Lipid nanoparticles silence tumor necrosis factor α to improve wound healing in diabetic mice

Abstract Diabetes mellitus is a mounting concern in the United States, as are the mortality and morbidity that result from its complications. Of particular concern, diabetes patients frequently suffer from impaired wound healing and resultant nonhealing diabetic foot ulcers. These ulcers overproduce tumor necrosis factor α (TNFα), which reduces wound bed cell migration and proliferation while encouraging apoptosis. Herein, we describe the use of siRNA‐loaded lipid nanoparticles (LNPs) as a potential wound treatment to combat an overzealous immune response and facilitate wound closure. LNPs were formulated with an ionizable, degradable lipidoid and siRNA specific for TNFα. Topical application of nanoparticles reduced TNFα mRNA expression in the wound by 40–55% in diabetic and nondiabetic mice. In diabetic mice, this TNFα knockdown accelerated wound healing compared to untreated controls. Together, these results serve as proof‐of‐concept that RNA interference therapy using LNPs can reduce the severity and duration of chronic diabetic wounds.


| INTRODUCTION
Diabetes is becoming a major crisis for the health-care industry. As the population ages, the number of those affected is rising, with 50% of Americans now suffering from prediabetes or diabetes. 1 Approximately, 90% of diabetic people suffer from Type II diabetes, with many individuals unaware that they have it. This delays early treatment and intervention, worsening the chance of developing a constellation of negative complications. 2 These complications include ischemia, peripheral neuropathy, atherosclerosis, kidney failure, and impaired wound healing. This study addresses the latter, as delayed wound healing can result in the formation of a chronic foot ulcer in up to 25% of diabetics. Unfortunately, chronic foot ulcers can lead to lower limb amputation, 3 which does not address wound pathology and has a 3-year survival rate of only 50%. As such, there is an imperative need for a treatment that addresses the inflammatory pathology of the diabetic wound bed in a way that prevents the formation of the ulcer. 4 Unlike acute wounds, diabetic foot ulcers do not proceed normally through the three general stages of wound healing: inflammation, re-epithelization, and wound remodeling. 5 Instead, diabetic wound bed cells undergo harmful phenotypic changes that impair their ability to react appropriately to the normal cytokine and growth factor cascade. 6,7 In addition, higher than normal numbers of inflammatory macrophages have been shown to reside within diabetic wounds, where they overproduce tumor necrosis factor alpha (TNFα), one of the most important inflammatory cytokines. [8][9][10][11] In normal wound healing, a small, tightly controlled amount of TNFα is required for fibroblast migration, proliferation, and remodeling. 12 When overproduced, however, TNFα damages the wound further by upregulating cellular apoptosis, reactive oxygen species production, and matrix degradation. 9,10,[13][14][15] Currently, the most common method of treatment for diabetic wounds is the use of moisture-retentive bandages, hyperbaric oxygen therapy, and antibiotics. 16,17 These treatments are often ineffective, as they do not address the chronic inflammation and other biological irregularities that caused the ulcer to form. Biomaterials research has focused on correcting these irregularities with an array of nanoparticles, hydrogels, nanofibrous meshes, and dressings that deliver drugs (e.g., growth factors) to the wound. 18 Alternatively, it may be possible to treat diabetic wounds using short interfering RNA, which can reduce the expression of problematic proteins. 19,20 Although RNA interference is a promising therapeutic strategy, its use in wound healing has been limited. There have been several reports of improved wound healing outcomes upon gene silencing of matrix metalloproteinases (MMPs) [21][22][23] and prolyl hydroxylase domain protein 2. 24,25 In separate studies, siRNA loaded in hydrogel dressings has induced modest knockdown of other gene targets, including xanthine dehydrogenase 26 and the tumor suppressor gene, p53, 27 in the diabetic wound.
In this study, we use RNA interference to examine how reduction in the inflammatory cytokine, TNFα, influences the wound healing process. Sufferers of other inflammatory diseases, like inflammatory bowel disease and rheumatoid arthritis, have benefited from treatment with antagonists of TNFα. [28][29][30] Diabetic wounds have improved with systemic anti-TNFα therapies 31 and antibody-based neutralization 32 but with an increased risk of infection and severe liver damage. 30,[32][33][34][35] Local downregulation of TNFα has the potential to correct chronic inflammation in the wound while avoiding the negative side effects associated with systemic suppression of TNFα.
Delivery of siRNA, whether systemically or topically, requires a vehicle that protects the RNA cargo and enables its transport across cell and endosomal membranes into the cytoplasm of target cells. 19 Fortunately, we have previously developed LNPs that potently deliver siRNA in vivo [36][37][38] and control inflammatory feedback loops via TNFα silencing in a macrophage-fibroblast co-culture model. 39 Herein, we show that LNPs can be topically delivered in solution to nondiabetic and diabetic mouse wounds to silence TNFα and improve wound healing outcomes.

| Nanoparticle formulation
Nanoparticles were formulated with the three-tailed version of lipidoid 306O 13 , 36 which was synthesized via the Michael addition of 3,3 0 -diaminodipropylamine (Acros Organics) to tridecyl acrylate (Pfaltz and Bauer) at a stoichiometric ratio of 1:3 as described previously. 37,39 The lipidoid was then purified over a silica column on a Teledyne ISCO chromatography system to isolate the three-tailed

| Nanoparticle characterization
Nanoparticles were diluted to a final siRNA concentration of 1 μg/mL in PBS. Percent siRNA entrapment was determined via the Quant-iT Ribogreen assay (Invitrogen) according to manufacturer instructions.
Nanoparticle size was measured with a Malvern Zetasizer Nano (Malvern Instruments, UK).

| Animal studies
All mouse experiments were approved by the Institutional Animal To calculate appropriate topical dosing, we referred to experiments conducted in 24 well plates in a previous study. 39  Administration was performed while the mouse was anesthetized with isoflurane, and the liquid was allowed to sit on the wound for up to 10 min without the mouse moving. By the time the mouse was removed from the isoflurane nose cone and allowed to regain consciousness (a process that took an additional 5 min), the liquid was almost entirely absorbed.

| Confocal microscopy
Mice were wounded and immediately received 10 μL of either PBS or 5 μM nanoparticle solution. The nanoparticles used in these studies were loaded with Cy5.5-labeled siRNA. Mice were sacrificed, and wounds were excised 2 hr after treatment with surgical scissors and immediately fixed in 4% formaldehyde solution. After a period of overnight fixation at 4 C, they were washed with PBS, permeabilized with 0.1% Triton-X100, and incubated for 2 hr with staining solutions.

| Wound area analysis
Mice were photographed in the same position next to the same object of known length each day of the experiment. Images were then processed using ImageJ to calculate wound area and the change in wound area for each day.

| Statistical analysis
All mean values are expressed as AE standard deviation. Unpaired Student's t-tests and one-way ANOVA tests were used where appropriate to evaluate statistical significance. A p <.05 was considered significant.

| Lipidoid nanoparticles reduce TNFα gene expression in double-wounded nondiabetic mice
Previously, we showed that LNPs effectively delivered siRNA and induced TNFα gene silencing in an in vitro co-culture wound model. 39 These experiments motivated the present study, in which we assessed the effect of TNFα gene silencing on the wound healing process in mice.
Although healthy animals do not require assistance in wound healing, we

| TNFα silencing accelerates wound healing in diabetic mice
After showing that TNFα knockdown was possible, we sought to treat diabetic mouse wounds with siRNA-loaded LNPs. These experiments used BKS.Cg-Dock7 m +/+ Lepr db /J mice, which are homozygous for the diabetes spontaneous gene mutation, Lepr db . These mice become

| DISCUSSION
Although siRNA is promising as a therapeutic, its use is often complicated by its large molecular size, negative charge, and the need for delivery vehicles to escape the endosome in sufficient numbers. 19 For these reasons, siRNA has been utilized only sporadically in literature studies of diabetic wound treatment and is usually delivered with polymer nanoparticles, nanofibrous meshes, or hydrogels. [21][22][23][24][25][26][27] The most common gene targets include members of the matrix metalloproteinase family, as these enzymes prevent healthy tissue reconstruction when upregulated in diabetic ulcers.
Although this is an important avenue of treatment, addressing the chronic inflammation endemic to diabetic foot ulcers represents another promising method.
Other inflammatory disease models, including plaque psoriasis and arthritis, have utilized polymer carriers for siTNFα to reduce chronic inflammation to good effect. 46,47 Some diabetic mouse studies have also used systemic anti-TNFα treatments, like neutralizing antibodies, 9,32,48 to hasten wound healing. To date, however, none have examined the effect of a topically applied compound intended to downregulate TNFα. In this regard, the specificity and local activity of siRNA makes it an ideal choice.
LNPs are among the most efficacious of existing RNA delivery systems. [49][50][51] Nanoparticles formulated with lipidoid, a type of ionizable lipid-like material, have been shown to potently deliver siRNA in vivo to several cell types, including hepatocytes, epithelial cells, and difficult to transfect cell lines, like immune cells. 38,39,[52][53][54] To transfect wound bed cells, lipidoid nanoparticles must remain in the wound tissue without being degraded by enzymes or becoming stuck in wound debris. Confocal images show that they remained in the tissue, most likely conforming to wound bed topology, and avoided degradation long enough to be taken up by cells (Figure 2). Topical application of lipidoid nanoparticles resulted in approximately 50% TNFα gene silencing within the diabetic wound ( Figure 3c). This knockdown brought TNFα expression almost down to baseline levels typical of a normoglycemic mouse. It has been shown that TNFα mRNA levels in wounded diabetic mice can be up to three times as high as wounded normoglycemic animals. 48 Our data show that reduction in TNFα levels curbs the inflammatory response in the wound and reduces wound area and healing time (Figures 3b   and 4). Treated wounds healed almost completely by Day 12,while untreated wounds were open as of Day 16. The untreated wounds in two of the five mice were so ulcerated that they opened again on Day 11, necessitating the sacrifice of those mice. Even for those mice, the treated wounds were healing faster than the PBS-treated wounds and had not reached the same level of ulceration.
Although it was beyond the scope of this study, we anticipate that siRNA therapy would be best used in combination with a moisture retentive dressing. These dressings are known to accelerate wound healing in general when compared to a dry or uncovered wound bed. 17 In addition to improved healing due to moisture in the wound, a dressing may also facilitate repeat applications of siRNA solution by softening the tissue and promoting solution uptake. The present study was limited to two applications of siRNA on Days 2 and 3, as scab formation prevented absorption of the treatment solution.
One of the challenges of studying inflammation in wound healing is the heterogeneity of the wound environment. The use of doublewounded mice combats this heterogeneity by providing a built-in control that accounts for mouse-to-mouse variation in inflammatory cytokine levels and the overall wound healing response. Our data indicate that TNFα expression was more similar among diabetic mice than among nondiabetic mice, differing by less than a factor of 2, rather than a factor of 3. Nanoparticle-treated wounds in nondiabetic mice also did not heal significantly faster than control wounds, at least in the short-term RNA interference study. This was not surprising, however, because nondiabetic mice do not suffer from an overproduction of TNFα that impedes healing. While normoglycemic mouse wounds halve in size by about 2-3 days, diabetic mice may require weeks to reach the same point. 55  Together, these data highlight the potential of siTNFα-loaded LNPs as an alternative therapeutic to address chronic inflammation, one of the major biological irregularities endemic to the diabetic wound. FIGURE 4 Treatment of diabetic mouse wounds with siRNA-loaded lipid nanoparticles enabled complete wound healing within 13 days. Mice were wounded on Day 1 and dosed with either PBS (black circles, control) or LNPs containing siRNA specific against TNFα (red squares) on Days 2 and 3. LNP treatment accelerated wound healing over a period of 2 weeks and closed wounds by Day 13 in three of the five mice. The remaining two mice required sacrifice prior to complete treated wound closure. Error bars represent s.d. significance is compared to PBS treated wounds on each day (n = 5, *p < .05, **p < .01, ***p < .001)