A multidomain peptide hydrogel-liposome composite for controlled release of a cyclic dinucleotide in oral cancer

Abstract

While immunotherapy is a promising treatment strategy for cancer, the majority of head and neck squamous cell carcinoma (HNSCC) patients treated with single-agent immunotherapy do not respond. Therefore, researchers are investigating combination treatments with immunostimulatory molecules that can maximize anti-tumor responses. Cyclic dinucleotides (CDNs) are STING agonists that hold promise in combination approaches, but they require frequent intratumoral administration when used in both preclinical models of HNSCC and clinical trials. To reduce administration frequency, we have created a peptide hydrogel–liposome composite system, K₂-Lip(CDN), for local and prolonged availability of CDN. We investigated the loading limits of cationic liposomes in both anionic (E₂) and cationic (K₂) peptide hydrogels and found that E₂ caused aggregation to occur at the desired lipid:peptide molar ratio, which led to our use of K₂ in the composite system. At a molar ratio of 8:1 lipid: peptide, K₂-Lip(CDN) formed a self-healing composite hydrogel and resulted in prolonged release of CDN in vitro and indocyanine green (ICG) in vivo. This composite material resulted in cellular cytotoxicity in ROC1 oral cancer cells in vitro and extended survival in an aggressive ROC1 tumor-bearing murine model compared to a single CDN injection. The overall survival using only a single dose of K₂-Lip(CDN) was equivalent to that achieved with six repeated doses of CDN. The material properties and extended-release capabilities displayed by peptide hydrogel–liposome composite systems could translate to broad treatment applications that require the delayed release of localized therapeutics.

1. Introduction

Next-generation drug delivery systems are needed against one of the world’s most prevalent diseases: cancer. Identified as an unregulated growth of cells, cancers are notorious for evading standard-of-care therapies. Conventional multi-modality therapy in oral squamous cell carcinoma, the most common head and neck squamous cell carcinoma (HNSCC) [1], has significant toxic side effects that contribute to morbidity and often fails to eradicate cancer. Multi-modality therapy includes surgery and often adjuvant chemoradiotherapy. Chemoradiotherapy uses a combination of chemotherapy and radiation therapy, both of which destroy cancer cells at the cost of healthy tissues. This conventional treatment for HNSCC patients results in high locoregional [2] and distant recurrence rates [3]. Unfortunately, patients with nonhuman papillomavirus (non-HPV)-associated HNSCC have a worse prognosis relative to patients with human papillomavirus (HPV)-positive HNSCC [4]. Thus, there is a strong interest in utilizing treatment strategies that localize therapeutics to prevent systemic toxicity and induce a robust, durable anti-tumor immune response to reduce cancer recurrence in non-HPV models.

Immunotherapy is a promising treatment modality that enhances the immune system’s ability to target and kill cancer cells. Unfortunately, current FDA-approved immunotherapies require frequent administration and are delivered systemically, which can result in toxicity and severe adverse events [5,6]. The FDA-approved immunotherapy for recurrent or metastatic head and neck cancer is immune checkpoint inhibitors (ICIs), specifically, anti-programmed cell death protein-1 (anti-PD-1). Another ICI includes anti-cytotoxic T-lymphocyte associated protein-4 (anti-CTLA-4), which has changed the course of cancer treatment in advanced melanoma [7,8]. CTLA-4 inhibitors in combination treatments have been found to be tolerable [9], but are yet to be FDA approved for HNSCC. Although immune checkpoint inhibitors release the “brakes” on lymphocyte activity, most HNSCC patients treated with single-agent immunotherapy do not respond, in large part due to the immunosuppressive tumor microenvironment [10]. Therefore, combining immune checkpoint inhibitors with localized immunostimulatory modulators can maximize anti-tumor responses.

Cyclic dinucleotides (CDNs) are a class of immunostimulatory modulators that have garnered significant attention over the past decade, following the 2011 discovery [11] that Stimulator of Interferon Genes (STING) functions as an immune sensor of CDNs, and deeper understanding of the pathway activation mechanisms that emerged in 2013 [12,13]. These cyclized nucleobases activate the cyclic GMP-AMP synthase (cGAS)-STING pathway, a proinflammatory pathway that recognizes cytosolic DNA. However, as with other immunotherapeutics, frequent administration is typically required. One example is the STING agonist cyclic dinucleotide dithio-(RP,RP)-[cyclic[A(2′,5′)pA(3′,5′)p]] (also known clinically as ML RR-S2 CDA and subsequently this cyclic dinucleotide is abbreviated “CDN”). In a phase I clinical trial of CDN, solid tumor patients received weekly intratumoral (IT) injections (NCT02675439). When administered CDN, patients’ lesion size was stable or decreased in 94% of evaluable, injected lesions, suggesting that prolonged availability of CDN in the tumor microenvironment can result in tumor regression [14]. Achieving extended bioavailability of CDN while mitigating multiple injections may be possible through a localized delivery platform designed to control payload release. This approach could increase effectiveness of immunomodulators and limit systemic side effects.

Self-assembled biomaterials have shown promise for prolonged, localized drug delivery. Formed through weak intermolecular forces such as electrostatic interactions, hydrophobic packing, and hydrogen bonding, these biomaterials can assemble under physiologically relevant conditions, offering protection and delayed release of the loaded drug [15–18]. Previously, our group has utilized a cationic, self-assembling lysine-based multidomain peptide (MDP) hydrogel [19–22] (here called “K₂”) for prolonged release of anionic CDN. This CDN-loaded hydrogel, denoted STINGel, resulted in statistically significant extended survival in an in vivo oral cancer model compared to a CDN- loaded collagen hydrogel [23]. These results are an example of the increased efficacy that can be achieved through controlled release of therapeutics. Other methods for increasing the efficacy of similar CDNs include the use of transition metal-CDN complexes loaded into silica nanoparticles [24] or CDN encapsulated in cationic liposomes [25–27]. Despite these promising studies concerned with increasing the bioavailability of CDNs, further improvements are needed to enhance therapeutic outcomes.

Peptide hydrogels suffer from rapid release of small molecules due to mesh sizes larger than the size of small molecules, while liposomal drug formulations fail to result in high accumulation at the site of interest due to rapid clearance after systemic injection [28]. Previous studies by others [29] and us [30] have attempted to circumvent these shortcomings with the creation of peptide hydrogel-liposome composite systems. Toward this effort, our lab utilized MDPs in combination with liposomes. The resulting composite system displayed tunable fluorophore release kinetics through alteration of peptide and lipid compositions [30]. Inspired by those results, we were curious whether a similar approach could be used to create a CDN-loaded composite for localized, delayed release (Fig. 1). As the concentration of liposomes needed to be much higher in these composites, we investigated the effects of liposome concentration on the composite’s material properties. This investigation revealed that self-healing hydrogels can still be formed at high lipid: peptide molar ratios (up to 8:1) when the liposome and peptide charges are matched. We then showed that these CDN-loaded composites result in prolonged release of CDN in vitro and ICG in vivo, and cytotoxicity of ROC1 oral cancer cells in vitro. Finally, we showed that a single intratumoral injection of this composite results in similar survival as six injections of free CDN. This study highlights the ability of MDP hydrogels to accommodate high concentrations of liposomal drugs for delayed release of therapeutics and the potential for composite biomaterials to decrease dosing frequency in cancer treatment.

Fig. 1.

Diagram of the self-assembling systems. A) Depiction of the self-assembly of K₂ MDP: Hydrophilic (green), hydrophobic (blue), and charged (red) regions of the assembly are highlighted. Formation of β-sheet-rich bi-layered nanofibers are induced by salt-mediated charge-charge quenching. Nanofiber entanglement results in a syringe-injectable nanofibrous hydrogel. B) Cationic liposomes used to encapsulate anionic cyclic dinucleotide (CDN) for increased cellular uptake. C) Composite system formed by the combination of nanofibrous K₂ hydrogel and CDN-loaded cationic liposomes. This system is syringe-injectable and results in delayed therapeutic release. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

2. Methods

2.1. Peptide synthesis

K₂ (primary sequence: (K₂(SL)₆K₂) (Fig. S1a–c) and E₂ (primary sequence: E₂(SL)₆E₂) (Fig. S1d–f) were synthesized using manual solidphase peptide synthesis with an Fmoc-protection strategy previously reported from our lab [30]. This synthesis resulted in an N-terminally acetylated and C-terminally amidated product. Matrix-assisted laser desorption/ionization time of flight mass spectrometry (MALDI-TOF) was used to confirm peptide molecular weight (Bruker Instruments, Billerica, MA) (Fig. S1b,e). Crude peptide was purified by high performance liquid chromatography (HPLC) using a Waters 1525 Binary HPLC Pump with an XBridge Protein BEH C4 OBD Prep column (Waters Corporation, Milford, MA). Mobile phases were water and acetonitrile, both with 0.05% trifluoroacetic acid (TFA), for K₂ and water and acetonitrile, both with 5 mM ammonium hydroxide (NH₄OH) and 4 mM acetic acid (AcOH), for E₂. (Fig. S1c,f).

2.2. Liposome synthesis

Liposomes were synthesized using the thin film hydration method previously described [30]. Lipids were purchased from Avanti Polar Lipids (Alabaster, AL) and cholesterol was purchased from Fisher Scientific (Hampton, NH). Briefly, chloroform solutions of cholesterol, 1,2-dioleoyl-3-trimethylammonium-propane (DOTAP) (Fig. S1g), hydrogenated soy choline (H. soy choline) (Fig. S1h), and 1,2-dimyristoyl-rac-glycero-3-methoxypolyethylene glycol-2000 (DMG-mPEG2000) were all added to a black capped test tube in a molar ratio of 37:20:40:3. Empty liposomes, denoted Lip(Empty), consisted of cholesterol, H. soy choline, and DOTAP in a molar ratio of 50:28:23. Chloroform was removed by evaporation using a stream of nitrogen and then placed under vacuum for 1 h, resulting in a dry lipid film. The lipid film was then hydrated with either 1.36 mg/mL (1.97 mM) CDN solution in 1× Hanks’ Balanced Salt Solution (HBSS) preheated to 60 °C, or with 1× HBSS alone for Lip(Empty), resulting in 100 mM of lipids. For ICG- liposomes used in the in vitro release experiment, the lipid film was hydrated with 200 μg/mL indocyanine green (ICG) in 1× Hanks’ Balanced Salt Solution (HBSS) preheated to 60 °C. For ICG-doped CDN liposomes used in the in vivo release experiment, the lipid film was hydrated with a mixture of 1.36 mg/mL CDN and 17 μg/mL indocyanine green (ICG) in 1× Hanks’ Balanced Salt Solution (HBSS) preheated to 60 °C. Hydration was allowed to occur for 30 min at 60 °C with 30 s of vortexing every 5 min. Liposomes were then subjected to 5 freeze-thaw cycles consisting of 3 min in liquid nitrogen followed by 5 min in a 60 °C water bath. Solutions were vortexed for 30 s after each thaw cycle. Liposomes were then extruded back and forth 15 times through 2 stacked 200 nm polycarbonate membranes using a Mini-Extruder (Avanti Polar Lipids Inc., Alabaster, AL). For Lip(Empty), subsequent passage through 100 nm membranes was achieved, followed by purification using a Sephadex G-50 column. CDN encapsulation efficiency was calculated using an Amicon Ultra Centrifugal Filter (MWCO 100,000) and analyzing the filtrate, as well as by bursting the liposomes remaining in the filter with Triton X-100 and running the sample on a Waters ACQUITY UPLC system with an ACQUITY UPLC Protein BEH C18 column. Water and acetonitrile with 5 mM NH₄OH and 4 mM AcOH were used as the mobile phases. A CDN standard curve was created using UPLC to ensure values were within linear range (Fig. S2a). Liposomes were then characterized for z-average, polydispersity index (PDI), and zeta potential by dynamic light scattering (DLS) using a Malvern Zen 3600 Zetasizer (Malvern Instruments Ltd., Malvern, U.K.) (Table S1).

2.3. K₂-Lip(CDN) preparation

A 20 mg/mL (11.2 mM) solution of MDP was made in Milli-Q water. Then, this peptide solution, 10× HBSS, and 100 mM CDN-loaded liposomes were mixed to give final concentrations of 10 mg/mL (5.6 mM) peptide, 1× HBSS, and 29 mg/mL (44 mM) liposomes (resulting in 0.67 mg/mL CDN). This composite system is denoted K₂-Lip(CDN) for in vitro and in vivo experiments. Composite systems were stored at 4 °C until use.

2.4. MDP-liposome loading limitations

A suspension of liposomes containing 100 mM of lipids (i.e. Lip- (High)) was diluted with 1× HBSS to create 50 mM liposomes (i.e. Lip- (Med)), 25 mM liposomes (i.e. Lip-(Low)), and 10 mM liposomes (i.e. Lip-(Very Low)). These four liposome samples were mixed with either K₂ or E₂ using the procedure described above (Table S2, S3). 3 μL of each composite material was analyzed using a Thermo Scientific Nanodrop 2000C Spectrophotometer (Waltham, MA) by measuring the absorbance at 800 nm. All subsequent experiments utilized CDN-loaded liposomes.

2.5. Oscillatory rheology

Viscoelastic properties of composite materials were analyzed using an AR-G2 rheometer (TA Instruments, New Castle, DE). 100 μL of gel was pipetted onto the stage. A 12 mm stainless-steel parallel plate and gap height of 500 μm were used for all experiments. Samples were allowed to equilibrate for 30 min at 1% strain and 1 rad/s. Frequency sweeps were run at 1% strain over 0.1–100 rad/s. Amplitude sweeps were run at 1 rad/s frequency over 0.01%–200% strain. The step-strain recovery test allowed gels to equilibrate for 30 min at 1% strain and 1 rad/s, followed by 1 min of 200% strain and 1 rad/s, and finally 30 min of recovery at 1% strain and 1 rad/s. For repeated step-strain recovery, gel was allowed to recover for 10 min at 1% strain and 1 rad/s, followed by 1 min of 200% strain and 1 rad/s. This process was repeated two more times for a total of three step-strain recoveries.

2.6. Cryo-EM

Lacey carbon grids were glow discharged twice for 1 min with a 5 mA discharge. 3 μL of sample was pipetted onto the grid, blotted, and plunged into liquid ethane using a Vitrobot. The grid was kept under liquid nitrogen for storage until imaging. An FEI Tecnai F20 transmission electron microscope (FEI Company, Hillsboro, OR) was used for imaging.

2.7. SEM

SEM samples were fixed overnight with 4% PFA and subsequently dehydrated with graded ethanol. Fixing was required to prevent bursting of liposomes during the dehydration process. Samples were dried using a critical point dryer and were later placed on carbon tape before being sputter coated with 5 nm of gold using the Denton Desk V Sputter System (Denton Vacuum, Moorestown, NJ). All samples were imaged using the Helios NanoLab 660 Scanning Electron Microscope (FEI Company, Hillsboro, OR).

2.8. In Vitro release kinetics

50 μL of hydrogel composite system was pipetted into a 1.5 mL Eppendorf tube and allowed to equilibrate at 4 °C overnight (n = 3). 450 μL of 1× HBSS was carefully pipetted onto the samples. Samples were then kept in a 37 °C incubator for the duration of the experiment. For CDN release, 3 μL of sample was taken at each timepoint and analyzed for absorbance using a Thermo Scientific Nanodrop 2000C Spectrophotometer. A 100% control was analyzed at every time point to accommodate for any changes in CDN molecule over time at 37 °C. Each sample’s time point was normalized to the 100% control for that time point and converted to percent release. For ICG release, a standard curve was created and run on a microplate reader (Tecan, Männedorf, Switzerland) to confirm the linear range (Fig. S2b). 100 μL of sample was taken at each timepoint, transferred to a black 96-well plate, and analyzed for fluorescence intensity using a microplate reader. An excitation wavelength of 742 nm was used along with an emission wavelength of 835 nm. An excitation bandwidth of 20 nm was used along with an emission bandwidth of 20 nm. After fluorescence measurements were taken, the 100 μL samples were then returned to their original Eppendorf tubes to continue the release experiment.

2.9. IVIS release kinetics

A standard curve of ICG dissolved in 1× HBSS was created and run on the IVIS Lumina K Series III to determine the linear range (PerkinElmer, Waltham, MA) (Fig. S2c). A concentration of 7.5 μg/mL ICG was used for all animal groups. All protocols were in accordance with the guidelines for the humane treatment of laboratory animals by the National Institute of Health and the Rice University Institutional Animal Care and Use Committee in the Animal Resource Facility at Rice University. BALB/c female mice (Jackson Laboratory, Bar Harbor, Maine) were anesthetized and injected in the subcutaneous space with ICG-CDN solution and ICG-loaded biomaterials (7.5 μg/mL ICG, 0.67 mg/mL CDN, 30 μL) (n = 6). Animals were then transferred to the IVIS manifold where they were kept under isoflurane anesthesia (0.25 L/min) and maintained warm on a heated stage. Animals were imaged with an XFOV-24 lens [Field of view (FOV)-E, 22.2 cm]. Photographs and fluorescence images were acquired using the Imaging Wizard feature on the Live Imaging software (PerkinElmer, Waltham, MA). Luminescent exposures were set to 0.5 s, with the binning set at medium, the excitation filter set to 745 nm, and the emission filter set to 840 nm. Animals were imaged 10 min after ICG injection. All other instrument acquisition parameters were maintained constant.

2.10. In Vitro cellular cytotoxicity in ROC1 oral cancer cells

70 μL hydrogels were coated on 16-well chambered glass slides (Thermo Fisher, Rochester, NY). Samples were allowed to shear recover 5–10 min before adding 30,000 cells in 200 μL media on top of each hydrogel. Half of the cell media was removed and replaced every 2 days to avoid disturbing the hydrogel and cells. Cell viability was determined at day 1 and day 3 by performing live/dead assays (n = 3/group). Live/dead staining solution was prepared in PBS with 2 μM Calcein AM for live cells (Biotium, Fremont, CA), 4 μM Ethidium homodimer III for dead cells (Biotium), and 2.5 μg/mL Hoechst 33342 for nuclei (Enzo Life Sciences, Farmingdale, NY). There was significant background staining of the hydrogels that resulted in reduced resolution of the blue Hoechst channel in confocal images, so this channel was excluded for analysis. Cell media was removed and the hydrogels were washed with PBS. Samples were then stained with 100 μL of prepared solution by incubating at RT for 15 min. After staining, the hydrogels were removed from the glass chambered slides and placed cell-side down against the glass surface to ensure consistent imaging. Samples were analyzed by z-stack imaging (150 μm) using an A1 Confocal Microscope (Nikon Instruments) with a 10× objective. Image processing was done using NIS Elements (Nikon Instruments), and live/dead cell counting was performed using Fiji (ImageJ) software.

2.11. In Vivo efficacy: establishing the oral cancer preclinical model

The murine oral cancer cell line, ROC1, was provided by Dr. Roberto Rangel (The University of Texas MD Anderson Cancer Center, Houston, TX) and maintained as previously described [31]. ROC1 cells were cultured at 37 °C with 5% CO₂ in media routinely used for maintaining this cell line [31]. The orthotopic tumor model was established by injecting 500,000 ROC1 cells into the maxillary vestibule of the left oral cavity of 6 to 8-week-old wild-type C57BL/6 J female mice (Jackson Laboratory, Bar Harbor, Maine). Tumor size measurements (in diameter) were taken 3 times per week to monitor tumor growth using digital calipers to measure the length and width of the tumor (tumor area). Euthanasia criteria were defined as a tumor diameter reaching 12 mm, tumor ulceration, or weight loss of greater than 20%. These factors were considered for all endpoints in tumor growth and survival data.

2.12. Intratumoral treatments of CDN and systemic delivery of immune checkpoint inhibitors

Once 6 to 8-week-old wild-type C57BL/6 J female mice had the ROC1 orthotopic tumor model established (16–25 mm² tumor area, typically 23 days of growth), 30 μLs of CDN at 20 μg/injection (MedChemExpress, Monmouth Junction, New Jersey) either unloaded or loaded in biomaterials, was injected intratumorally (Table S4). Additionally, all treatments received a total of 100 μLs of immune checkpoint inhibitors (ICIs) composed of anti-PD-1 at 200 μg/injection (BioXCell, Hanover, New Hampshire) and anti-CTLA-4 at 100 μg/injection (Bio-XCell, Hanover, New Hampshire) intraperitoneally (IP). Data was then analyzed for Kaplan-Meier survival and tumor growth curves. Mice were maintained in standard housing conditions for the duration of the studies. All protocols were in accordance with the guidelines for the humane treatment of laboratory animals by the National Institutes of Health, the Animal Welfare Committee, and the Center for Laboratory Animal Medicine and Care at The University of Texas Health Science Center at Houston.

2.13. Statistical methods

Statistical analyses for aggregate detection were performed using One-way Anova with GraphPad Prism. Statistical analyses for in vitro cell cytotoxicity data were performed using Two-way Anova with Tukey’s Multiple Comparisons with GraphPad Prism. Statistical analyses for Kaplan-Meier survival curves were performed using the log-rank/Mantel-Cox test with GraphPad Prism. p values less than 0.05 were considered statistically significant.

3. Results and discussion

3.1. Material properties of K₂-Lip systems

We hypothesized that the delayed release of CDN from MDP-liposome composites (MDP-Lip) could increase the therapeutic efficacy of CDN in an oral cancer preclinical model. Our choice of liposome formulation was inspired by work from Ji et al. [32] that showed a 10:1 N/P (amine:phosphate) ratio resulted in maximum CDN encapsulation efficiency. DMG-mPEG-2000 was added to prevent aggregation between liposomes, with hydrogenated soy choline as the helper lipid and cholesterol for membrane stability. Our previous MDP-Lip systems were comprised of relatively low lipid:peptide molar ratio (< 2:1) [30], whereas this study required a much higher ratio to facilitate the targeted CDN drug loading (0.67 mg/mL) while keeping the N:P ratio 10:1. We calculated a starting liposome concentration of 100 mM would be needed. Therefore, we studied the effects of liposome concentration on the composite system’s material properties.

To evaluate whether K₂ or E₂ was the optimal MDP to use in composite systems with cationic liposomes, we investigated the effects of increasing liposome concentration (Lip(Empty)) on the composite systems’ opacity. As K₂ and E₂ do not absorb 800 nm light, the scattering of this wavelength of light can be used to measure the formation of aggregates, which can result from interactions between lipid and peptide. We found that K₂ could accommodate a final concentration of 44 mM of cationic liposomes (Lip-High) without changes in the scattering of 800 nm light (Fig. S3a), whereas E₂ resulted in scattering which increased with increasing cationic lipid concentration (Fig. S3b, Table S2). Cryo-EM was run on CDN-loaded E₂-Lip(Very Low), as this was the highest lipid concentration that could be combined with E₂S without aggregate formation. Images revealed deformed cationic liposomes in the presence of anionic E₂ nanofibers, showing a nanoscale picture of how E₂ de-stabilizes these liposomes even at low lipid concentrations (Fig. S3c). These results motivated our use of K₂ in all further experiments.

A high encapsulation efficiency of 90% was achieved using the previously mentioned liposomal formulation. The encapsulation efficiency was determined by centrifuging the liposome suspension in a centrifugal filter, analyzing the free CDN concentration in the filtrate, and normalizing it to the total concentration of CDN in the liposome suspension. As electrostatic attraction between cationic DOTAP and anionic CDN is responsible for liposomal encapsulation, we suspect that the use of polyvalent ionic buffer may slightly decrease encapsulation. Ji et al. showed that 100% encapsulation could be achieved with the appropriate N:P ratio, but their liposomal formulation was also made in the zwitterionic HEPES buffer [32]. We used HBSS in our formulation because polyvalent ions are required for MDP hydrogel formation. However, the phosphate anions in HBSS may compete with CDN for electrostatic interactions with DOTAP and prevent 100% encapsulation from being achieved. Given the high encapsulation efficiency and the success of free CDN mixed with K₂ [23], we decided to move forward with this mixture of 10% free CDN/90% liposomal CDN without purification to simplify the liposome preparation process.

Oscillatory rheology experiments showed that K₂ forms self-healing hydrogels in the presence of CDN-loaded Lip(Low), Lip(Med), and Lip (High) liposomes (Fig. 2a–b, Fig. S4a). An increase in storage modulus, G’, was seen with increasing liposome concentration, similar to results seen with other liposome-hydrogel composite systems [33]. Repeated shear-thinning of K₂-Lip(High) showed the composite’s ability to self-heal multiple times, suggesting that K₂ can withstand the loading of high concentrations of liposomes while retaining robust materials properties (Fig. 2b). This was a surprising discovery, since K₂-Lip(High) is primarily lipid-based by mass, though it still retains the self-healing ability of the peptide hydrogel. These results show the potential for MDP hydrogels to deliver high concentrations of liposomal therapeutics by syringe without degradation of materials properties.

Fig. 2.

Material properties of CDN-loaded K₂-Lip systems. Oscillatory rheology experiments of MDP hydrogels with varying concentrations of liposomes including a) frequency sweep and b) step-strain recovery test. c) Intensity distribution of K₂-Lip(High) and Lip(High) using DLS. d) SEM image of CDN-loaded K₂-Lip(High). Lip (Low) = 11 mM liposomes, Lip(Med) = 22 mM liposomes, Lip(High) = 44 mM liposomes.

To assess whether the liposomes retained their structure while at such high concentrations in K₂, we used dynamic light scattering (DLS) to study the changes in scattering with increasing liposome concentrations. We suspected that if the liposomes’ structure was being compromised from interacting with K₂, the scattering profile would be different between liposomes that were and were not combined with K₂. The scattering profile of the intact CDN-loaded K₂-Lip(High) composite was the same as free Lip(High) (Fig. 2c), indicating that the liposomes retain their size and shape after combination with K₂. SEM images of K₂-Lip(High) after fixation (Fig. 2d) reveal a porous, nanofibrous network typical of a peptide hydrogel (Fig. S4b). After running Cryo-EM on CDN-loaded K₂-Lip(High), the composites appeared too concentrated to acquire high-resolution images even at a 1:100 dilution, though thick, fibrous structures can be seen surrounding what appear to be intact liposomes (Fig. S4c). Cryo-EM images of liposomes alone (Fig. S4d) were also taken and showed their expected spherical structures. After confirming that K₂ and 44 mM liposomes loaded with CDN (K₂-Lip(CDN)) formed self-healing composites, in vitro and in vivo experiments were conducted to assess drug-delivering capabilities.

3.2. In vitro release of CDN and ICG

An equilibrium in vitro release assay compared the release of CDN between K₂-Lip(CDN) and STINGel. We found that K₂-Lip(CDN) released 46% of total CDN by the 24-h timepoint, whereas STINGel released 96% over the same period (Fig. 3a). These results confirmed the ability of charge-matched MDP-liposome composite systems to delay the release of CDN in vitro substantially longer than K₂ without liposomes. These and other in vitro release assays are fast, easy, and inexpensive, but they do not encompass many important factors in drug release that exist in vivo. Therefore, they are best utilized for direct comparisons of the relative ability of a material to control release rather than an approximation of practical release efficiency. To assess the in vivo relevance of these findings, we studied the release kinetics of these systems in murine models after subcutaneous injection.

Fig. 3.

Release kinetics from biomaterial systems. a) In vitro release kinetics of CDN from K₂-Lip(CDN) and STINGel (n = 3 for both groups). b) In vivo release kinetics of ICG from K₂-Lip(CDN), STINGel, Lip(CDN), and an ICG/CDN mixture (n = 6 for all groups). c) Schematic of subcutaneous biomaterial injection for in vivo release. d) Representative IVIS images of ICG release in vivo at days 0, 1, 18, 37, and 51.

To assess whether our composite system could delay the release of an anionic small molecule in vivo, indocyanine green (ICG) was chosen as a model drug for release. ICG is a convenient molecule to use for measuring release kinetics due to its red-light excitation and emission wavelengths, making it visible using fluorescence imaging of live animals. To confirm that ICG was a suitable molecule to use, an in vitro ICG release assay was performed. ICG-encapsulated liposomes were created and loaded into K₂, and ICG release from this material was monitored and compared to release from a K₂ control hydrogel. Over 24 h, only 4% of ICG was released from the composite system, whereas 21% was released from K₂ (Fig. S5). The magnitude of ICG release was smaller than that of CDN, however it is important to note crucial differences between the two in vitro release experiments. Solubility differences between ICG and CDN limited the amount of ICG that could be loaded into liposomes. Therefore, the concentration of ICG in the composite system (88 μg/mL) was much lower than the concentration of CDN in the composite system (670 μg/mL). However, as both in vitro experiments showed the same general trend of delayed release achieved by the composite system, we decided to move forward with ICG in the in vivo release experiment.

3.2.1. In vivo release of ICG

CDN-loaded liposomes doped with ICG were generated to track the in vivo release of ICG from our composite system using IVIS (Table S1). BALB/c female mice were injected with 30 μL of each group (ICG/CDN mixture, ICG/CDN liposomes, ICG STINGel, ICG/CDN liposomes in K₂) and imaged over time (Fig. 3b–d). The concentration of ICG was determined using a standard curve generated in 1× HBSS and quantified via imaging on the IVIS system (Fig. S2c). To minimize fluorescence quenching, 7.5 μg/mL ICG was used as it was also within the linear range.

Free ICG/CDN mixture left the site of injection very rapidly as the intensity was approximately 50% that of the other groups at the 10-min time point (Fig. 3b). This was expected since the mixture was not loaded in any biomaterial to keep the fluorophore localized. Though signal can be seen up until day 10, the intensity was extremely low (10⁸ total radiance efficiency) compared to all other groups (10¹⁰ total radiance efficiency) after 12 h, suggesting that the majority of free ICG is cleared within the first day (Fig. 3b). STINGel released ICG over a period of 2.5 weeks, highlighting the advantage of our previous hydrogel system [23] for delayed release compared to free injections. CDN-loaded liposomes were visible at the injection site of all mice up to 4 weeks, which could be due to the limited vasculature in the subcutaneous space. As these liposomes are ~140 nm (Table S1), they do not diffuse as readily as the free fluorophore. K₂-Lip(CDN) had the longest observable release of all four groups. On day 51, K₂-Lip(CDN) could still be seen in all mice, and by day 58, it was the only group that continued to have detectable ICG in all mice. These results were promising, as they showed that high concentrations of liposomes could be loaded into the K₂ peptide hydrogel for longer release of small molecules than either biomaterial alone. Having shown extended release of small molecules from K₂-Lip(CDN) both in vitro and in vivo, we wanted to assess its use as a drug delivery system in a preclinically-relevant oral cancer model.

3.3. In Vitro cellular cytotoxicity in ROC1 oral cancer cells

Cellular viability of ROC1 oral cancer cells was evaluated at days 1 and 3 to determine cytotoxicity of K₂-Lip(CDN). On day 1, the viability was high for cells plated on K₂-Lip(Empty), K₂-(Empty), K₂-Lip(CDN) and STINGel hydrogels (Fig. S6a–d). On day 3, the ROC1 cells on the K₂-Lip(Empty), K₂-(Empty), and STINGel remained viable (Fig. S6e–g), whereas cells on the K₂-Lip(CDN) hydrogel showed decreased viability (Fig. S6h). Cellular viability was significantly higher in the K₂-(Empty) at 99.7% and STINGel at 99.4% compared to K₂-Lip(CDN) at 69.2% (Fig. S6i), indicating that K₂-Lip(CDN) is cytotoxic to ROC1 oral cancer cells. This data further motivated the use of K₂-Lip(CDN) for intratumoral injection in an oral cancer animal model.

3.4. Treatment of CDN results in brief tumor regression in ROC1 oral tumors and cellular cytotoxicity

To establish a clinically relevant preclinical model of HNSCC, we chose oral squamous cell carcinoma, as it constitutes the most prevalent subtype of HNSCC. We utilized a non-HPV-associated ROC1 murine cell line that contains tobacco mutation signatures similar to the ones found in human oral cancer [31]. We established the ROC1 cell line in the oral cavity of mice. Once the tumors reached an area between 16 and 25 mm², typically 23 days post-inoculation, mice were treated intratumorally with either CDN or saline (Fig. 4a). We categorized tumor responses by partial responses or complete responses. Partial responses indicate there was tumor regression while complete responses indicate complete tumor regression without tumor recurrence. After six doses of CDN administration, CDN-treated mice experienced tumor regression and partial responses were observed in 2 out of 8 mice (Fig. 4b, c). However, following the first six doses of CDN administration, continued intratumoral CDN showed no benefit and tumors recurred. The saline group had tumor growth without any instances of tumor regression (Fig. 4d). ROC1 tumor-bearing mice treated with CDN had extended survival compared to the saline group (Fig. 4e). After observing tumor regression with CDN treatments for a two-week duration, we were encouraged to evaluate a biomaterials-based approach for local and sustained delivery of CDN to the tumor microenvironment in combination with immune checkpoint inhibitors.

Fig. 4.

ROC1 tumor-bearing mice treated with repeated intratumoral CDN results in brief tumor regression and survival. a) Schematic of CDN intratumoral (IT) treatment regimen for tumor-bearing mice. b) Mean tumor growth curves up until the first mouse was euthanized. Error bars indicate standard deviation. Individual tumor growth curves of mice treated with multiple c) CDN (n = 8) or d) Saline (n = 8) injections. Two out of eight mice in the CDN group had a partial response indicated by tumor regression, while the saline group did not have any instances of tumor regression. e) Kaplan Meier survival curve of the treatment groups. **CDN vs Saline, p < 0.008.

3.5. Locally delivered composite combined with systemic delivery of ICIs in the ROC1 preclinical model

Previous ROC1 preclinical studies indicate that CDN administration in combination with an immune checkpoint inhibitor results in tumor regression and extended survival in tongue tumors [31]. We therefore treated ROC1 oral cavity tumor-bearing mice with or without CDN-loaded biomaterials and ICIs (Table S4, Fig. 5a). When treated with K₂-Lip(CDN) + ICIs, tumors regressed 1-week post-treatments (Fig. 5b). Specifically, one of the twenty-four ROC1-tumor bearing mice did not have a palpable tumor in the buccal space 56 days post-inoculation. Only one mouse from the K₂-Lip(CDN) + ICIs group lost significant weight by 63 days post-inoculation (Fig. S7a). A second K₂-Lip(CDN) + ICI-treated mouse had tumor regression by 80 days post-inoculation, but ultimately the tumor reoccurred, and the mouse reached the humane endpoint. In this case, the mice exhibited brief tumor regression, indicating a partial response. Lip(CDN) + ICIs and STINGel + ICIs groups did not display tumor regression like the K₂-Lip(CDN) + ICIs group (Fig. 5c, d). This suggests the need for the single dose of CDN-loaded liposomes to be in the hydrogel carrier to effectively deliver CDN intratumorally. The body weights of the mice were regularly monitored, and there was no noticeable weight loss from any of the other treatment groups (Fig. S7b–h). CDN 6× + ICIs had tumor regression after two weeks of treatments, but ultimately the tumors recurred (Fig. 5e), while CDN 1× + ICIs only had one instance of tumor regression (Fig. 5f). These findings indicate that a single dose of CDN without a biomaterial to localize the small molecule is insufficient for this aggressive oral cancer model. The tumor growth curves for mice treated with ICIs (Fig. 5g) and those given empty controls with ICIs (Fig. S8a, b) did not display tumor regression. This suggests that the addition of CDN improves treatment in the oral cancer model. The K₂-Lip(CDN) + ICIs-treated group was significantly different compared to CDN 1× + ICIs and STINGel + ICIs (Fig. 6). This indicates that K₂-Lip(CDN) + ICIs extends survival compared to a single dose of CDN with or without the hydrogel. There was no significant difference between K₂-Lip(CDN) + ICIs and CDN 6× + ICIs, indicating that 20 μg of CDN delivered via K₂-Lip(CDN) enables similar survival as 120 μg of free CDN delivered over two weeks.

Fig. 5.

Tumor growth of the different experimental groups based on euthanasia timepoints resulting from excessive tumor burden. a) Schematic of treatment regimen for tumor-bearing mice administered CDN intratumorally (IT) and ICIs intraperitoneally (IP). Individual tumor growth curves of mice treated with b) K₂-Lip (CDN) + ICIs (n = 24), c) Lip(CDN) + ICIs (n = 14), d) STINGel + ICIs (n = 8), e) CDN 6× + ICIs (n = 14), f) CDN 1× + ICIs (n = 14), or g) ICIs (n = 14). Bolded fractions indicate partial responses.

Fig. 6.

Kaplan Meier survival curve of the treatment groups. The total experimental period was 86 days post-tumor cell inoculation. Only one mouse from the K₂-Lip(CDN) + ICIs-treated C57BL/6 mice survived until the endpoint of the study. ns: not significant, *p < 0.05, **p < 0.002, ***p < 0.0002, ****p < 0.0001.

There was slight extended survival of Lip(CDN) + ICIs compared to CDN 1× + ICIs and STINGel + ICIs, but ultimately these CDN-loaded liposomes injected without a hydrogel carrier were not able to extend survival as effectively as repeated intratumoral injections of CDN 6× + ICIs. The CDN 6× + ICIs treated group had extended survival compared to the CDN 1× + ICIs, Lip(CDN) + ICIs, STINGel + ICIs and ICIs. All treatment groups had extended survival compared to the K₂-(Empty) + ICIs, with the exception of Lip(Empty) + ICIs (Fig. S8c). The CDN 1× + ICIs group did not have extended survival compared to the Lip(Empty) + ICIs. The tumor-bearing mice that were treated with ICIs had extended survival compared to CDN 1× + ICIs. By removing the K₂-Lip(CDN) + ICIs’ long-term survivor, we did not observe significant differences between the treatment groups (Fig. S9). We have previously seen CDN-loaded MDP hydrogels induce regression and extended survival in early-stage HPV-associated murine tumors [23]. Here, similar effects were observed in established non-HPV tumors. Future work should consider treating early-stage non-HPV tumors to assess potential for complete responses. Additionally, the initial rapid release of STING agonist can result in excessive activation of local innate immunity and global inflammation [34] making it more suitable for tumors to thrive. A study found that conducting either high or repetitive doses of CDN intratumorally results in systemic distribution, while low-dose immunogenic regimens induce local activation of tumor-specific CD8⁺ effector T cells that are responsible for durable anti-tumor immunity and can be enhanced with checkpoint inhibitors [35]. Therefore, it is important to consider the release kinetics of CDN to achieve anti-tumor responses and the immune infiltration of lymphoid organs and tumors. Previous studies have found a lipid-based nanoparticle platform carrying Mn²⁺ and STING agonist activated innate immune cells and T cell populations in lymphoid tissues in an acute myeloid leukemia model [36].

Future studies will characterize the immune landscape of the ROC1 orthotopic tumor model and associated lymphoid tissues to inform the design of combination treatment strategies. Further improvements to the composite system can also be made through a dose-response study to determine if the therapeutic dosage can be lowered even more. As this composite system is modular, incorporation of different liposomal formulations of small molecules could be interesting to investigate. Overall, the lack of significant differences between K₂-Lip(CDN) and the six doses of CDN suggests that the K₂-Lip system prolonged availability of CDN locally.

4. Conclusion

With tumor regression seen after multiple doses of CDN in a ROC1 tumor-bearing murine model, we hypothesized that a biomaterials approach could extend the release of CDN and ultimately decrease the number of doses needed for tumor regression. A charged-matched MDP-liposome composite system was chosen to test our hypothesis, and with this, we discovered the high-loading capacity of cationic liposomes in a K₂ hydrogel. Even at a high molar ratio of 8:1 lipid:peptide, a self-healing composite hydrogel formed and showed delayed release kinetics of CDN in vitro and ICG/CDN in vivo. The K₂-Lip(CDN) in vitro indicated cellular cytotoxicity in the ROC1 oral cancer cells after 3 days. In a ROC1 tumor-bearing murine model, we found CDN loaded in K₂-Lip (CDN) had extended survival compared to both the single injection and STINGel. The overall survival between a single dose of K₂-Lip(CDN) and six repeated doses of CDN are equivalent, supporting the hypothesis that K₂-Lip systems can extend the in vivo persistence of CDN locally. In addition to fewer doses, the peptide liposome formulation required one-sixth the total dose of CDN to achieve equivalent results. The favorable materials properties and extended-release capabilities could allow MDP-liposome composite systems to be used to treat other diseases that require the delayed release of localized therapeutics.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.jconrel.2026.114730.

Data availability

Data will be made available on request.