Use of Therapeutic RNAs to Accelerate Wound Healing in Diabetic Rabbit Wounds

Abstract

Introduction:

Diabetes mellitus (DM) affects over 422 million people globally. Patients with DM are subject to a myriad of complications, of which diabetic foot ulcers (DFUs) are the most common with ∼25% chance of developing these wounds throughout their lifetime.

Innovation:

Currently there are no therapeutic RNAs approved for use in DFUs. Use of dressings containing novel layer-by-layer (LbL)-formulated therapeutic RNAs that inhibit PHD2 and miR-210 can significantly improve diabetic wound healing. These dressings provide sustained release of therapeutic RNAs to the wounds locally without systemic side effects.

Clinical Problem Addressed:

Diabetic foot wounds are difficult to heal and often result in significant patient morbidity and mortality.

Materials and Methods:

We used the diabetic neuroischemic rabbit model of impaired wound healing. Diabetes was induced in the rabbits with alloxan, and neuroischemia was induced by ligating the central neurovascular bundle of each ear. Four 6-mm full-thickness wounds were created on each ear. A LbL technique was used to conformally coat the wound dressings with chemically modified RNAs, including an antisense oligonucleotide (antimiR) targeting microRNA-210 (miR-210), an short synthetic hairpin RNA (sshRNA) targeting PHD2, or both.

Results:

Wound healing was improved by the antimiR-210 but not the PHD2-sshRNA. Specific knockdown of miR-210 in tissue as measured by RT-qPCR was ∼8 Ct greater than nonspecific controls, and this apparent level of knockdown (>99%) suggests that delivery to the tissue is highly efficient at the administered dose.

Discussion:

Healing of ischemic/neuropathic wounds in diabetic rabbits was accelerated upon inhibition of miR-210 by LbL delivery to the wound bed. miR-210 inhibition was achieved using a chemically modified antisense RNA.

Aristidis Veves, MD, DSc

Brian H. Johnston, PhD

INTRODUCTION

Diabetes affects ∼30 million adults in the United States, with these numbers expected to double by the year 2050. Approximately one quarter of these patients develop at least one foot ulcer in their lifetime, increasing their risk of lower limb amputations.¹ The recurrence rate of diabetic ulcers is over 60%, and with each subsequent ulceration, the amputation rate is 12%. Thus, nonhealing diabetic wounds are a tremendous challenge for patients and caregivers alike. Despite standard of care treatment, diabetic foot ulcers (DFUs) continue to have a low healing rate.^2,3 Therefore, new treatments are needed to improve outcomes in the future for this demographic of patients.

Some of the dysfunction in wound healing has been linked to a diminished cellular hypoxia response, of which the master regulator is the transcription factor hypoxia-inducible factor-1 alpha (HIF-1α). HIF-1α regulates cell homeostasis as oxygen concentrations change to promote wound healing.⁴ HIF-1α levels are negatively regulated by prolyl hydroxylase domain-containing protein 2 (PHD2), a crucial cellular oxygen sensor that, in normoxia, hydroxylates two specific proline residues in HIF-1α, tagging it for rapid degradation using the proteasome pathway.^5–8 However, in the hypoxic conditions of normal acute wounds, hydroxylation and degradation of HIF-1α are reduced, resulting in increased translocation of HIF-1α to the nucleus where it transactivates factors involved in vasculogenesis, angiogenesis, reepithelialization, and cell survival that promote wound healing.^5–7,9

These factors include vascular endothelial growth factor, erythropoietin, platelet-derived growth factor, fibroblast growth factor, transforming growth factor-beta, stromal cell-derived factor 1, and heat shock protein 1. In patients with diabetes, induction of these factors is hampered due to high-glucose induced modification of p300, a cofactor of HIF-1α transactivation.¹⁰ Stabilization of HIF-1α may be achieved through inhibition of PHD2 expression,^5–7,11,12 and such intervention can improve wound closure in diabetic mice as we and others have shown.^{5–7,11–13}

Hypoxia also induces the expression of certain microRNAs (miRNAs), called hypoxamirs, that are involved in wound healing.^14,15 One such miRNA, miR-210, is upregulated by HIF-1α^16–18 and targets E2F3, a cell cycle regulator that facilitates the G1/S transition.¹⁹ This activity is thought to be responsible for the finding that miR-210 attenuates keratinocyte proliferation and reepithelialization in ischemic wound healing.²⁰ Hypoxia-induced upregulation of miR-210 also represses mitochondrial respiration and associated downstream functions.²¹ These characteristics lower the need for oxygen in an ischemic environment, promoting cell survival, but are detrimental to wound healing as tissue repair requires oxidative metabolism and cell proliferation.²⁰

In view of the ability of PHD2 inhibition to stabilize HIF-1α and of miR-210 to attenuate reepithelialization, we have explored the hypothesis that inhibiting PHD2 and miR-210, either alone or in combination, can promote wound healing in patients with diabetes.¹³ To inhibit PHD2, we used short synthetic hairpin RNAs (sshRNAs),^22–24 which are potent effectors of RNA interference and have potential advantages over siRNA. They are synthesized as a single chemical entity which simplifies their production. They have reduced off target activity since the passenger strand lacks the terminal 5′-phosphate necessary to enter the RNA-induced silencing complex as a guide sequence. And, unlike shRNAs expressed from a gene therapy vector, they are delivered as chemically-modified synthetic molecules whose dosage and half-life can be more precisely controlled.

To inhibit miR-210, a chemically-modified antisense oligonucleotide (antimiR) targeting miR-210 was used. Previously, we have shown that this approach was effective in accelerating wound closure in a db/db mouse model¹³ when the oligonucleotides were formulated and delivered using layer-by-layer (LbL) technology.^25–27 This strategy allows the conformal coating of these oligonucleotides onto the surface of commercial wound dressings to enable their localized delivery to the wound bed in a facile, clinically translatable method. Importantly, polymers within the coating that enhance transfection into cells are co-adsorbed with the oligonucleotides. Systematic characterization of this formulation was recently performed. Two days after placing dressings containing antimiR, we observed interactions predominately with endothelial cells, but also to a lesser extent with interfollicular keratinocytes. In a proof-of-concept study in db/db mice, these dressings inhibited an endothelial-predominant antiangiogenic miRNA to promote increased wound closure and sex-dependent increases in vascularization.²⁸

Because there is no perfect animal model of diabetes, it is prudent—and expected by regulatory authorities—that a new therapeutic candidate should show efficacy and safety in two different animal models of diabetic wound healing before advancing to clinical trials. Among the recommended models are the db/db mouse that we previously used, as well as the diabetic, ischemic rabbit ear model. The latter enables two characteristic features of human diabetic wounds, hypoxia and neuropathy, that are not present in the db/db mouse model. In this study, we report on the efficacy of LbL-formulated therapeutic RNAs in this rabbit model of diabetic wound healing with neuroischemia.

CLINICAL PROBLEM ADDRESSED

DFUs remain a substantial problem for patients and health care providers. Treatment advances will significantly help patient quality of life, as well as improve outcomes. Enhanced wound healing will decrease the likelihood that the patient will suffer from infection or amputation.

METHODS

Preparation of sshRNAs and antimiR oligonucleotides

sshRNAs and antimiRs were chemically synthesized and HPLC purified by Integrated DNA Technologies (Coralville, Iowa). The sequences of PHD2 sshRNA SG302 m1 and antimiR-210 SG608 are as follows [m = 2′-O-methyl modification; ZEN = N,N-diethyl-4-(4-nitronaphthalen-1-ylazo)-phenylamine].

SG302 m1: 5′-UAUACCUCCACUUACCUUGmUmUCmAAmGGmUAAGUGGAmGG mUAmUATT-3′

SG608: 5′-mU/ZEN/mCmAmGmCmCmGmCmUmGmUmCmAmCmAmCmGmCmAmCmA/3ZEN-3′

sshRNAs were annealed in aqueous solution to convert any dimeric species to monomeric hairpin conformations and analyzed by denaturing PAGE as described in Dallas et al.²⁹ AntimiR oligonucleotides were solubilized in RNase-free H₂O.

Cell culture

The human kidney cell line 293FT and human keratinocyte cell line HaCaT were cultured in Dulbecco's Modified Eagle's Medium (Cambrex) with 10% fetal bovine serum (Hyclone), 2 mM L-glutamine, and 1 mM sodium pyruvate. Human fetal lung fibroblast line MRC-5 was cultured in Minimal Essential Medium and 10% fetal calf serum.

RT-qPCR of mRNA and miRNA

cDNA for mRNA analysis was synthesized from 10 μL of total RNA samples using the High-Capacity cDNA Kit (Applied Biosystems). cDNA for miRNA quantification was reverse transcribed using the TaqMan-MicroRNA Reverse Transcription Kit (Applied Biosystems). RT-qPCR amplification was performed using 5X HOT FIREPol Probe qPCR Mix Plus (ROX) (Solis BioDyne), and the appropriate TaqMan probes (Applied Biosystems): human PHD2 (Hs00254392_m1), human GAPDH (Hs99999905_m1), miR-210 (00512).

In vitro PHD2 knockdown luciferase reporter assay

A dual luciferase reporter was constructed by subcloning the rabbit PHD2 mRNA coding sequences into the psiCHECK™-2 plasmid (Promega) downstream of the Renilla luciferase reporter. Transfections using the transfection agent Lipofectamine 2000 (Invitrogen) and dual luciferase assays and analysis were performed as described.¹³ Percent silencing was calculated relative to cells transfected with the reporter plasmid in the absence of any sshRNA. Dose curves were plotted, and IC₅₀ values were computed using GraphPad Prism.

In vitro PHD2 knockdown RT-qPCR assay

Triplicate transfections of PHD2 sshRNAs at various concentrations along with nonspecific control sshRNAs into HaCaT cells were performed, and total RNA was extracted as described by Dallas et al.¹³ mRNA levels were quantified using the ΔΔCt method,³⁰ normalizing to GAPDH. Dose curves were plotted, and IC₅₀ values were computed using GraphPad Prism.

miR-210 dual luciferase biosensor assay

A miR-210 biosensor plasmid (pSG247) described previously¹³ was used to confirm the activity of the antimiR targeting miR-210 (SG608) as described previously.¹³ pSG247 contains an rLuc gene having target sites for miR-210; reduction of miR-210 levels due to antimiR activity results in increased levels of rLuc through its derepression.

Formulation solution preparation

Sodium acetate buffer at pH 5.2 and 6 was diluted to 100 and 10 mM using nuclease-free water. Dextran sulfate (DS, VWR, 500 kDa) was dissolved at 1 mg/mL in 100 mM pH 5.2 sodium acetate buffer. Linear polyethylene imine (LPEI, Polysciences, Inc., 25 kDa transfection grade) was dissolved at 0.25 mg/mL in 100 mM pH 6 sodium acetate buffer. Poly2 was synthesized in-house following the literature,³¹ and a solution was made daily at 1 mg/mL in 100 mM pH 5.2 sodium acetate buffer. All polymer solutions were filtered through a 0.2 μm filter before use.

Formulation of LbL dressings

The LbL technique was used to conformally coat the wound dressings.^25,27,31 Dressings had the formulation [Poly2/DS]₂₀[LPEI/RNA]₅₀ for monotherapy and [Poly2/DS]₂₀[LPEI/RNA]₁₀₀ for combination therapy. This enabled matching the loadings of each RNA in the combination therapy films to that in the monotherapy.

More details on the formulation methods are provided in previous work.^26,28 Briefly, Tegaderm^® woven mesh (3M) was cleaned with ethanol and treated in a Harrick plasma cleaner (Ithaca, NY) on high for 10 min under an oxygen environment. After plasma treatment, Tegaderm pieces were placed immediately into the first solution containing Poly2 and allowed to incubate at room temperature for 1 h. Base layers were then assembled at pH 5.2, and RNA layers were assembled at pH 6. Each bath of RNA solution was a 29 μg/mL (2.24 μM) solution of sshRNA and a 22 μg/mL (2.82 μM) solution of antimiR. For combination therapies, the total RNA concentration stayed constant, so the bath contained a mixture of 34 μL of each RNA. Every 25 layers, the coated dressings were allowed to dry for 10 min, and the solution tanks were also refilled to account for evaporation and deposition.

Once formulation was completed, dressings were allowed to air dry for at least 1 h and then stored at −80°C until use, at which time they were fashioned into 7 × 7 mm squares using a scalpel blade.

Animal model: Diabetes induction and neuroischemia wounding surgery

We used a total of eight New Zealand white rabbits that were 4 months old weighing ∼3 kg each. Diabetes was induced in the rabbits using an alloxan based method.³² For the creation of the neuroischemia wounds, we utilized a previously described model of neuroischemia in rabbits³³ in which the central and caudal neurovascular bundles in the ears are ligated. After neuroischemia was induced, four 6 mm full-thickness wounds were created on each ear using a biopsy punch. Wounds were treated with Tegaderm mesh squares formulated as described above and then covered with an uncoated occlusive Tegaderm dressing.

In the first experiment, 5 rabbits were used for a total of 40 wounds. Wounds were treated with mesh squares that had been LbL formulated with the following RNAs: PHD2 sshRNAs SG328 (n = 6) or SG317 (n = 10), miR-210 antimiR SG608 (n = 8), or nonspecific control sshRNA SG221c (n = 8), as well as uncoated mesh squares lacking RNA (n = 8).

A second follow-up experiment was then designed to further test the treatment group that performed best in the first experiment. Three additional rabbits were used for the second experiment using the same diabetic and neuroischemia model with four wounds per ear. Twelve wounds were treated with the best-performing treatment and 12 wounds with uncoated control mesh for a total of 24 wounds. Each animal had control treatments on one ear and active treatments on the other ear to control for inherent differences in healing among animals while avoiding possible confounding effects of RNA migrating between adjacent wounds. The study was approved by the Beth Israel Deaconess Medical Center IACUC.

Wound measurements and imaging

Wound measurements were taken immediately after surgery using electronic calipers to determine the vertical and horizontal diameters. The exterior occlusive dressings were changed every 2 days with subsequent measurements taking place between dressing changes. The wound base was not disturbed nor was the wound cleansed during dressing changes. Wound areas were calculated based on these measurements. Wound healing curves were created using areas normalized to day-zero dimensions.

Sacrifice and histology

Twenty-one days after wounding, animals were euthanized with FatalPlus. After cessation of heart sounds and breathing, the wounds were harvested. Histology was performed with staining for hematoxylin and eosin stain (H&E), Trichrome, and immunofluorescence. Analysis was performed in a blinded manner by a pathologist who reviewed images for inflammatory cells, spindle cells, and vessels.

Quantification of miR-210 levels in rabbit wounds

Wounds of one rabbit were treated with SG317 (PHD2-Targeting sshRNA), control mesh dressing, Tegaderm, SG608 (antimiR-210), or SG221 (nonspecific control) for 4 days before sacrifice. Tissue sections were immediately immersed in RNAlater solution (Invitrogen), incubated overnight at 4°C, and then RNAlater was removed and sections were stored at −80°C until testing. A measure of 1.2 g of lysing Matrix “D” beads (MP Biomedicals) and 1 mL of QIAZOL (Qiagen) were added, and the samples were processed with a FastPrep-24 homogenizer (MP Biomedicals, Solon, OH) by four 60-second cycles (6 m s⁻¹) followed by cooling on ice. Cellular debris was removed by centrifugation at 12,000 g for 2 min. The homogenate was extracted with chloroform, and the aqueous layer was used to isolate total RNA using an RNeasy Plus Universal Mini Kit (Qiagen). RT-qPCR analysis of miRNA was performed as described above. miR-210 levels were quantified using a standard curve with a synthetic miR-210 RNA oligonucleotide with a 5′-phosphate modification.

Statistical analyses

Statistical analyses were performed using the Minitab Statistical Package Version 21.3 (Minitab, Inc., State College, PA). As there were eight wounds per animal and each wound cannot be considered as independent, we used the mixed effects model to perform repeated measurement analysis and compare the efficacy of each treatment. Comparisons of differences among the various treatment time points were performed using either one-way analysis of variance, when more than two treatments were compared, or a Student's t-test to evaluate differences in histological analysis between wounds treated with miR-210 antimiR and Tegaderm alone. A p-value of <0.05 was considered statistically significant.

RESULTS

Identification of potent sshRNA inhibitors of rabbit PHD2 in vitro

The initial lead sshRNA SG302¹³ that silences human PHD2 mRNA with high potency targets the corresponding target site of rabbit PHD2 mRNA with a single mismatch. We therefore designed a sshRNA sequence (SG317) that perfectly matches this analogous site on the rabbit PHD2 mRNA with our standard sshRNA architecture (19-bp blunt-ended stem with 2-nucleotide loop). In addition to this target site, we aligned the human and rabbit PHD2 sequences and selected 25 sites that were conserved between the two species and that in silico design algorithms predicted would be active. These were then screened for efficacy in cell culture. For this initial in vitro screen, the sshRNAs were synthesized without chemical modifications and tested for activity in both rabbit and human sequence contexts. Before testing, all sshRNAs were annealed to convert any multimeric structures to hairpin monomers by heat denaturing and snap-cooling.

We first screened for activity against rabbit PHD2 using a dual luciferase reporter assay. sshRNAs were cotransfected with the reporter plasmid into 293FT cells at 8 concentrations spanning the range 0.003 nM to 10 nM. Forty-eight hours after transfection, the cells were lysed, and both Renilla and firefly luciferase activity levels were measured and IC₅₀ values calculated. A nonspecific sshRNA transfection (SG221c) was included as a control. For the most potent sshRNAs from the first pass screen, activities were confirmed for human PHD2 in a human keratinocyte line, HaCaT cells. Dose–response curves for these sshRNAs were generated by transfection into HaCaT cells, and endogenous PHD2 mRNA levels were measured using TaqMan qPCR. From this second screen we identified, in addition to SG317, another highly potent sshRNA lead (SG328; Table 1) to carry forward for chemical modification pattern optimization.

Table 1.

Lead unmodified short synthetic hairpin RNAs have similarly high potency in both rabbit and human sequence contexts

PHD2 sshRNA	IC50, pM Rabbit PHD2 Target	IC50, pM Human PHD2 Target
Site 1 (SG317/SG302)	9.513	7.848
Site 2 (SG328)	10.93	9.905

SG317 and SG302 target analogous sites on rabbit and human PHD2 and differ by a single nucleotide, and SG328 targets a site that is conserved in the two species.

sshRNA, short synthetic hairpin RNA.

Chemical modification and optimization of rabbit PHD2-sshRNA

The most effective PHD2-sshRNA sequences (Table 1) were modified with 2′-OMe groups in two nucleotides comprising the loop and at alternating nucleotides in the sense strand except at the slicer site as described in our previous studies^23,29 to make them more suitable for in vivo studies. In addition, we screened for potency sshRNAs having larger numbers of chemical modifications by incorporating 2′-OMe modifications in parts of the guide arm in over 20 different patterns (Fig. 1 shows these patterns for SG328). We found highly active sshRNAs in which 2′-OMe modifications were positioned toward the 3′-end of the guide strand opposite an unmodified (2′-OH) residue in the passenger strand. Dose–response curves and IC50s for SG328-derived sshRNAs, as well as unmodified SG328, are shown in Fig. 2.

Figure 1.

Modification patterns for SG328.m1-SG328.m22 sshRNAs targeting human and rabbit PHD2 mRNA synthesized and tested in vitro cell culture. Gray (2′-OMe), white (2′OH), and red-outlined (guide strand sequence). sshRNA, short synthetic hairpin RNA.

Figure 2.

Dose–response curves and IC50 values for various chemical modification patterns screened for initial lead SG328. (A) SG328.m5–m8; (B) SG328.m9–m12; (C) SG328.m13–14; (D) SG328.m15–m17; (E) SG328.m18–m22; and (F) D SG328.m21.

From this screen, SG328.m20 was chosen as having high overall potency and enhanced stability due to a higher degree of chemical modification. We found that the same modification pattern was suitable for the SG317 site as well (SG317.m5; data not shown). Since the increased level of modification showed no reduction in potency, we carried both leads forward for the in vivo rabbit studies.

Activity of a potent antisense inhibitor (antimiR) of miRNA-210

We previously identified a potent antisense inhibitor targeting miR-210 (SG608) that improved healing in a db/db mouse model.¹³ miR-210 is conserved among mammals, so there was no need for species-specific versions of SG608 to be made for the rabbit model or for human use. SG608 contains N,N-diethyl-4-(4-nitronaphthalen-1-ylazo)-phenylamine (“ZEN”) moieties at both the 5′- and 3′-ends, and all residues are 2′-OMe-modified. Activity of SG608 was confirmed using a luciferase reporter biosensor (pSG247) assay.¹³ In this assay, rLuc is derepressed to an extent that correlates with antimiR efficacy. SG608-transfected cells exhibited derepression of rLuc expression by up to threefold compared with cells lacking antimir, consistent with our previous studies¹³ (Fig. 3).

Figure 3.

Dual luciferase biosensor assay to measure the inhibition of miR-210 by antimiRs. Derepression of rLuc signal shows specific activity of miR-210-targeting antimiR SG608 relative to nonspecific control antimiR. pSG247, biosensor encoding plasmid alone (no-inhibitor normalization control); NSC, nonspecific control RNA.

LbL-formulated antimiR-210 (SG608) accelerates diabetic wound healing in neuroischemic rabbits

As the first step of the analysis, we evaluated the efficacy of various modified RNAs or combinations thereof in wounds that were created on diabetic, neuroischemic rabbit ears as described above. Wounds were treated with Tegaderm mesh squares that had been LbL formulated with the following RNAs: PHD2 sshRNA (SG328 or SG317), miR-210 antimiR (SG608), or negative control sshRNA (SG221c), as well as uncoated mesh lacking RNA. Of these treatments, SG608 alone showed the greatest acceleration in the rate of wound healing (Fig. 4A), although the differences did not reach statistical significance presumably due to the underpowered sample size.

Figure 4.

Healing curves for wounds in the neuroischemic ear of diabetic rabbits treated with therapeutic RNAs and controls. (A) Treatment with miR-210-antimiR SG608 provided the fastest healing, especially between Days 4 and 10 postwounding, and was selected for further investigation. SG317 = SG317.m5; SG328 = SG328.m20. p < 0.001 (Mixed Effects Model), n = 8 for each treatment. (B) Direct comparison of SG608 treatment with substrate alone, showing that this miR-210-antimiR improved wound healing, p = 0.038 (Mixed Effects Model), n = 12 for each treatment.

In a second experiment, we focused on this treatment alone with larger numbers and proper statistical power. Three rabbits (24 wounds) were treated with SG608-formulated mesh or mesh alone (n = 12 wounds for each treatment). This time the mesh squares were replaced with fresh ones having the same formulation on day three (D3) postsurgery on the assumption that this would increase the total amount of therapeutic RNA delivered to tissue (Dressing replacement is also a better mimic of the expected use in humans, where regular dressing changes are standard practice.). This experiment showed that SG608 improved wound healing compared to Tegaderm (p = 0.038) (Fig. 4B). The first differences were observed at day 6, and this trend continued until Day 20 where wound area remaining unclosed was 5% for SG608 versus 18% for the control. Complete data are shown in Supplementary Figs. S1 and S2.

Histological analysis showed no significant adverse effects from LbL antimiR-210 treatment

Histological analysis using H&E and Masson's trichrome stain (MTS) showed no significant difference between wounds treated with SG608 dressings and those treated with unformulated mesh. Measurements showed similar levels of inflammatory cells, blood vessels, and spindle cells.

Quantification of miR-210 RNA levels in tissue using RT-qPCR

miR-210 levels in total RNA isolated from wound tissue (day 4) were measured using RT-qPCR. The levels of miR-210 measured for the negative controls ranged from 22.20 to 22.62 Ct, showing consistent recovery of RNA. A standard curve was generated using a synthetic miR-210 RNA oligonucleotide and used to estimate miR-210 copy numbers per ng of total RNA. Controls/untreated were within the dynamic range (linear portion of the standard curve) (4136 copies/ng). Strikingly, the controls of the SG608 samples (30.45) were ∼8 Ct higher. This corresponds to 256-fold less than the controls. This apparent level of knockdown (>99%) suggests that delivery to the tissue is highly efficient at the administered dose. The SG608-treated samples (at the background of the standard curve) were calculated to contain only 13 copies/ng.

DISCUSSION

In this study, we tested the use of modified RNAs that target the HIF-1α pathway in a rabbit model. This large animal model has been previously found beneficial for studying diabetic wound healing. Ligation of the central and caudal neurovascular bundles reduces blood flow by ∼70%, similar to patients with severe limb ischemia. In addition, disruption of the peripheral nerves is also important to replicate the neuropathy that patients with diabetes suffer from and the impact it has on wound healing. Normally, neuropeptides such as Neuropeptide Y, Substance P, and calcitonin gene related peptide regulate the expression of numerous cytokines related to wound healing. However, with a diminished neuropeptide presence, important cytokines such as interleukin (IL)-6, IL-8, IL-10, and tumor necrosis factor-alpha become dysregulated, further inhibiting cell proliferation, neovascularization, and reepithelialization.³⁴

In particular, we used SG317.m5 and SG328.m20, chemically modified sshRNAs targeting PHD2, and SG608, a modified antimiR-210. While our previous work showed improved wound healing with PHD2-targeting sshRNA in a diabetic (db/db) mouse model,¹³ in the present study targeting of PHD2 produced no significant acceleration of healing. While it is unclear why knocking down PHD2 was helpful in the mouse model but apparently not in the rabbit model, it should be noted that those mouse experiments were performed in animals that lacked the neuroischemia features of the rabbit model. It has been shown in a mouse model³⁵ that, in addition to helping maintain cell homeostasis in response to hypoxia, HIF-1α also works to regulate neuropeptides. In the context of neurovascular bundle ligation, the positive effect seen by PDH2 to activate HIF-1α may not translate into improved wound healing due to its inability to upregulate neuromodulators. However, additional experiments are needed to test this hypothesis.

The effectiveness of inhibiting miR-210 alone suggests that this miRNA plays a significant role in chronic wounds. miR-210 is upregulated by binding of HIF-1α to a hypoxia responsive element in the promoter of miR-210,¹⁶ and hence, its levels are elevated when HIF-1α is stabilized. Multiple studies have shown the importance of miR-210 in wound healing, angiogenesis, and oxidative stress.²

Intradermal injection of lipid nanoparticles bearing antimiR-210 rescued keratinocyte proliferation in ischemic wounds in mice.²⁰ In contrast, delivery of a miR-210 mimic to wounds in diabetic mice was also found to improve healing,³⁶ perhaps due to a positive effect on angiogenesis³⁷ at lower doses and an inverse effect at higher doses.¹⁷ The levels of miR-210 in diabetic wounds appear to reflect a balance between increases due to hypoxia and decreases due to hyperglycemia.³⁶ The literature consensus is that miR-210 generally has a detrimental effect on healing.^17,20,21

The other important concept that this study supports is that LbL can be an effective formulation approach for packaging and localized release of various RNA oligonucleotide therapies. LbL formulations allow for topical application of therapeutic oligonucleotides, providing high local concentrations with release extending over several days and with very low systemic exposure.

CONCLUSION

Healing of ischemic/neuropathic wounds in diabetic rabbits was accelerated upon inhibition of miR-210 by LbL delivery to the wound bed of a chemically modified antisense RNA targeting this microRNA.

TAKE-HOME MESSAGES

• Wound healing was statistically improved by the antimiR-210.

• Use of PHD2-sshRNA or combination therapy did not statistically improve wound healing.

• Neuroischemia in diabetic rabbits can serve as a good model for diabetic wound healing.

• LbL can be an effective formulation approach for packaging and localized release of various RNA oligonucleotide therapies with very low systemic exposure.

ACKNOWLEDGMENTS AND FUNDING SOURCES

This study was supported by NIH SBIR grant R43DK121668 to B.H.J., U.S. Army grant W81XWH2110235 to P.T.H., NIH grant R01CA235375 to P.T.H., and NIH award F30DK130564 to A.G.B. Core facilities at MIT were supported, in part, by the Koch Institute Support (core) Grant P30CA14051 from the National Cancer Institute. A.G.B. acknowledges additional support from NIH grants T32GM007753 and T32GM144273 and the MIT Termeer Fellowship of Medical Engineering and Science. A.V. received funding from the National Rongxiang Xu Foundation. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

AUTHOR DISCLOSURE AND GHOSTWRITING

No ghost writers were used for this article.

ABOUT THE AUTHORS

Aristidis Veves, MD, DSc, is professor of surgery at Harvard and is the chairman of the Rongxiang Xu Center for Regenerative Therapeutics. Brian H. Johnston, PhD, is the founder and CEO of SomaGenics, Inc., Brandon J. Sumpio, MD, is a vascular surgeon trainee. Paula T. Hammond, PhD, is the Institute Professor and Vice Provost for Faculty at MIT.

Abbreviations and Acronyms

DM

diabetes mellitus

DS

dextran sulfate

H&E

hematoxylin and eosin

HIF-1a

hypoxia-inducible factor-1 alpha

IL

interleukin

LbL

layer-by-layer

LPEI

linear polyethylene imine

sshRNA

short synthetic hairpin RNA

antimiR

antisense oligonucleotide

miRNA

microRNA

PHD2

prolyl hydroxylase domain-containing protein 2