Systematic engineering of Layer-by-Layer anti-miR coated wound dressings with tunable release kinetics

Abstract

The Layer-by-Layer (LbL) technique enables conformal coating of commercial wound dressings with oligonucleotide therapeutics and polymeric excipients to achieve controlled delivery to non-healing wounds. We previously demonstrated a proof-of-concept LbL dressing containing miRNA inhibitors (anti-miRs); however, release occurred on a single time scale. Because wound healing is dynamic, temporally tuning anti-miR delivery to the wound may alter therapeutic effects. Here, we provide an approach to modulate oligonucleotide release kinetics from coated wound dressings, exploring the design space of excipients, formulation parameters, and oligonucleotide structure. Incorporation of Laponite nanoclay prolonged release, while poly(acrylic acid) accelerated release kinetics. Hydrolyzable poly(β-amino esters) with varying hydrolysis rates were also used to further tune release kinetics. A fast-release formulation demonstrated 40% release over 4 hours, while a slow-release formulation took 24 hours to release the same amount. We evaluated these dressings in an excisional chronic wound model in diabetic (db/db) mice using anti-miR-92a (92ai), selected for its ability to enhance angiogenic signaling and promote wound closure. Dressings coated with 92ai accelerated healing compared to non-targeting scramble anti-miR (Scr) controls, with closure rates accelerated at times corresponding to release kinetics. Although ultimate wound closure was similar across release conditions, analyses of gene expression, cellular composition, histology, and immunofluorescence suggested distinct underlying mechanisms. Differences were particularly evident in macrophage phenotype and angiogenesis, suggesting that release timing influences biological pathways active during repair. Overall, these findings demonstrate that LbL dressings can be engineered for temporally controlled oligonucleotide delivery and that release kinetics may shape mechanisms driving wound closure.

Graphical Abstract

1. Introduction

Chronic wounds, or ulcers, are a common complication of diabetes that result in reduced quality of life, increased mortality, and high medical costs.^1–4 Current treatments are insufficient, as they do not target the molecular underpinnings of the impaired healing state, leaving patients with wounds that persist or recur. The impaired healing state is partially regulated by microRNAs (miRs), which function at the post-transcriptional level to modulate gene expression.^5–7 Aberrant expression of miRs can be targeted with anti-sense oligonucleotide therapeutics called anti-miRs; however, anti-miRs require carrier systems for their effective delivery.⁸

Our lab designed the Layer-by-Layer (LbL) electrostatic-driven thin film assembly process for the conformal coating of various therapeutics onto surfaces for their controlled delivery.⁹ Previously, we leveraged LbL to create conformal coatings on commercial wound dressings or sutures that control the release of oligonucleotide therapies into the wound bed.^10–13 More recently, we adapted these dressings for the delivery of single stranded anti-miR oligonucleotide therapy.^14–16 In a recent proof-of-concept study demonstrating our approach, we loaded pro-angiogenic anti-miR-92a (92ai) onto the dressings. We showed that controlled delivery from our coated wound dressings enabled preferential delivery of 92ai to endothelial cells, increased wound closure, and increased sex-dependent angiogenesis.¹⁴

One remaining question after this previous study was how release rate of 92ai alters its efficacy. Two contrasting hypotheses exist: 1) faster release may lead to higher levels of anti-miR therapy in the wound cells at once, increasing miR inhibition and downstream effects during a critical early phase of wound healing. This is because anti-miRs work in an anti-sense manner, sequestering highly expressed miRs and outcompeting their binding to target mRNAs.¹⁷ 2) Slower release kinetics may target later phases of wound healing, including angiogenesis in the proliferative phase.¹⁸ Preliminary evidence towards both hypotheses exists. Some studies suggest a threshold effect of dosing anti-miRs, and thus higher bolus doses may not increase efficacy.^19,20 Other studies have suggested that delivery of the pro-angiogenic growth factor vascular endothelial growth factor (VEGF) works best when delivered in a sustained manner as opposed to a bolus release.²¹ It remains unknown how these previous studies may apply to anti-miR delivery.

The LbL platform is modular, allowing the incorporation of different components and excipients that can modulate release kinetics to answer our question. For example, modification of polymer molecular weight, charge density, and pK_a of component polyelectrolytes can all affect release kinetics.²² Other excipients can also be incorporated into the assemblies, including those that allow additional covalent or non-covalent crosslinking or those that use high aspect ratio nanomaterials to limit diffusion within and out of the films.^23–25 Laponite nanoclay is one such material that can be used as a diffusional barrier to slow release kinetics from electrostatic assemblies.²⁶ It has been used in several drug delivery systems for its biocompatibility and in LbL to control the release of protein and small molecules.^27–29 It has a high aspect ratio and, due to its charge profile, deposits parallel to the underlying dressing. This allows the clay to increase film tortuosity and greatly reduce diffusion of the encapsulated therapeutic.³⁰

In this study, we explore how changing excipients in coated wound dressings alters release kinetic profiles of loaded oligonucleotide therapy. We highlight the engineering of these dressings through systematic investigation. Using a similar formulation with few materials alterations, we find that poly(acrylic acid) accelerates release while high aspect ratio Laponite nanoclay slows release of oligonucleotide from the dressings. We then use these dressings to investigate our tunable release system in a murine db/db excisional diabetic ulcer model. We observe that 92ai speeds wound closure over a non-targeting scramble anti-miR (Scr) control, with the timing of wound closure acceleration dependent on dressing release kinetics. Further probing suggests potential differences in macrophage phenotype and angiogenesis between fast- and slow-release dressings. Ultimately, this study demonstrates how engineering and formulation science can be used to modulate the release of single stranded anti-miR therapy from coated wound dressings to access different stages of the wound healing process.

2. Methods and materials

2.1. Formulation materials and solution preparation

Poly(acrylic acid) (PAA) with an average molecular weight of 20 kDa, poly(styrenesulfonic acid) (PSS) with an average molecular weight of 18 kDa, and linear poly(ethyeleneimine) (LPEI) transfection grade with an average molecular weight of 25 kDa were purchased from Polysciences, Inc. Dextran sulfate (DS) with an average molecular weight of 500 kDa was obtained from VWR. Laponite XLG nanoclay was sourced from Southern Clay Products. Poly2 and Poly3 are of the poly(β-aminoester) class, and their synthesis is detailed in our previous work. As previously reported MALDI-TOF based molecular weight for Poly2 was M_n=3.8 kDa; M_w=4.4 kDa and for Poly3 was M_n=2.3 kDa/M_w=2.6 kDa. Poly3 has higher hydrophobicity, lower pK_a, higher cooperativity in deprotonation, and stronger RNA binding compared to Poly2, leading to slower release kinetics from LbL coatings and higher transfection efficacy.²²

Oligonucleotides used in this study (Table S1) are a scramble non-targeting anti-miR with AlexaFluor 647 fluorophore (anti-miR-AF647) or without (Scr), anti-miR targeting miR-92a (92ai), and a dsDNA mimic of siRNA labeled with AF647 (siD7) used for cheaper cost and greater stability.³¹ Lyophilized oligonucleotides were dissolved in ultrapure water to 50 μM and stored at −20 or −80°C until use, per manufacturer instructions. Stock concentrations were validated with a Nanodrop UV/Vis spectrophotometer (Thermo Fisher) on the RNA-40 or DNA-50 setting, as appropriate.

PAA, PSS, DS, Poly2, and Poly3 were dissolved at 1 mg/mL in pH 5.2 100 mM sodium acetate buffer. PBAEs were dissolved freshly on the day of formulation due to their degradable nature. Laponite nanoclay was dissolved at 1 mg/mL in ultrapure water without buffer. LPEI was dissolved at 0.25 mg/mL in pH 6 100 mM sodium acetate buffer. For base layers, wash buffer contained pH 5.2 10 mM sodium acetate. Sodium acetate buffer at pH 6 and 10 mM was used for washes during RNA layers and for further dilution of oligonucleotides.

2.2. Layer-by-layer assembly

LbL assembly proceeded as described in our previous work.¹⁴ Briefly, Tegaderm^™ Non-Adherent Contact Layer wound dressings (3M) were cut to size, soaked in 70% ethanol, plasma cleaned, and deposited in a solution containing the first layer of polyelectrolyte. Tegaderm^™ woven nylon dressings were chosen since they are porous and non-adherent, commonly used for wound-healing applications, generally biocompatible, and amenable to surface functionalization through polymer adsorption. A slide stainer automated assembly of the various polyelectrolytes into a thin film coating on the Tegaderm^™ wound dressing surface. Polymers were deposited for 10 minutes, RNA was deposited for 15 minutes, and washes lasted 30 seconds. Once coated wound dressings dried, they were stored at −80°C.

2.3. Total loading and In vitro release kinetics

Total loading measurements and in vitro release kinetics occurred similar to our previous work,¹⁴ leveraging fluorescently labeled oligonucleotides. Total loading was assessed in two ways, as denoted in the text. Early experiments disassembled the thin film coatings using high salt conditions (3M sodium chloride overnight), and later experiments performed disassembly over time using a high protein solution of 100% fetal bovine serum (FBS) for 6–10 days. Standard curves were made in respective diluents. In vitro release kinetics were performed in 1% w/v bovine serum albumin (BSA) in 1x phosphate buffered saline (PBS) or in 100% FBS.

2.4. Cell culture and In vitro transfection experiments

Human umbilical vein endothelial cells (HUVECs) derived from pooled donor sources were purchased from Lonza and used for studies investigating transfection of anti-miRs released from wound dressings. Culture media was EGM-2, which contains added growth factors and supplements (Lonza). Periodic testing for mycoplasma was performed by the MIT Koch Institute ES Cell and Transgenics core using a MycoAlert kit (Lonza). Cells were grown at 37°C and 5% CO₂ using a humidified incubation chamber. Cell culture followed best practices.

2.5. Mouse wound studies

To recapitulate the healing of diabetic ulcers, a full-thickness excisional wound model in genetically diabetic (BKS.Cg-Dock7^m+/+Lepr^db/J, homozygous for Lepr^db, db/db) mice was employed. This model is well studied in the literature and thus provides a comparison to other studies.^32–34 Given that we observed differences in vascularization with 92ai in male mice, but not females in previous work,¹⁴ we proceeded to use only male mice in this study. Mice were 11–14 weeks old. To better assess both earlier and later aspects of wound healing, endpoints were at both 5 days and 9 days post-wounding. The primary outcomes of the efficacy study were: 1) wound closure, as measured by digital photography, 2) granulation tissue thickness, as measured with morphometric analysis, and 3) CD31+ cell count, as measured with flow cytometry. 8 wounds/treatment/release kinetics/endpoint were used, for 32 mice and 64 wounds total. For some preliminary studies, C57Bl6/J mice were used.

2.5.1. Model

All animal research was vetted and approved by the MIT Committee on Animal Care (protocols #0121-004-24, 2310000591) and by the Army Animal Care and Use Review Office (protocol #PR202365). Information on the model, husbandry, pre-op care, wounding and dressing application, post-op care, and tissue harvesting for this model are well-described in our previous work.¹⁴ Note that dressings were not reapplied in this study, as they have been in our previous work, since release kinetics was the independent variable and re-application might obfuscate effects.

Briefly, one day before wounding, a blood glucose of greater than 300 mg/dL was confirmed with a glucometer (Contour). Then, mice were shaved, and a Nair depilatory was used to remove fur. The following day, mice were anesthetized with isoflurane, provided analgesic with sustained release buprenorphine, wounded bilaterally with a 6 mm biopsy punch to a full-thickness wound depth, and dressed with coated wound dressings. The assembly was covered with a clear adhesive to keep the dressings on the wound and prevent the mice from manipulating their wounds. Formulation was randomized by mouse. Mice were allowed to recover, and their comfort was maintained over the remainder of the study. Endpoints were on days 5 or 9. The mice with an endpoint of day 9 had coated dressings removed on day 5 to prevent tissue infiltration into the non-degradable nylon, and a new adhesive was placed on these wounds.

2.5.2. IVIS imaging

For studies looking at release kinetics from wound dressings coated with siD7 or anti-miR-AF647 in vivo, an in vivo imaging system (IVIS, PerkinElmer) was used. At various points after dressing application, mice were anesthetized with isoflurane and their wounds imaged. An excitation wavelength of 640 nm and an emission wavelength of 680 nm were used. LivingImage (PerkinElmer) software was used for quantification. Fluorescence was normalized to unwounded untreated skin or non-fluorescent control treated wounds.

2.5.3. Gross wound imaging

To assess the closure of wounds over time, the gross wound was imaged at days 0, 5, 7, and 9, adjacent to a surgical ruler. A digital camera on the iPhone 14 (Apple) was used. ImageJ was used to outline the tongue of epithelialization into the new wound and measure the open wound area. Mice with the day 9 endpoint were used. The wound area over time was normalized to the initial wound area. Area under the curve (AUC) of wound closure was approximated with the trapezoid rule. Healing rate was calculated between each two time points as the linearization of relative wound area over time, and that for 92ai treated wounds normalized by the average for Scr treated wounds by release kinetics. The time point for healing rate was the average of the two time points.

2.5.4. Tissue harvesting

To perform histology, flow cytometry, and RT-qPCR of the wounds, tissue was harvested on day 5 or 9. At the endpoint of the study, mice were euthanized. Adhesives and any remaining dressing material were carefully removed from the wound bed. Forceps and scissors were used to excise the wound with a few mm of circumferential clearance. The wound was halved; half was placed in zinc fixative (Becton Dickinson) for histology. For the other wound half, subcutaneous fat was scraped using blunt dissection. 1/6 of the wound was placed in Qiazol (Qiagen) in MACs M tubes (Miltenyi Biotec) for later RNA extraction. The remaining 1/3 of the wound was placed in PBS to keep the tissue moist prior to dissociation for flow cytometry studies. All tissue was kept on ice for preservation until further processing. Unwounded, untreated (UW, UT) was also collected from four mice per treatment condition from the caudal aspect of the shaved/naired dorsum.

2.6. RNA extraction and RT-qPCR

An octodissector (Miltenyi Biotec) was used to homogenize the wound tissue in RNA extraction buffer. The “RNA” setting was used. For in vitro cell samples collected for RNA extraction, cells were washed once with PBS and then Qiazol was added to the cells for at least 5 minutes.

RNA was then extracted from all samples using a miRNEasy kit (Qiagen), following the manufacturer instructions. The resulting RNA was of high quality (most samples had 260/280>1.9). RNA was extracted from the columns using 20 μL of ultrapure water. cDNA synthesis for downstream miRNA expression analysis was performed with a miRCURY reverse transcriptase cDNA synthesis kit (Qiagen) and cDNA synthesis for downstream mRNA expression analysis was performed using as Quantitect reverse transcriptase kit (Qiagen). cDNA library synthesis was carried out according to manufacturer instructions. Synthesized cDNA was stored on ice for short term use or at −20°C for long term storage. Total RNA was stored at −80°C.

RT-qPCR was carried out as described previously.¹⁴ Primers used to assess mRNA expression are shown in Table S2, and primers used to assess miRNA expression are shown in Table S3. qPCR with SYBR green readout was carried out using a Quantstudio Pro 6 (Applied Biosystems). For in vivo wounds, analysis was performed using the ΔΔC_t method, where normalization was set to both reference gene(s) and to the contralateral control receiving the control therapy (i.e. 92ai relative to Scr treated wound). In the in vivo efficacy study, reference transcript expression was pooled using the BestKeeper method.³⁵ For miRNA analysis, these references included SNORD65 and SNORD68. For mRNA analysis, these references included Gapdh, Hprt1, Ywhaz, and Rps18. Analysis of gene expression from in vitro experiments was performed similarly, using secondary normalization to the average expression of Scr treated cells. In these human cells, SNORD48 was used as the reference transcript for miRNA analysis, and ACTB was used as the reference transcript for mRNA analysis.

2.7. Histology and morphometric analysis

Preparation of histological sections was performed by the MIT Koch Institute histology core. The full procedure for staining with hematoxylin and eosin (H&E) is provided in our former work.¹⁴ Sectioned slides were scanned with an Aperio Slide Scanner (Leica), and ImageScope was used for morphometric analysis, as previously described.

2.8. Immunofluorescence staining

Immunofluorescence staining was performed by the BIDMC Confocal Imaging Core Facility using the antibodies in Table S4. Unstained slides were deparaffinized and rehydrated. Antigen retrieval was performed with 10 mM sodium citrate pH 6.0 buffer. Slides were then permeabilized with 0.05% Triton X-100 in 2% BSA (Jackson ImmunoResearch Lab) for 15 minutes at room temperature, washed triply in tris-buffered saline (TBS), and incubated with 2% BSA for an hour at room temperature. Slides were incubated overnight at 4°C with the primary antibodies, washed triply in TBS with 0.1% tween-20, and then incubated with secondary antibodies for 90 minutes. Nuclei were counterstained for 2 minutes with 1:10000 Hoechst 33342 (Life Technologies) followed by 3 washes in TBS. Prolong gold anti-fade mounting media (Life Technologies) was layered on the slides, and cover slips were placed.

2.9. Tissue dissociation and flow cytometry

To study the distribution of wound cells and their phenotypes, wound tissue was dissociated as in our previous work.¹⁴ Flow staining was performed using the panel in Table S5. In addition to wound tissue, UW, UT skin was also stained. UW, UT skin was additionally used for the single color and fluorescence minus one (FMO) controls.

2.9.1. Initial efficacy study

This study used a similar gating strategy to that in our earlier work and was adapted from Sakamoto et al.;^14,36 however, the protocol was modified to accommodate both surface and intracellular markers. Cells were spun at 500 x g for 3 min at 4 °C and resuspended in 1:500 Zombie UV Fixable viability dye. After 15 minutes of incubation at room temperature, cells were washed twice with buffer. Fc block was performed using 1:200 Fc antibody in buffer for 5 minutes on ice. Cells were spun down and resuspended in surface antibodies for 25 minutes at 4 °C. After washing with buffer, cells were fixed for 15 minutes at room temperature using fixation/permeabilization solution from a Transcription Factor Staining Buffer Set (eBioscience). Following another wash with buffer, the cells were permeabilized with the kit’s permeabilization buffer for 15 minutes at room temperature. The intracellular staining antibodies were then diluted 1:200 in permeabilization buffer and incubated for 40 minutes at room temperature, protecting the cells from light. After staining, two washes were performed with permeabilization buffer followed by a final wash with buffer. Ultracomp beads and Arc amine reactive compensation beads (both Invitrogen) were also stained for compensation.

Flow cytometry was performed on a BD FACS Fortessa flow cytometry analyzer. Zombie UV (515/30) was read with a 355 nm laser. BV421 (450/50) was read with the 405 nm laser. AF488 (530/30), PE (575/25), and PE-Cy7 (780/60) were read on the 488 nm laser. eFluor660 (670/30) was read on the 640 nm laser. Analysis was per the diagram in Figure S1.¹⁴

2.9.2. Add-on efficacy study

Based on the desire to better capture CD31 staining and to differentiate the CD68⁺ macrophages by phenotype, we performed additional staining and flow cytometry on the cells stained as above. The cells were spun down by centrifuging at 500xg and room temperature for 3 minutes. The additional antibodies for CD31-APC-Cy7, CD86-BV605, and CD206-PE-Cy5 were diluted 1:50 given that the cells were previously fixed, and cells were incubated with these antibodies at room temperature in the dark for 40 minutes. Cells were washed twice with buffer and then re-fixed. After an additional wash with buffer, flow cytometry was performed. Compensation beads were stained in parallel.

To accommodate the additional colors used in this panel, we used a BD FACS Symphony A3. Zombie UV (515/30) was read with a 355 nm laser. BV421 (450/50) and BV605 (610/20) were read with the 405 nm laser. AF488 (515/20) was read with a 488 nm laser. PE (586/15), PE-Cy5 (670/30) and PE-Cy7 (780/60) were read on the 561 nm laser. eFluor660 (670/30) and APC-Cy7 (780/60) were read on the 640 nm laser. Analysis was performed as above and using the gating in Figure S2.

2.10. Statistics

Statistical testing was implemented using Graphpad Prism 10 (Dotmatics) and R studio with R 4.3.1. For each experiment, the statistical procedure is noted. Except for ordinal histology scoring, parametric statistics were used due to an assumption of normality. Ordinal histology scoring employed a cumulative link model. Outliers were removed using ROUT with Q=1% and excluded from analysis. Any missing data other than outliers is also noted, as appropriate. A family-wise type I error rate of 5% was chosen by convention, and corrections were used to account for multiple hypothesis testing. Unless noted, significant statistical tests and tests approaching significance (p≤0.1) are plotted or enumerated in the figure caption. Results of testing are not reported if the results showed non-significance or not approaching significance.

3. Results and Discussion

3.1. Laponite nanoclay can be used in various coating architectures to modulate release kinetics of a model oligonucleotide

In our initial proof-of-concept study of LbL anti-miR coated wound dressings, we demonstrated only a single formulation of excipients. This formulation released oligonucleotide cargo mostly within the first day both in vitro and in vivo.¹⁴ This is a relatively fast release rate compared to the wound healing process, which occurs on the order of days to weeks. We wanted to engineer slow-release formulations, which could allow sustained oligonucleotide therapy delivery over multiple days of the wound healing process. We first tested Laponite nanoclay, which has a high aspect ratio that slows diffusion and which we have used previously to slow the release kinetics of proteins and small molecules.^28,29 Its structure contains partial negative charges on a large flat surface, enabling it to deposit parallel to the plane of the scaffold surface and increasing tortuosity (Figure 1A). The simplest approach to incorporating Laponite was to replace the polyanion in our original formulation, dextran sulfate (DS), with Laponite. Thus, we could compare the original formulation incorporating DS with the slow-release formulation incorporating Laponite, denoted as [Poly2/DS]₂₀[LPEI/RNA]₅₀ and [Poly2/Laponite]₂₀[LPEI/RNA]₅₀, respectively. These formulations have two components - a base bilayer and an RNA-containing bilayer; DS and Laponite were only in the base bilayer. In theory, the base bilayer might only release after the RNA layers in a simple surface erosion system and might not affect RNA release kinetics. Thus, a one-component tetralayer architecture containing Laponite was also tested ([LPEI/RNA/Poly2/Laponite]₃₀). This formulation leveraged the same constituents as the two-component Laponite architecture but in a different order of assembly to minimize confounding factors (Figure 1B).

Figure 1. Laponite can be used as a diffusional barrier to slow release of oligonucleotide therapy from dressings.

A) Schematic depicting Laponite nanoclay’s ability to increase tortuosity through its high aspect ratio, thus limiting diffusion that can cause burst release. Schematic adapted from Howard et al., 2022, with permission from Elsevier.²⁹ Structure of Laponite from Tomás et al., 2018, with permission from Elsevier.²⁶ B) Laponite is incorporated into the thin coating architecture using either a two-component bilayer (bi) or single-component teratalayer (tetra) architecture. Original formulation is as in Berger et al., 2023.¹⁴ C) Release of a fluorescently labeled siRNA mimic made of dsDNA (siD7) from Poly2-containing formulations. Release in 1% BSA in 1x PBS. Normalized to the average total loading for that formulation. n=2–3. D) Total loading of siD7 into the Poly2 architectures. n=2–6. One-way ANOVA with Tukey’s multiple comparisons. E) Poly3 is also incorporated in the place of Poly2 to alter release kinetics. F) Total loading of siD7 into the two Poly3 architectures. n=3. Student’s t-test. All shown as mean±SD. **p≤0.01. ns=not significant.

We loaded the coatings with a fluorescently-labeled dsDNA mimic of siRNA (siD7).³¹ We then compared release of the slow two-component bilayer and slow single-component tetralayer formulations containing Laponite nanoclay to our original formulation through release in 1% BSA in PBS. While the original formulation released about 60% of its therapeutic within the first day, the slow two-component bilayer and slow single-component tetralayer formulations released only about 20–30% of their therapeutic over the same timeframe. By 10 days, the original formulation was completely released, while the slow single-component tetralayer formulation still contained about 30% of its loaded siRNA and the slow two-component bilayer formulation still contained about 60% of its loaded cargo (Figure 1C).

Despite confirming slow-release with both Laponite-containing architectures, we noted that the absolute amount of loaded RNA was unequal between the formulations. The one-component tetralayer formulations had lower total loading than the slow two-component bilayer or original formulations (Figure 1D). To minimize any confounding in biological effect from the delivery of different total doses of therapeutic, we decided to proceed forward leveraging Laponite in a slow two-component bilayer formulation rather than one-component tetralayer architecture to better match it to the original formulation.

We endeavored to slow the release from our Laponite-containing formulation even further, providing greater differentiation from the original formulation. In the original formulation, changing the PBAE from Poly2 to a more hydrophobic Poly3 slows release of RNA,²² and we suspected that it could similarly further slow release kinetics from the slow-release formulation (Figure 1E). Prior to testing these Laponite- and Poly3-containing formulations, we confirmed that changing the PBAE from Poly2 to Poly3 did not change the total loaded dose (Figure 1F).

The mechanisms for decreased release rates observed in the two component bilayer systems may be due to one or more mechanisms. As reported in previous work utilizing similar architectures for siRNA release, we found through SEM imaging that release likely occurs through gradual break up and disruption of the film as opposed to surface erosion.¹² By decreasing the rate at which the base underlayer breaks down upon swelling with water, the overall rate of film disruption and release from the top layers is likely to be slowed. Furthermore, there is a well-known effect in LbL thin films called interlayer diffusion, where polyelectrolytes can move freely between layers.^37,38 Typically, Laponite has been shown to reduce interlayer diffusion,²⁸ but it has been demonstrated that during assembly steps, which leverage salt and pH shifts, the Laponite can rearrange within the film,³⁹ which would block diffusion of siRNA during release. In all cases, Laponite helps to exclude water from the PBAE, which further slows its hydrolysis.

3.2. Slow-release Poly3 dressings exhibit bioactivity in vitro and in vivo

Having demonstrated a slower release formulation containing Laponite and Poly3 (Figure 2A), we sought to verify that it could indeed deliver functional anti-miR. We loaded anti-miR-92a (92ai) given our previous proof-of-concept and other studies suggesting its effects when delivered to wounds.^14,40–42 Coated dressings containing Scr were used as a control as opposed to untreated cells or wounds to account for the potential effects of the nucleic acid, dressing, and polymer excipients themselves. We first confirmed that 92ai released from coated wound dressings could inhibit miR-92a expression in human umbilical vein endothelial cells (HUVECs), a representative endothelial cell line (Figure 2B). At 24 hours, we observed decreased levels of miR-92a expression with dressings containing active therapy released to cells (92ai) compared to those containing a scramble sequence non-targeting anti-miR (Scr), although we did not observe de-repression of miR-92a target ITGA5 at this timepoint (Figures 2C). In a full thickness wound model in C57Bl6/J mice (Figure 2D), we confirmed miR-92a inhibition and de-repression of Itga5. We saw that these expression changes were particularly apparent on day 4 (Figures 2E). Through these experiments, we confirmed that our slow-release dressings were able to deliver anti-miRs in a way that retained their function while promoting uptake and endosomal escape.

Figure 2. Slow-release 92ai dressings exhibit bioactivity in vitro and in vivo.

A) A formulation containing Poly3 and Laponite used as the slow-release formulation. B) Commercially available woven nylon wound dressings coated with anti-miR-92a (92ai) were incubated with endothelial cells for 24 hours before assessing gene expression. C) miR-92a and downstream target (ITGA5) expression with 92ai-containing wound dressings compared to those containing a scramble sequence anti-miR (Scr). RT-qPCR with ΔC_t method to SNORD48 for miRNA and ACTB for downstream target de-repression. Expression normalized to that with Scr at each time point. n=6 biological replicates from two independent cell platings. Student’s t-test. D) C57Bl6/J mice received bilateral dorsal full-thickness skin wounds and were immediately dressed with slow-release dressings containing Scr or 92ai. E) miR-92a and downstream target (Itga5) expression. ΔΔC_t method, normalizing to SNORD65 for miRNA and to Gapdh for de-repression. Comparison of wounds receiving 92ai-coated dressings to their contralateral control receiving a Scr-coated dressing. Student’s t-test comparing to a fold change of 1. n=4, ignoring a sample that did not amplify. All shown as mean±SD. *p≤0.05, ***p≤0.001.

3.3. Oligonucleotide therapy structure also impacts release kinetics

We next wanted to ensure that any release kinetics observed in vitro were recapitulated in vivo. This is important since wounds contain protein at levels higher than the 1% used for in vitro release,^43,44 and protein can shield charges in a manner that disassembles LbL thin films. Wounds also contain inflammatory cells and reactive oxygen species, which further complicate delivery.⁸ The wound dressings were loaded with siD7, placed on mouse wounds in C57B6/J mice, and measured at various time points with IVIS to assess the amount of siD7 remaining on the dressing or released to the immediate vicinity in the wound bed (Figure 3A). Comparing the original formulation to the slow formulation, both containing Poly3 (Figure 3B), we observe that the original formulation releases about 80% of its siRNA within 12 hours. In contrast, the slow formulation released only 30% and continued to elute slowly over the next 4 days, albeit with large variability between wounds (Figures 3C and S3A). This result was encouraging as it suggested that Laponite slows the release of siRNA in vivo, as it does in vitro.

Figure 3. In vivo release kinetics of the original and slow-release formulation are dependent on the oligonucleotide.

A) Dressings were applied to dorsal dermal wounds, and the fluorescent oligonucleotide remaining was tracked over time with IVIS imaging. B) Dressing formulations for original and slow bilayer dressings, both containing Poly3. C) Dressings of the original or slow-release architecture were loaded with fluorescent siRNA (siD7), and release tracked over time. n=4 wounds/condition in 4 mice. D) Dressings of the same architectures were assembled with fluorescent anti-miR (anti-miR-AF647). n=4 wounds/condition in 8 mice. All shown as mean±SD.

We repeated this experiment using dressings coated with single stranded anti-miR rather than double stranded siD7. We loaded dressings with a fluorescent version of scramble sequence anti-miR (anti-miR-AF647) and released them into the wound bed in full-thickness mouse wounds over 5 days, as before. We observed that the release of anti-miR from the original and slow formulations nearly overlapped, suggesting that differences in oligonucleotide structure between siRNA and anti-miR also impact release (Figures 3D and S3B). We realized that we would have to re-engineer the original formulation to release even faster, but first we sought to understand why release differed greatly between siRNA and anti-miR.

We suspected that wound proteins may have mediated differences in release kinetics between in vitro and in vivo experiments. Wound effluent is thought to contain about 3–5% protein,^43,44 more than half of which is comprised of albumin. We initially modeled this using a pure albumin solution with 1% BSA in PBS; however, to better recapitulate the protein content of wound effluent, we also released into commercial 100% FBS, which contains 3–5% protein (Figure S4A). In 1% BSA in PBS, the slow formulation with siD7 displayed slower release than the original formulation with siD7, but the original formulation released with the same kinetics irrespective of the therapy being siD7 or anti-miR-AF647 (Figure S4B). In contrast, in 100% FBS solution, the anti-miR formulation released slower than those containing siD7 (Figure S4C). We observe differences in early release between the original and slow-release formulation with anti-miR-AF647, but they release more similarly than those dressings with siD7. This led us to postulate that the ability for single stranded products to bind more strongly to polycations could confer resistance to charge shielding by proteins and thus reduce film disassembly.⁴⁵ While our in vitro results did not completely recapitulate our in vivo findings, it does suggest that our system can be tunable for both siRNA and single stranded anti-miR, but that they interact differently within the coatings and exhibit different release kinetics. All subsequent in vitro release experiments were performed in 100% FBS and leveraged anti-miR-AF647 to tune our system specifically for anti-miR delivery.

3.4. PAA can be used to engineer a fast-release wound dressing formulation for anti-miR delivery

Given that the slow formulation with anti-miR-AF647 indeed demonstrated release over 5 days, which is likely sufficient to release into the proliferative phase of wound healing, we sought to engineer a formulation that would release on a faster time scale to give a bolus of anti-miR quickly upon application. We endeavored to keep the formulation similar to the slow formulation to limit the design space and minimize combinatorial effects from changing multiple excipients. We leveraged knowledge of thin film assembly and were guided by two design principles: 1) lower molecular weight polymers act as a less tortuous network and may speed diffusion from the thin films and 2) carboxylated polymers have a higher pKa than sulfated ones, which enables them to be less charge dense and thus form lowered numbers of electrostatic interactions within the film. To test this hypothesis, we compared our original and slow formulations to those containing poly(acrylic acid) (PAA, ~20kDa) or poly(styrenesulfonic acid) (PSS, ~18 kDa) (Figure 4A), with the replacement of DS or Laponite with PAA or PSS. PAA greatly sped release kinetics in 100% FBS in vitro, releasing 80% of loaded therapy within the first day. In contrast, PSS demonstrated slowed release kinetics similarly to that observed with the Laponite-containing slow formulation, potentially due to both a higher charge density of the polymer at assembly pH and exclusion of water from greater hydrophobicity of its backbone (Figure 4B).

Figure 4. Different anionic excipients modulate release kinetics from two-component bilayer formulations.

A) Poly(acrylic acid) (PAA), ~20 kDa, and Poly(styrene sulfonic acid) (PSS), ~18 kDa, can be used in place of dextran sulfate (DS), ~500 kDa, or Laponite nanoclay (slow) as anionic excipients to modulate release kinetics. All formulations loaded with fluorescent anti-miR (anti-miR-AF647) as the oligonucleotide. B) Release kinetics into 100% FBS. Release normalized to final time point of release after about 2 weeks. Inset shows early time points. n=3. Mean±SD.

3.5. Systematic optimization of PAA-containing dressing coatings enables a fast-release formulation

Based on these initial tests, we continued to optimize a fast-release formulation containing PAA (hereby referred to as the fast formulation). We leveraged multiple factors to control the release rate: 1) fewer layers might speed time to complete surface erosion or provide a less tortuous network, 2) Poly2 releases faster than Poly3, as we have previously shown,²² and 3) reducing the polymer molecular weight may speed release kinetics by reducing tortuosity. In addition to modulating release kinetics, we also wanted to equalize the anti-miR loading between the fast- and slow-release dressings. This was particularly important since we found that total loading and release kinetics of anti-miR were intrinsically linked, suggesting that bolus release may be due to non-specific interactions between anti-miR and polyelectrolyte. Previously, it was determined that RNA deposition in LbL is diffusion-limited; that is, by changing the concentration of RNA in the dipping bath, the amount deposited into the thin film can be modulated.¹⁰ We leveraged these factors (Figure 5A) to further optimize the dressings.

Figure 5. A fast-release formulation containing PAA as the excipient can be optimized to speed release kinetics.

A) Dressings containing anti-miR-AF647 were formulated as [PBAE/PAA]₂₀[LPEI/anti-miR-AF647]_n. The PBAE, number of RNA layers (n), anti-miR concentration in the dipping bath, and PAA molecular weight were optimized. B) Release kinetics of various fast-release architectures compared to those with a slow-release formulation. C) Normalization of release kinetics. For B-C, release into 100% FBS. Inset shows early time points. n=3–4. D) Final formulations for fast- versus slow-release dressings. E) Total loading on the final fast- and slow-release dressing formulations incorporating anti-miR-AF647 from B. Total release measured after 10 days of release in 100% FBS. n=3–4. F) Highlighting the relative release kinetics of the final fast- and slow-release dressing formulations from C. Release in 100% FBS. Inset shows early time points. n=3–4. G) Area under the curve (AUC) of the normalized cumulative release in F. Interpolating to an endpoint of 127 hours to align the two release profiles. n=3–4. All shown as mean±SD. Student’s t test. ***p≤0.001, ns - not significant.

Releasing into 100% FBS, we found that formulation B containing Poly2, 25 RNA layers, 50% of the original concentration of anti-miR used in the dipping bath (0.96 μM rather than 1.92 μM), and a 20 kDa PAA provided the most rapid bolus release kinetics. Using Poly3, more layers, lower anti-miR concentration, and a smaller 2 kDa PAA did not speed release kinetics (Figure 5B). Formulation B released 2.36 μg/cm² of anti-miR within 45 minutes, compared to the slow formulation, which released only 0.35 μg/cm² of anti-miR within the first 30 minutes (Figure 5B). This correlates to 30% and 5% of the total loaded anti-miR, respectively (Figure 5C). After initial differences in bolus release within the first 24 hours, the release kinetics equalized in rate between the formulations. This indicated a biphasic release system, where Laponite slows initial bolus release compared to PAA, but not subsequent release. These optimized dressing formulations (Figure 5D) loaded a similar total amount of anti-miR-AF647 (Figure 5E). Measuring the area under the curve (AUC) to understand the dose-time exposure between the fast- and slow-release formulations (Figure 5F), we observed a significantly reduced AUC in the slow formulation, with the fast formulation increased release kinetics over the slow formulation by 36% (Figure 5G).

3.6. Diabetic wounds exhibit enhanced closure with 92ai over Scr, and the timing of this effect is dependent on release kinetics

Having established a system for locally delivering anti-miR on multiple time scales, we wished to establish whether our final fast- and slow-release dressings maintained bioactivity in vivo. We also wanted to observe if there might be any differences in wound closure and angiogenesis by release kinetics. The final fast formulation had the architecture [Poly2/PAA]₂₀[LPEI/RNA]₅₀ and the final slow formulation had the architecture [Poly3/Laponite]₂₀[LPEI/RNA]₅₀ (Figure 5D). We leveraged a well-established full thickness excisional wound model in genetically diabetic mice to investigate the effects of release kinetics on efficacy of dressings loaded with 92ai vs Scr. Given that we had also previously established sex-dependent differences in 92ai on its role in promoting vascularization, we proceeded with all male mice to disambiguate effects.¹⁴ This decision was also consistent with previous studies that have only used male mice in related research.^{41,42,46–49} Immediately upon wounding, dressings were placed on the wounds such as to cover their edges. Wounds were imaged on days 0, 5, 7, and 9 to assess their macroscopic closure rate. In addition, wound tissue was analyzed from mice on days 5 and 9 with RT-qPCR, flow cytometry, histology, and immunofluorescence staining (Figure 6A).

Figure 6. 92ai delivered from coated wound dressings promotes wound closure.

A) Genetically diabetic mice (db/db) were wounded bilaterally with a full-thickness excisional wound. Dressings containing Scr or 92ai and fast- or slow-release kinetics were immediately applied to the wounds. Healing was tracked in the day 9 endpoint group, with wound tissue collected at 5- and 9-days post-injury. B) Representative gross wound images used for quantification of the wound area remaining over time. Epithelial edge determined on each image to find the open wound area and depicted with a dashed yellow line. Scale bar 5 mm. C) Quantification of the wound area over time, normalized to initial wound size. Inset shows Scr vs 92ai, independent of release kinetics. Blinded to condition on analysis. n=8 wounds/condition/timepoint. Three-way ANOVA with factors being release kinetics, treatment (Scr vs. 92ai), and time. Matching by sample over time and Geisser-Greenhouse sphericity correction. D) Area under the healing curve (AUC). n=8 wounds/condition. Ordinary one-way ANOVA with Tukey correction for pairwise comparisons. E) Linear estimation of healing rate between time points for 92ai dressings, normalized to that for Scr wounds of the same release kinetics. A value greater than 1 represents faster healing in the 92ai wound than the contralateral Scr control wound. n=8 wounds/condition/timepoint. Repeated measures two-way ANOVA with Geisser-Greenhouse sphericity correction and factors as time and release kinetics. Matching by sample over time. Šidák correction for pairwise comparisons within each time point. All shown as mean±SD. *p≤0.05, **p≤0.01, ***p≤0.001, ****p≤0.0001.

Wounds were initially of similar size between the different groups (Figure S5A). Over time, all wounds approached complete closure using a combination of re-epithelialization and contraction. Wound imaging suggested that 92ai-treated wounds achieved greater closure and approached complete healing more closely than Scr-treated wounds (Figure 6B). Quantification of the open wound area by a reviewer blinded to the condition demonstrated significantly quicker healing in wounds treated with 92ai than those treated with Scr; however, no differences were detected between release kinetics (Figure 6C). To better quantify the relationship between open wound area and time, we also calculated an area under the curve (AUC), corroborating our finding that 92ai significantly reduces the AUC compared to Scr, independent of release kinetics (Figure 6D). Healing was independent of the initial wound size, suggesting differences in wound healing rate were not due to the spread in initial wound size between groups (Figure S5B). Healing was about 14% faster with 92ai over Scr when treated with the fast formulation and 12% faster with the slow formulation, comparable to the 17% faster healing we observed in our previous proof-of-concept study where wound dressings were re-applied after 2 days to increase the therapeutic dosage. This suggests that although the effect on wound closure was small, it was biologically meaningful in this model.

We wondered if release kinetics may alter the time scale on which accelerations in wound healing are noted. We quantified this by calculating the rate of healing over time, normalizing wounds receiving 92ai to those receiving Scr with the same release kinetics. We observed a significant effect of both release kinetics and their interaction with time. In the first interval between days 0 and 5, the healing rate appeared faster in the fast-release group than the slow-release group, although the difference was not significant. In the second interval, the slow-release dressings demonstrated significantly higher relative closure than the fast-release dressings. By the last interval of days 7 through 9, the healing rates had normalized and were the same irrespective of release kinetics (Figure 6E). These findings are comparable to our previous study with the original formulation, where the closure rate with 92ai was 1.9–2.2 times higher than Scr over the first 4 days with little acceleration beyond that time point, most consistent with fast-release dressings from this study.¹⁴ Ultimately, we show earlier effects of 92ai therapy on wound closure with fast-release dressings than slow-release dressings, as expected.

3.7. 92ai modulates various aspects of macrophage phenotype and endothelial cell abundance, differing by release kinetics

Having established the effect of 92ai on wound closure and observed how release kinetics alters the temporal aspects of that effect, we performed an analysis of potential molecular and cellular mediators. First, we assessed the cellular components of wound healing using flow cytometry on dissociated wound tissue. After dissociation, there was a consistent level of live cells across treatment groups (Figure S6), enabling flow cytometry, although a few samples had very low yield. We assessed cell types (CD68⁺ macrophages and their M1- or M2-subtypes along with CD31⁺ endothelial cells) and cell phenotypes (Ki67⁺ proliferation, ITGA5 protein expression, and αSMA protein expression) (Figures 7A, S7A). Wound tissue was compared to unwounded, untreated (UW, UT) skin as a baseline.

Figure 7. Release kinetics of 92ai delivery from dressings may alter the mechanism of healing, as assessed by flow cytometry.

A) Upon reaching the experimental endpoint, wound tissue was harvested within a few mm radius of the wound edge, dissociated, stained for various markers, and analyzed with flow cytometry. Gating strategy for flow cytometry in Figures S1 and S2. Expression of key markers for B) CD68⁺ macrophages, their C) M1/M2 phenotype, D) total M2-like macrophages (CD45⁺CD68⁺CD206⁺), and E) CD31⁺ endothelial cells. Dashed line is average of unwounded, untreated (UW, UT) skin from 4 mice at each time point. Two-way ANOVA with Tukey corrected pairwise comparisons. Two factors are treatment and release kinetics. All graphs represents mean±SD. n=8 wounds/treatment/timepoint, excluding samples with inadequate cell count or with outlier detection (ROUT Q=1% threshold). Pairwise comparisons plotted if significant or approaching significance (p≤0.1). *p≤0.05, **p≤0.01.

We first examined macrophages, as they play a central role in regulating wound healing and wound angiogenesis. Macrophage accumulation in wounded skin was higher than that in UW, UT skin, as expected, and it trended towards increased prevalence in wounds treated with 92ai versus Scr at day 5 (Figure S7B). The effects of release kinetics also trended towards a significant increase with slow dressings over fast dressings. By day 9, there was a trend towards difference with the interaction of treatment and release kinetics on macrophage accumulation (Figure 7B).

Macrophages can act in both pro-inflammatory and anti-inflammatory ways depending upon spatial and temporal differences, and they may receive miR-92a signaling from endothelial cells.⁵⁰ While the role of M2-like anti-inflammatory macrophages has long been understood in wound healing,^51,52 recent research profiling cells from chronic wounds suggests that M1-like pro-inflammatory macrophages also play an important role in promoting ulcer healing.^53–56 Thus, we wanted to see if there were underlying differences in the phenotype of wound macrophages. No differences were observed on day 5 (Figure S7C). On day 9, we found a significant interaction between treatment group and release kinetics (Figure 7C). Specifically, there was a trend towards decreased M2/M1 ratio with 92ai versus Scr when tested in slow-release dressings but not in fast-release dressings. This led to a difference in the M2/M1 ratio by release kinetics, where fast 92ai dressings exhibited a significantly higher ratio compared to slow 92ai dressings (Figure 7C). This was mediated mostly by differences in the M2 macrophage population, not the M1 macrophage population (Figure S8), as we observed a significant interaction between release kinetics and treatment for the prior. Fast-release dressings with 92ai increased M2 macrophages over Scr, although not in a statistically significant manner. In contrast, slow-release dressings with 92ai significantly decreased M2 macrophage levels compared to Scr; however, fast-release Scr dressings exhibited significantly fewer M2 macrophages on day 9 than slow-release Scr dressings, suggesting a potential role for the materials contained in the dressing themselves to affect macrophage phenotype (Figure 7D). Overall, these results suggest that fast- and slow-release dressings may have had different effects on macrophage phenotype.

We next probed the level of endothelial cells, which express CD31, as a proxy for vascularization. In the wounds treated with Scr, very low levels of endothelial cells were observed, but at higher levels than UW, UT skin. On day 5, there was a trend towards an increased endothelial cell population, expanding the CD31⁺ population 3-fold from an average of 0.18% with Scr to 0.58% with 92ai. This effect was not mediated by release kinetics, as an increase, while not significant, was observed in both fast and slow formulations. However, while there appeared to be some samples that received fast 92ai that raised the average, the whole population shifted towards increased CD31⁺ cells for the wounds treated with slow 92ai dressings, suggesting that the effect may be more robust in slow- than fast-release dressings. By day 9, there were no significant effects of 92ai on the endothelial cell population fraction (Figure 7E). Overall, this suggests that at least part of the increased wound closure with 92ai may be due to increased angiogenesis, with more uniform effects in the slow- versus fast-release dressings.

Consistent with the literature and expected findings,⁵⁷ Ki67 (a marker of cellular proliferation) increased in wounds over UW, UT skin, but no differences were observed with 92ai versus Scr or fast- versus slow-release (Figure S7D). We also assessed expression of ITGA5, which plays key roles in mechanotransduction, cell adhesion, and cell migration.⁵⁸ miR-92a deficiency is thought to de-repress ITGA5 in endothelial cells to promote a more migratory, pro-angiogenic phenotype.⁴⁹ In addition to its role in endothelial cells, ITGA5 expression in macrophages alters phenotype and promotes angiogenesis within the context of myocardial infarctions.⁵⁹ On both days, levels of ITGA5 were increased over those in UW, UT skin, with trending towards de-repression in the 92ai group versus the Scr groups at day 9, and with more effects observed in the fast rather than the slow dressings (Figure S7E). Finally, assessment of αSMA expression, which is a key marker for contraction-producing myofibroblasts and is present in the smooth muscle of the vasculature,^60,61 revealed significantly decreased αSMA expression on day 5 in wounds receiving 92ai versus Scr dressings, independent of the release kinetics (Figure S7F), which could be due to a temporal shift in the inflammatory and proliferative phases. Together, our flow cytometry results suggest potential differences in macrophage phenotypes between fast- and slow-release dressings, which might mediate healing.

3.8. Gene-level expression in wound tissue suggests differences in angiogenesis and ECM by release kinetics

We also probed gene expression at endpoints of day 5 or 9 (Figure 8A). Gene expression was normalized to the BestKeeper index for enhanced robustness to perturbations (Figure S9).^35,62 On day 5, we observed a significant decrease in the expression of Pten with slow-release dressings (Figure 8B), which could suggest a pro-angiogenic environment through AKT signaling. On day 9, Hif1a was downregulated in wounds receiving 92ai fast formulations. This downregulation was similarly seen in our previous study for male mouse wounds that displayed increased angiogenesis.¹⁴ It could represent that increased blood vessels have now reduced hypoxia in these wounds. Mmp9 expression was also significantly upregulated for fast 92ai dressings over their Scr controls. Typically, matrix metalloproteinase 9 (MMP9) is thought to inhibit wound healing through overactive extracellular matrix remodeling and to be upregulated in non-healing chronic diabetic ulcers.⁶³ By knocking out or inhibiting MMP9, mixed phenotypes have been seen, with some studies reporting improved healing and others reporting impaired healing,^64,65 including studies from our own group on release of an siRNA against MMP9.¹² MMP9 is also upregulated in hypoxia to induce angiogenesis and releases growth factors from the ECM.^63,66–68 Il10 trended to increased expression in slow-release 92ai dressings. IL10 is thought to be anti-inflammatory, which is contrary to the decreased M2-like macrophages we observed on flow at this time-point, although it could represent the wound’s attempt to transition to a more anti-inflammatory phenotype in this context or could be a response to the overall lower levels of macrophages for slow release 92ai dressings. Finally, we also observe increased Vegf expression with 92ai delivered from slow-release dressings (Figure 8B). Higher baseline VEGF expression has been associated with improved healing of diabetic ulcers.^54,69–71

Figure 8. Release kinetics of 92ai delivery from dressings may alter the mechanism of healing, as assessed by PCR and tissue section immunofluorescence.

A) Wound tissue is collected and homogenized, RNA extracted, and RT-qPCR performed. B) Differentially expressed wound healing gene expression in wounds treated with Scr or 92ai-containing dressings and fast- or slow-release kinetics. RT-qPCR with ΔΔC_t method normalized to Bestkeeper of Gapdh, Hprt1, Ywhaz, and Rps18. Expression within 92ai wounds normalized further to that in the contralateral wound receiving Scr. Non-differentially expressed genes shown in Figure S10. Student’s t-test comparing to a fold change of 1. Mean±SD. C) Representative CD31 immunofluorescence of tissue sections. Vascularization (Vasc) scoring on CD31-stained samples. Scored by reviewer blinded to treatment condition using an ordinal scale of vascularization: 0 – none, 1 – mild, 2 – moderate, 3 – marked. Median. Analysis with ordinal cumulative link model of the effects of treatment group, release kinetics, and their interaction on vascularization score. Model assessed with type III ANODE with Wald Chi-square test. Data represents. n=8 wounds/treatment/timepoint, excluding samples that did not section or amplify or with outlier detection (ROUT Q=1% threshold). Pairwise comparisons plotted if significant or approaching significance (p≤0.1). *p≤0.05, ***p≤0.001.

No differences in expression were observed for miR-92a’s downstream target (Itga5) or for several general wound healing genes related to the extracellular matrix (ECM) and fibrosis, immune regulation, vascularization, and mesenchymal mitogens (Figure S10A). A small but significant inhibition of miR-92a was observed with fast-release dressings at day 9, but the same was not true for slow dressings or at day 5 (Figure S10B).

Taken together, these results suggest some differences in gene expression between fast- and slow-release dressings, potentially underlying the macrophage phenotypes and endothelial cell differences that we observed with flow cytometry. Together, it could suggest earlier progression through angiogenesis for fast-release dressings, especially given reduced Hif1a, and leading to increased Mmp9 as the ECM remodels. Slow-release dressings may still impact earlier stages of angiogenesis, driven by reduced Pten at first and then increased Vegf.

3.9. 92ai alters vascularization in a release-kinetics dependent manner

Finally, we evaluated wound healing on a histological level. We observed no difference in average granulation tissue thickness by treatment group or release kinetics (Figure S11A–B). A granulation tissue score adapted from the literature was also used to assess granulation tissue in both the dermal and subcutaneous levels, where a higher score indicates more marked granulation tissue formation.⁴¹ Few effects were observed in the average score between the levels, although there was a trend towards an effect of release kinetics at day 9 (Figure S11C). We similarly did not see changes with 92ai versus Scr in our previous study,¹⁴ even though we did use older mice in the study. We note that the inconsistency of sectioning the wound bed may contribute to the inability to detect differences in three-dimensional tissue.

Noting this limitation of histology from a single section, we additionally sought to characterize new vascular networks forming in the tissue. We first quantified the number of CD31⁺ vessels per unit area of tissue, but we did not see differences between the treatment groups and release kinetics (Figure S12). We also scored samples for their vascularization since quantification is limited to only 2–3 areas of the tissue per sample. Scoring, with higher scores indicating more vascularization, was performed by a reviewer blinded to conditions and provided a gestalt of vascularization across the whole tissue. We did not observe differences in vascularization score on day 5, as was observed in flow cytometry of these samples. This could be because CD31⁺ cells themselves start as more disorganized, small vessels or vessel fragments that eventually organize into mature vessels. On day 9, there was a significant effect of both the treatment and interaction between treatment and release kinetics on vascularization score. Pairwise comparisons revealed trending towards increased vascularization score with 92ai slow dressings over 92ai fast dressings (Figure 8C). These results were mostly recapitulated in the scoring of H&E sections, although there appeared to be an interaction between treatment and release kinetics at the earlier time point (Figure S13). There appeared to be no differences in αSMA by treatment group or release kinetics (Figure S14). Immunofluorescent measurement of Ki67⁺ did show an effect of release kinetics on day 5, but otherwise there were no effects of treatment, release kinetics, or their interaction (Figure S15). Together with flow cytometry data, this suggests that 92ai released from slow-release dressings may be more effective at modulating angiogenesis.

3.10. Limitations

Our initial release kinetics testing was limited by assumptions around what defines complete release for normalization and how well in vitro matrices model wound effluent. Some of the formulations exhibited spatial heterogeneity in loading, which may have further compounded release experiments and ultimately bioactivity. We coated one example commercial wound product, but the properties of the underlying dressing can affect LbL adsorption and thus release kinetics; further optimization would be needed to control release from coatings on other wound dressing materials. As in our previous work,¹⁴ the db/db wound healing model has significant challenges in recapitulating the re-epithelialization process observed in human diabetic ulcers. A high degree of wound contraction leads to healing of even untreated wounds. Splinting may force wounds to leverage re-epithelialization over contraction.^72,73 Ultimately, a large animal model, such as a porcine skin model, may better recapitulate diabetic wound healing, as others have used.^74,75 Although we find some evidence for changes in macrophage phenotype and vascularization within the wounds, further molecular and cellular analyses are needed to better define the phenotypes. Probing at earlier time points could provide more molecular insight, based on other studies and our preliminary data.⁴⁰ Additionally, techniques such as bulk RNA sequencing may help find putative targets driving changes observed at a transcriptional level beyond those tested through RT-qPCR, and single cell sequencing could help better understand the phenotype of endothelial cells and macrophages in the wound bed. Finally, we study the delivery of only one representative anti-miR, but targeting different miRNAs implicated in impaired wound healing could be possible with our platform approach.

4. Conclusions

Chronic wounds remain a critical challenge for patients with diabetes, and current therapies are inadequate. We have previously developed and demonstrated coated dressings that release anti-miR, but we have only examined effects on a single time scale. Here, we leveraged different polyanions and Laponite nanoclay to systematically engineer the release kinetics of 92ai from LbL coatings on commercial woven wound dressings. We demonstrated how several aspects of the formulation, including assembly conditions, contribute to anti-miR loading and release kinetics. We then used these to understand how the release kinetics of 92ai from wound dressings into the wound affects efficacy. Together, our data suggests that 92ai accelerates wound healing in different temporal manners dependent on release kinetics, but that the overall cumulative effect on wound closure is similar. In exploratory mechanistic studies, we show that there may also be differences in how angiogenesis and inflammation contribute to these effects dependent upon release kinetics. Angiogenesis may play more of a role with slow-release dressings, while anti-inflammatory macrophages may play more of a role with fast-release dressings, or fast-release dressings may progress through the wound healing phases faster than slow-release dressings. Further work is necessary to define the mechanisms underlying 92ai delivered to wounds on different time scales. In sum, this study provides a framework for modulating the release kinetics of anti-miR from coated wound dressing formulations and suggests the ways in which temporal control of 92ai delivery may impact efficacy. As we envision an industrially-manufactured, off-the-shelf bandage strategy, understanding both the formulation and biological science behind our dressings will be critical to furthering their clinical translation.

Highlights.

• Layer-by-Layer assembly is used to coat oligonucleotides onto wound dressings.

• We engineer the excipients in these coatings to alter release kinetics.

• Laponite nanoclay incorporation slows oligonucleotide release from the dressings.

• Slow- and fast-release dressings with anti-miR-92a speed ulcer healing in mice.

• The acceleration of closure and mechanisms of healing differ by release kinetics.