and S17A–B). Erythrocyte accumulation was also observed across models and to the highest degree in CT26 6 h after treatment (4.4-fold increase; Fig. 5C and S17A). Neutrophil infiltration was high at 24 h after treatment across models. A tendency for a lower neutrophil accumulation was observed in the MC38, Renca and B16-F10 models where therapeutic activity of the 1V270-micelles was also lower compared to CT26 (Fig. 5D and S17A–B). The better responding models showed tendencies of having TMEs 24 h post treatment with higher proportions of overall immune cells and neutrophils while having lower proportions of eosinophils, CD4⁺ T cells, CD8⁺ T cells, and NKT cells (Fig. S17A). Activation of immune cells occurred across all analyzed immune cell types and models with no correlations to efficacy (Fig. S17C, D). This indicates that neutrophils together with depletion of other cell types is an important mechanism of 1V270-micelle efficacy.

1V270-micelles can be administered safely to non-human primates

Clinical advancement of 1V270-micelles is dependent on the tolerability at systemic dosing. Systemic tolerability was therefore explored in cynomolgus monkeys as these are considered an optimal translational model based on the TLR7 expression profile and function being similar to those in human.^28,29 Three cynomolgus monkeys were intra-individually dosed with escalating dose-levels of 0.01, 0.03, 0.1, 0.3, 0.9, and 2.7 mg/kg 1V270-micelles by intravenous infusions over 20 min with 14 d intervals. Mortality, clinical signs, body weight, food consumption, body temperature, heart rate, and blood pressure were assessed. Blood samples were taken prior and after each infusion for blood chemistry and biomarker analysis (Fig. 6A).

No clinical signs were observed at dose levels ≤0.9 mg/kg. At 2.7 mg/kg (interspecies dose level corresponding to the efficacious dose of 11 mg/kg in mice, Table S3), 2 monkeys displayed clinical signs including sneezing, runny nose, and slightly elevated rectal temperature. Symptoms were transient and lasted for 1 d after infusion. There was no effect on body weight, heart rate and blood pressure. Additionally, no effect was observed for investigated blood chemistry (Table S4).

Noteworthy changes in biomarkers included a transient increase in the CRP level with a peak response 24 h after each administration with a non-linear dose relationship and return to baseline ∼48 h post end of infusion (Fig. 6B). A panel of 10 cytokines was measured pre-dose, 2, 4, 8, 24, and 72 h post end of infusion. The potentially clinically relevant pharmacodynamic biomarkers IP-10, MCP-1 and IL-1RA, showed a dose-dependent induction while IL-6 showed a non-linear dose relationship and remained low at the dose levels ≤2.7 mg/kg (Fig. 6C). IFN-α2a, IFN-γ, IL-12p70, IL-15, MCP-2 and TNF-α were only weakly affected (Fig. 6C and Table S5).

The plasma concentration of 1V270 was determined to evaluate dose effects. The shape of the mean plasma concentration-time profiles was comparable over the whole investigated dose range (Fig. 6D). The plasma t_½ ranged between 1.71 h at 0.01 mg/kg to 2.23 h at 2.7 mg/kg. A dose-proportional increase in exposure was seen ≤0.9 mg/kg. At 2.7 mg/kg, a more than dose proportional increase in exposure was observed, where a 3-fold increase in dose from 0.9 to 2.7 mg/kg led to a 7.8-fold increase in mean area under the curve (AUC)_2-72h values. Volume of distribution (V_z) was comparable at all dose levels constituting 49.5 ml/kg and indicated little tissue distribution at the investigated dose levels. All measured pharmacokinetic parameters are summarized in Table S6.

In summary, escalating dose-levels of 1V270-micelles up to 2.7 mg/kg administered intravenously to cynomolgus monkeys were well tolerated at interspecies equivalent dose comparisons that provided excellent therapeutic effects in murine cancer models.

Discussion

Despite positive preclinical results for TLR7 agonists, the clinical use of TLR7 agonists is still limited to topical application. Several TLR7 agonists have been clinically investigated but have shown limited therapeutic index in humans due to systemic cytokine responses.^10,12,30 Here, we demonstrate that formulating 1V270 into micelles provided an efficacious anti-cancer therapy in mice while being well tolerable in monkeys at the corresponding interspecies equivalent dose level.

Managing systemic cytokine levels following immunotherapy is important for managing cytokine release syndrome in patients, which can limit the therapeutic index and thus ultimately treatment effectiveness.^31,32 Both magnitude and duration of cytokine induction appear to be highly variable between different TLR7 agonists which may be due to differences in pharmacokinetics as well as cellular targeting and uptake which underline the importance of the drug delivery strategies.²³ In this study, 1V270-micelles displayed a relatively short blood circulation time which limited systemic exposure. This may explain the low and brief induction of systemic pro-inflammatory cytokines observed in cynomolgus monkeys. Indicating a favorable safety profile of the drug once administered in patients.

The inherent challenge of using nanoparticles systemically in patients due to anti-PEG responses has limited anti-cancer nanomedicine to primarily include cytotoxic agents as these mitigate humoral detection by killing or otherwise impairing B cells.^33,34 Indeed, previous studies have shown that TLR agonists in nanoparticles cause potent anti-PEG responses leading to a rapid clearance of nanoparticles, which causes severe hypersensitivity reactions and loss of therapeutic efficacy in both animals and patients.^16,18,34 Here, we show that by formulating 1V270 into micelles, a similar circulatory profile could be observed between the first and successive administrations in rats and non-human primates. We also showed that the anti-PEG response was much lower for 1V270-micelles compared to 1V270-liposomes. Although these antibodies could bind liposomes, they could not bind micelles in an ex vivo setting. Together, this indicates that micellar formulations may overcome the challenges associated with antibody-mediated accelerated blood clearance for liposomal formulations and allow repeated stimulation of the TME. However, the exact effect of particle characteristics remains to be investigated.

Modulating the suppressive TME into an environment that supports anti-cancer immune phenotypes is essential for successful immunotherapy.³⁵ Although previous reports on modulating effects on the TME by TLR7 agonists are numerous, we consider this study to be amongst the most comprehensive investigations of the TME following TLR7 agonist administration. The TME was examined for acute effects across different cancer models with varying responsiveness and the adaptive effects in the well responding CT26 model. The early response included a high accumulation of erythrocytes accompanied by activation of a variety of immunological effector cells 6 h after 1V270-micelle treatment. After 24 h, effector cells remained activated, neutrophils were massively recruited to the TME, and many cell types, including cancer cells, were depleted from the TME. Meanwhile, multiple antigen presenting cells displayed strong upregulation of co-stimulatory molecules in tdLNs. This was followed by re-infiltration in tumor of antigen specific CD8⁺ T cells. This, together with persistent neutrophil infiltration, characterized the TME of only responding tumors for at least 9 d after treatment. In this study, multiple cancer models were used to address efficacy of the 1V270-micelles, and a wide range of responses was observed. To this end, others have underlined the importance of NK cells and CD8⁺ T cells for the efficacy of TLR7 agonists using depletion studies.^36–38 Yet, it is not clear exactly what determines responsiveness to 1V270-micelles without further experiments. Tumor mutational burden presents one significant aspect that has commonly been used to predict responsiveness for certain immunotherapies by estimating the potential for T cells to mount a response with proper (re-)activation.³⁹ Indeed, previous studies have shown that the CT26 and MC38 which responded well to 1V270-micelles treatment have a high mutational burden, while the low responding model B16-F10 has a low mutational burden. However, the Renca model, which responded well to 1V270-micelle treatment, also has a low mutational burden.⁴⁰ Therefore, it may be speculated that the potent acute effects observed in this study will occur across most tumor types and the magnitude of changes can to some extent predict responsiveness. Additional effect on efficacy most likely depends on whether an adaptive immune response can be mounted and maintained.

To our knowledge, no previous studies have reported an acute inflammatory state in the TME to the same extent together with a significant drop in cellular viability momentarily after administration of TLR7 agonists. This effect could be explained by the high levels of TNF-α and IL-1α present in tumors after treatment as these cytokines can induce tumor hemorrhagic necrosis.^41,42 This is in line with observations of single stranded RNA viruses (that can be recognized by TLR7) such as influenza, severe acute respiratory syndrome, and Ebola viruses, which can cause hemorrhagic viral infections.⁴³

Tumor-infiltrating neutrophils are commonly associated with a negative prognosis due to their pro-tumorigenic phenotype (polymorphonuclear-MDSCs) that is induced by chronic low-grade inflammation.^44,45 However, in response to potent signaling, neutrophils can also have an anti-tumorigenic phenotype that directly kills cancer cells and strongly potentiates the activity of CD8⁺ T cells.^2,45,46 Notably, we found that in addition to the high accumulation of neutrophils, tumor-infiltrating neutrophils had upregulated CX3CR1 expression in responding tumors. While the role of CX3CR1 in neutrophils is not known we speculate, based on CX3CR1 function in other immune cells, that neutrophils continuously migrate into the TME in a CX3CR1-mediated fashion.^47–49 Further experiments involving neutrophil depletion or neutrophil modification models are needed to fully understand their involvement.

In summary, this study presents a promising strategy for increasing the therapeutic index of TLR7 agonists for anti-cancer therapy by providing a manageable systemic cytokine profile, avoiding humoral detection and efficiently modulating the TME by bridging the innate and adaptive immune response. The immediate response in tumors to 1V270-micelles included hemorrhage, a large influx of neutrophils, depletion of most other cell types and a broad activation of effector cells. Simultaneously, tdLNs displayed a large increase in co-stimulatory molecules to support development of an adaptive immune response. 1V270-micelles also provided long-term anti-cancer memory, rescued anti-PD1 sensitivity in some models and worked combinatorically in others. This makes 1V270-micelles highly attractive for clinical advancement as a novel therapeutic approach to increase patient response rates.

Acknowledgments

We thank SCIBLU genomic core facility at Lund University for their help with RNA analysis. Figure 2A and 5A were created in BioRender. Kjaer, A. https://BioRender.com/8axkbqd and https://BioRender.com/ng5ljjt.

Author contributions

E.C., C.S., R.M., S.P., J.R.H., A.E.H., S.J., and T.L.A. conceived and designed the experiments. E.C., C.S., R.M., M.B., S.P., H.R.H., M.P., P.K., M.J. and T.L.A. performed the experiments. S.P., J.R.H., A.E.H., S.J. and T.L.A. administered and supervised the project. The manuscript was written by E.C., C.S., R.M., M.B., S.P., H.R.H., J.R.H., A.E.H., and T.L.A. All authors discussed the results and critically reviewed the manuscript.

Esben Christensen (Conceptualization [Equal], Data curation [Equal], Formal analysis [Lead], Investigation [Lead], Methodology [Lead], Project administration [Lead], Visualization [Equal]), Camilla Stavnsbjerg (Conceptualization [Equal], Data curation [Lead], Formal analysis [Lead], Investigation [Lead], Methodology [Lead], Project administration [Lead]), Rasmus Münter (Conceptualization [Lead], Data curation [Supporting], Formal analysis [Supporting], Investigation [Equal], Methodology [Equal], Project administration [Supporting], Visualization [Supporting]), Martin Bak (Investigation [Supporting], Methodology [Supporting]), Svetlana Panina (Conceptualization [Lead], Data curation [Equal], Formal analysis [Equal], Investigation [Lead], Methodology [Equal], Project administration [Equal], Supervision [Equal], Visualization [Supporting]), Hólmfridur R. Halldórsdóttir (Investigation [Supporting], Methodology [Supporting]), Morten Petersen (Investigation [Supporting], Methodology [Supporting], Visualization [Supporting]), Paul Kempen (Investigation [Supporting], Methodology [Supporting], Visualization [Supporting]), Mikael Jensen (Investigation [Supporting]), Andreas Kjaer (Project administration [Supporting], Resources [Equal], Supervision [Supporting]), Jonas R. Henriksen (Conceptualization [Equal], Project administration [Supporting], Resources [Supporting], Supervision [Equal]), Anders E. Hansen (Conceptualization [Lead], Investigation [Supporting], Project administration [Equal], Supervision [Lead]), Simon Jensen (Conceptualization [Lead], Funding acquisition [Lead], Project administration [Lead], Resources [Lead], Supervision [Lead]), and Thomas L. Andresen (Conceptualization [Lead], Funding acquisition [Lead], Investigation [Supporting], Project administration [Equal], Resources [Lead], Supervision [Lead])

Supplementary material

Supplementary material is available at The Journal of Immunology online.

Funding

This work was supported by the European Council grant (ERC-2012-StG_20111109), The Lundbeck Foundation Fellowship grant, The Novo Foundation Synergy Grant, The Danish Research Council for Independent Research Sapere Aude grant and Innovation Fund Denmark—Grand Solutions.

Conflicts of interest

E.C., M.B., S.P., M.P. and S.J. were employees of MonTa Biosciences at the time of the research. S.P., M.P., S.J., A.E.H., J.R.H. and T.L.A. hold stock and/or stock options in MonTa Biosciences. E.C., C.S., R.M., J.R.H., A.E.H., S.J. and T.L.A. are inventors on a provisional patent related to the 1V270-micelle technology included in this study (patent no. WO2021053163). H.R.H., P.K., M.J., and A.K. have no competing interests.

Data availability

All data are available in the main text or in supplementary materials. Raw data and statistical analysis are available at DOI 10.17632/pbx4mg9nwt.1. Additional requests will be provided upon reasonable requests.

References